PRE2017 4 Groep1

Wiki Report

Group members

Thomas Boot	0988095	Industrial Engineering
Jelte Dirks	0908196	Computer Science and Engineering
Maurits van Riezen	1050246	Software Science
Linh Tran	0936651	Electrical Engineering
Roan Weterings	0888129	Psychology and Technology

Introduction

This wiki summarizes the project we have been working on for the past weeks. For the Course 'Robots Everywhere' we have been asked to develop a new application concerning robotics. In the following wiki, the entire process for developing this new application will be given. Below, the progress of the project is represented. The chronological work document (what we did each week) is given as well. The first step towards this development plan, was brainstorming for several options.

The Final Gantt chart can be seen below:

File:OGO Robots Everywhere Gantt V6.pdf

Brainstorm

After initializing the project, the group has come together to discuss several options regarding robotics. They are listed below:

Festival	A smartwear bag with necessities for First Aid and Organisation personnel. Drone will automatically bring new supplies when necessary.
Extreme Sports/Exploration	Smartwear monitors health, etc. Drone will fly in to bring First aid equipment for self-help, location data will be used to send First Aid personnel if necessary.
Cameraman	Smartwear monitors heartrate of many or all people at an event. A location with the highest average heartrate has the most exciting event. The drone will fly to the most exciting event to film footage.
Police aid	Drone can fly around for easy patrolling, the drone can be sent to a specific location as a scout, the drone has an easier time chasing someone.
Shock band (unethical)	People who leave trash anywhere but a recycling bin will get a small shock. The band has an NFC chip for payments within the event so people will wear and use it, eliminating the hassle with plastic chips and coins as well.

After some discussion, the subject that seemed most interesting to tackle was that of Extreme Sports (highlighted in bold).

Preliminary Research

After defining the subject of the project, research has been done regarding the content of the subject. Before actually developing a new application for extreme sporters, it should be researched whether or not there is an actually issue to solve. Without an actual benefit for the user, the application would have little use. Below, our problem statement and objectives for our application are given. Furthermore, this decision is supported with statistics.

Problem Statement

Extreme sporters find themselves in dangerous situations and are difficult to reach whenever they are in danger. What they need, especially mountaineers, is a method to keep track of their health and alert the rescue posts when necessary.

Objectives

In order to solve the problem at hand, we will design a SmartSuit. The combination of a drone and smartwear will monitor the health of mountaineers, warn them in time of potential dangers, and send help when necessary. In order to achieve this, the SmartSuit should comply to the following things:

• Provide preventive and qualitative warnings regarding bodily functions. The SmartSuit should be equipped with sensors which monitor all relevant bodily functions.
• Dispatch a drone whenever the user is in critical condition. This drone can provide preliminary care and can guide the rescue posts to the patient in a swifter manner than ever before.

For this application, the SmartSuit will be focused on mountaineers. However, in future research, the use of this suit might be extended to other extreme sports.

Requirements

The SmartSuit should be able to fulfill following requirements:

• The SmartSuit should be equipped with sensors and monitor the user's vital functions. These sensor readings should be able to alert the user when in risk/ danger.
• The SmartSuit must be able to at least monitor the vital functions: breathing, blood circulation and temperature.
• The SmartSuit should be able to send preventive warnings by triangulating the sensor data and thus estimating the time before arriving in critical condition (using an algorithm).
• The SmartSuit should be able to "communicate" with the user, i.e. send out a warning when approaching critical conditions.
• The SmartSuit needs a location tracker (GPS), in order for the rescue squad and drone to track the patient's whereabouts. The drone needs a GPS, microphone, speaker, and camera in order to be eyes on site for the medical personnel.

Users

Main users: mountaineers (may be extended to hikers, rock climbers, etc. in future research)

Statistics

In order to make sure this application would be an actually benefit for the current mountaineers, we have done some research. This research has shown the number of extreme sporters (namely mountaineers) and the number of actual accidents. Both these findings supported our assumption that a SmartSuit could be a valuable product.

Mountaineers/Rock-climbers

• 47.2 million people have gone trekking (hiking/backpacking) in the united states alone from spring 2016 to spring 2017. (SFIA)

• “Around 211,000 people (aged 14+, living in England) go climbing or hill walking at least once a month and 84,000 take part at least once a week.” (Gardner, 2015)

• “25 million people are climbing regularly” (International Federation of Sport Climbing)

• “Mountain biking has an estimated 8.6 million participants, making it the second-most popular extreme sport” (Jones, 2008)

• 2.53 million people have gone climbing (traditional/ice/mountaineering) in 2017 in the United States alone. (SFIA)

• The Royal Dutch Climbing and Mountaineering Federation (NKBV) has +- 59.000 members in 2015 and are growing every year. (NKBV, 2015)
  • 62% male vs 38% female
  • 14% avid climbers, 31% alpinist, 53% mountaineers, 2% other activities

Injuries and causes

It is very difficult to provide actual statistics and data regarding the number of accidents in general. No actual worldwide database with such data can be found. However, we did find statistics regarding injuries in the US. These data are fairly outdated (up till 2006). However, they can be used as a guideline. By this we mean that for the last few decades, the number of accidents has been fluctuating around 100 - 150, by which one could assume these number can be approximated for 2017 as well.

The statistics show that every year, approximately 100 to 150 injuries related to mountaineering occur in the United States alone.

Figure 1: US Mountaineering accidents compared to population overview

From these accidents, fractures, lacerations and bruises are the most common type of accident. However, these can hardly be prevented. Our SmartSuit can be an asset here, in such way that the user can directly communicate with the drone/ rescue post. However, preventive warnings are of little use here.

If you look one step further, you notice that the most common accident caused by external factors are Hypothermia, Frostbite and mountain sickness. By monitoring vital functions, our SmartSuit could provide preventive warnings and reduce the risk of accidents.

In short, the SmartSuit can reduce the number of accidents by providing warnings for all externally-induced causes (such as Hypothermia). For the other causes, such as fractures, the suit can be used to monitor the bodily functions while help is on the way (e.g. oxygen level) while also provide communication between the user, rescue post and drone. Lastly, using the GPS tracker in the suit, the rescue post can locate the person in question and reduce time needed to provide help.

Figure 2: US Mountaineering accidents by type of injury

(Abegg, 2006)

Based on these findings, it has shown that a large amount of people practice extreme sporters (especially mountaineers). Furthermore, they experience a lack of proper healthcare. By extension, we believe our SmartSuit provides an actual benefit for the user, a solution to an actual issue.

User, Society and Enterprise Aspects

In the following aspect, the USE aspects regarding this application are closely considered.

User Needs Research

The following research forms the foundations for our assumptions. Do notice that we have provided an overview of all research conducted. This does not mean all research has been applied in our project.

Extreme sports in Extreme conditions

Dr. M. Malashenkova (2016), exercise physiologist, has given a definition of extreme sports: “The definition “Extreme” in relation to sport is performed in a hazardous environment and involves great risk. In the modern world of extreme sports, a number of factors require an athlete to have maximum concentration, cope with the stress and physical and emotional mobilisation capabilities. Common to all of these sports are risk-taking, pushing limits (physical and legal) and having fun.” As extreme sports, she recognises: “trekking, paragliding, rock climbing, mountain bike, snorkeling, hot air ballooning, hand gliding, wind surfing, canoeing, sailing, skydiving, surfing, bungee jumping, scuba diving, snowboarding, and skiing” (Malashenkova, 2016). These sports are practiced in a wide variety of locations and in a wide variety of extreme natural conditions to do with “hypoxia, altitude, speed, atmospheric pressure, wind, and temperature”. Some people are capable of adapting to these extreme conditions by increasing functional reserves, though it is unclear if everyone can adapt to such extremes. When conducting research in this area, special attention should be given to safety, medical monitoring, and psychological testing of participants (Malashenkova, 2016).

Sports in extreme conditions: the impact of exercise in cold temperatures on asthma and bronchial hyper-responsiveness in athletes (only abstract available)

Athletes performing outdoor endurance winter sports frequently report exercise-induced asthma (EIA) and bronchial hyperresponsiveness (BHR). EIA is likely caused by the increase in breathing rate; water and heat loss are elevated, and in combination with the increased breathing rate can lead to inflammation of the airways. This can lead to increased parasympathetic nervous activity, likely leading to BHR. Sporters in these conditions ought to be regularly assessed in terms of lung function and BHR. These conditions can be alleviated or cured with medicinal treatment (Carlsen, 2012).

Extreme sports: Extreme physiology. Exercise‐induced pulmonary oedema (only abstract available)

During an extreme triathlon event in australia, certain participants were afflicted with dyspnoea (shortness of breath for an abnormal duration), haemoptysis (coughing up blood), and pulmonary oedema (fluid accumulation in the lungs) (Ma and Dutch, 2013).

Sports and extreme conditions. Cardiovascular incidence in long term exertion and extreme temperatures (heat, cold) (only abstract available)

Extreme sports tend to result in a higher body temperature and more sweating, which can result in dehydration and therefore a lower blood volume. This dehydration can also lead to an inability to regulate body temperature leading to thermal stress and injury such as heat stroke. Extended periods of sweating can lead to hyponatremia; decreased sodium concentration in one’s blood, leading to headaches, nausea, balancing issues, confusion, seizures and coma. Chances of thermal stress due to heat (hyperthermia) can be increased by a hot environment, as well as elevated levels of air humidity. Cold temperatures can result in hypothermia and frostbite (Melin and Savourey, 2001).

Emerging Environmental and Weather Challenges in Outdoor Sports

Because of climate change effects seen around the globe, current advice concerning extreme sports in extreme environments may well have become insufficient. plain weather indications no longer allow for an accurate estimation of heat or cold related illnesses and injuries. Several environmental and weather challenges include:

• Heat (treat heat related illnesses by developing cardiorespiratory fitness, using pre-cooling and ingestion of cold air, water or ice, acclimatization, and hydration and salt balance strategies)
• Ultraviolet exposure (skin cancers and sunburn, use sunscreen and UVR (Ultra Violet Ray) protective textiles.)
• Lightning (lethal injuries, use weather reports, taking shelter if necessary)
• Air pollution (deteriorating lung functionality, inflammation, immune system issues, bronchitis, asthma, etc.) (higher fitness level, train away from cars)
• Cold (hypothermia, frostbite, asthma, cardiovascular events, hallucinations, exacerbation through hypoxia, use protective clothing to prevent heat loss, no constricting clothing)
• Altitude (lower or higher pressure, hypoxia, high altitude sicknesses: pulmonary edema, cerebral edema. Use altitude acclimatization.)
• Snow and avalanche (asphyxia, compression, hypercapnia, hypoxia, use education, safety gears)
• Exercise induced asthma and bronchial hyperresponsiveness (use pollen distribution forecasts, antihistamines, immunotherapy, air acclimatisation gear (for cold air))

(Brocherie, Girard and Millet, 2015).

Extreme Sports: Injuries and Medical Coverage

Common injuries sustained from extreme sports include: head injury, wrist injury, fractures, internal injuries, microtrauma to the scrotum, ankle injury, knee injury, overuse injury, stress fractures, ligament and tendon injury, finger injury, concussion, abdominal injury, sunburn, dehydration, hyponatremia, and sleep deprivation. Some new extreme sports even include marathons on the south pole, or in the desert. Protective gear is advised to help alleviate some risk factors. There is a need for better medical coverage, better design of protective equipment, and assistance in event planning. currently it is difficult to handle injuries during a race, and equally difficult to arrange evacuations, because medical personnel needs the same advanced skills as the participants to reach them (Young, 2002).

Extreme Sports as a Precursor to Environmental Sustainability

Extreme sports gained a reputation for being for risk seeking adrenaline junkies, without much recognition for how extreme sports influence one’s relationship with the natural world. The reason to participate in extreme sports is not as shallow as just the adrenaline rush; they trigger deep personal changes in courage and humility amongst other construct (Brymer and Oades, 2009). The emphasis lies on how the sports change the relationship with nature, and how it is experienced (Brymer, Downey and Gray, 2009).

Performing in extreme sports works as a demonstration of human power, resilience, and robustness, which is done because society makes people feel powerless and insignificant. (Le Breton, 2000; Palmer, 2000). According to people in favour of ecopsychology, activities in nature are beneficial to psychological well being. They help kickstart combating environmental problems because they increase interest in the natural world beyond seeing it as a mere resource. This is because these activities help us recognise and realise we are part of the natural world, which helps people to actually adopt more environmentally sustainable practices (Brymer, Downey and Gray, 2009). This means participating in extreme sports would be beneficial to society, the environment, and individuals; provided it can be done in a safer or more controlled manner.

The extreme sports experience: A research report

Participants of extreme sports tend to report 5 main aspects of and/or reasons for participating; Commitment and skill (high levels of preparation and practice) Defining the boundaries (high risk, limited outcome possibilities) On risk (labeling them as risk or thrill seekers is “missing the point”) Feelings of accomplishment and personal insight (“empowering and making life easier to deal with Extraordinary experiences akin to Maslow’s peak experiences (“altered perceptions of time and space, floating and flying, calm and stillness, and self validation experiences” The conclusion is that extreme sports are not about risk taking, according to participants (Brymer, 2009).

Rock climbing injury rates and associated risk factors in a general climbing population (abstract only)

Self-reported measures research; 4.2 injuries per 1000 climbing hours, most injuries are in the realm of overuse (93%). Inflammatory tissue damages in fingers and wrists occurred most often. Older climbers have a lower risk of re-injuring something, whereas males have the highest chances.

Limits to human performance: elevated risks on high mountains

Many mountaineers are exposed to hypoxia, cold, and dehydration. The higher the mountain the lower the success rate and the higher the chance of death upon descending. The most important limiting factor in long term sustained human inhabitation of a certain location is barometric pressure. Decreased barometric pressure leads to decreased oxygen availability, which leads to physical stress. Hypoxia and dehydration exacerbate influence of cold temperatures. Wind chills the body even faster, which under these conditions makes temperatures drop another 25 degrees centigrade.

User demands/needs for Mountaineering

Based on the research conducted above, as well as the answers from the people who filled out a questionnaire, we defined following user needs.

Extreme sports in general:

• Experiencing raw, awe-inspiring nature
• Proving one’s own skills to oneself
• Acquiring mental health benefits
• Being reached by First Aid in an easier and faster way (Especially important for mountaineering)

Specific needs relevant for Mountaineering/Rock climbing:

For each of the following conditions, we have reasoned what sensors our SmartSuit could use to provide the user with warnings.

• Heights-related conditions:
  • Hypoxia/Hypercapnia (lack of oxygen) → Oxygen level sensor (to measure the amount of oxygen in the blood)
  • Pulmonary oedema (excessive amounts of fluid in lung tissue and spaces), cerebral oedema (excess fluid in brain) → Altimeter/Barometer to monitor speed of ascension (ascending too fast is the main cause)
  • Higher ultraviolet exposure → Sunblock textiles (Merino wool has been chosen after user survey, which also protects against harmful uv exposure)
• Cold-induced injuries:
  • Hypothermia (too low body temperature)/Frostbite → Temperature sensors
  • Cardiovascular events (heart attacks, etc.) → Heart rate monitor
  • Hallucinations → Earlier detection of hypothermia (preventive warnings needed)
  • Exacerbates if combined with hypoxia and/or dehydration and/or wind
• Overexertion & Overuse:
  • Dehydration/Hyponatremia (too little water or too little sodium in blood, respectively) → Galvanic skin response, Blood sensor
  • Inflammatory tissue damages to fingers and wrists → Finger pressure sensors

Other needs

• Appropriate price-quality trade-off for SmartSuit
• Ease of use
• Being warned before any actual harm occurs (whenever this is realistic)
• Privacy (data should not be randomly stored and shared)
• Comfortable design of the SmartSuit, not a burden to wear.

Society's demands/needs:

The following aspects are relevant for society:

• Not paying too much for healthcare (keeping expensive incidents to a minimum)
• Increasing the efficiency of healthcare.
• Advancing technological development in order to make people's lives better.

Enterprise demands/needs:

The following aspects are of importance to companies:

• Clear niche market to target and differentiate in.
• Increasing profit and customer satisfaction.
• Keep competitive advantage, use SotA technology to distinguish oneself.
• Increasing corporate social responsibility. Produce products that increase people’s safety to improve one’s brand image.

Since our SmartSuit is mostly relevant for individual users, the User needs outweigh the Society and Enterprise needs. Only the User needs will remain within the real focus of this particular research. Societal and Enterprise impact are a field for future research.

USE Scenario's

In order to once again demonstrate the real use of a SmartSuit, we have described some scenarios in which mountaineers could find themselves. For each of these scenarios, the use of our SmartSuit is demonstrated. These scenarios have been based on conversations with mountaineers, survey results and logical reasoning.

• A professional mountain climber is climbing a large snowy mountain. Even though he is well prepared, is well aware of the risks and has brought all relevant supplies, he cannot control all variables. When climbing, a blizzard hits, blocking all sight. His heart rate is lowering due to the cold, and in response he tries to hurry (in order to find shelter). In doing so, he overapplies tension to his fingers (severing nerves, which he is not aware of due to the cold). The tension sensors can sense this, and in combination with a lowered heart rate and decrease in skin temperature can deduce the fact that the hiker is likely in danger. It emits a distress signal to the nearby rescue post. A drone is dispatched. Using GPS tracking location, the drone can locate the hiker and guide the rescue squad there. The hiker, who in mean time has received a message that he should stop climbing and take some rest, can communicate through the drone. By preventively alarming the rescue post, the hiker can be brought to the hospital and treat his hands. With a bit of luck, he might just regain feeling in his fingers.

• A young athlete is seeking for a new way to get a thrill. He decides to go mountain climbing. However, he is not that experienced yet, and is not fully aware of all dangers. When starting, he rushes up the mountain in a streak of adrenaline. However, he quickly feels dizzy and light-headed. The Oximetric sensor in the SmartSuit detects a drop in oxygen level. In combination with a lowered heart rate, the smart-tech warns the athlete that he is climbing too fast and is experiencing altitude sickness. A drone is dispatched with extra oxygen to get the athlete back on track.

• A hiker is lost in the snowy mountain tops. He starts to panic and realizes he is running out of food supplies, and does not know what to do. He emits a distress beacon that is picked up by the rescue post. The drone scouts the area and finds the shortest path to the hiker. The drone can then either instruct the hiker back to the road or lead the rescue squad to the hiker.

• After a rough day, a hiker goes to sleep. That night, his toes get too cold, and the system detects a danger to frostbite. The tech emits alarms to the hiker. 1) The hiker wakes up and can prevent frostbite in time. 2) The hiker does not wake up and the next morning wakes up to find out his toes are affected by frostbite. Being a professional hiker, he knows that rewarming his toes only increases the risk of tissue damage (since he is still exposed to the cold and refreezing could occur). Luckily, the smart tech already detected a drop in temperature and allerted the medics. The medics get there in time to bring the hiker home and treat his foot before any permanent damage is caused.

Taking several scenarios into account. The use of a SmartSuit (in combination with a drone) could drastically increase healthcare efficiency, and prevent pretty horrible situations. The SotA in this tech does not fully rely on the use of a drone, but is more centered around the diagnostic ability of the suit as well as the preventive warnings this suit emits.

Survey Results

A survey was conducted with mountaineers/rock climbers as the target audience. Only three participants responded, one of which a mountaineering teacher. However, all three participant represented actual mountaineers (or mountaineering teacher)and the all responded in a similar manner. Therefore, we have elected to use their answers as a guideline for our research. We are fully aware three respondents are not reliable, but given the nature of their responses and the lack of other participants, we still have drawn some conclusions from the survey.

The survey itself has not been posted on the wiki, since this would only ensure an information overload, however a short summary of the main results is given. These results have been used to support our assumptions, to validate our USE scenarios and to guide the design of the SmartSuit.

The most often occuring issues/conditions reported were falling, long term muscle injuries, and exhaustion, the latter also being identified as a cause of other injuries. The main symptoms of these issues being chafing wounds, muscle injuries and low blood sugar levels. Clothing hardly gets damaged, though they do prefer to cover up as much as possible to protect against the sun. The daily height to climb was reported as being dependant on fitness according to the teacher. 7 single pitches and 2000m were mentioned as well. The participants would be ok with a suit that monitors their bodily functions, as long as the data is not available to just anyone and they have a say in it. The respondents would like the garment to be a t-shirt, as well as light and comfortable and not too noticeable, long sleeve merino shirts were mentioned as a good option. Average climbing times were reported to be 3, 7, and 8 hours. The participants would prefer the suit to measure heart rate, blood pressure, breaths per minute, body temperature, blood sugar levels, stress hormones, oxygen, and pressure. They prefer notifications to be received through sounds or vibrations, with more exact information being available on their phone. One participant would only like to see the odds that the warning was accurate when they ask for it, the rest would like to receive this anyway. One participant would like the suit to last for 10 hours on a charge, two for 72 hours (which will be used to determine the number of batteries needed to power the SmartSuit). The permissible charging times for the suit were reported to be two hours, six hours and eight hours. All participants would prefer a wired connection between suit components over a bluetooth connection even though this is more fragile, to conserve power. The participants would agree to the suit sending their data to nearby rescue posts, but only if used for alarms, and they want the opportunity to disable this functionality. The teacher likes the suit for “guided trips to high altitude mountains with clients who are not self-sufficient”, and another participant mentioned the suit should add to the currently available gear like a phone or gps watch. They also mention they would like it for solitary extreme sporting, but no as much for when in groups as they would rely on others for monitoring.

Drone

This section briefly discusses the application of a drone in the SmartSuit. Since we quickly discovered we would not have time to research both the drone and the SmartSuit itself, we have kept this part at a minimum. In other words, we have done some preliminary research, yet this should be elaborated upon in future research.

Need/Viability

For the suit to be viable, it is essential that actions can be taken once the suit has detected a critical health problem that needs stante pede attention. When such an issue occurs in a city, or even rural area with decent infrastructure, it is fairly easy for emergency medical services to get to the person in question. In fact, when looking at data from 2015 (Mell et al. 2015), gathered from 485 EMS agencies all over the United States, the average time of an EMS unit to arrive at the location where the 911 call was placed was 7 minutes for urban areas. For rural areas, this time was over 14 minutes. The areas were classified using ZIP code information, with rural having a population of less than 2.500, suburban between 2.500 and 50.000 and urban more than 50.000. 625 of the 1.275.529 instances were marked as outliers and were excluded in the study.

One can imagine that in extreme conditions, EMS is not as readily available as on ground level with proper infrastructure. For this reason, companies or non profit organisations arise that provide these services, but not in a traditional manner as you would expect. For instance, the ‘Himalayan Medics’ provide courses and training, besides the regular ‘search & rescue’ services. It is when trained professionals are with you on your journey, that actions can be taken stante pede. If no medical professional is present, someone has to be sent up (or down) the mountain to aid the person that needs help. The time it takes for aid to arrive depends on the location, and can possibly take several hours. Take for instance the mount Everest, the shortest climb from camp to camp -on average- is 3 hours, when the climber is acclimatized to the harsh conditions at such altitudes and crossing sections of mountains like the Khumbu Icefall is definitely not speeding up the process. (s.n., 2016)

To reach a 7 minute response time on average, or let us say 15 minute response time on average, with hand-carried emergency services in such extreme conditions is plain impossible. For this reason, there should be an alternative with decent response times that is triggered when the suit detects an anomaly. Besides an apt response time, adequate care is also very if not more important. It is therefore crucial to not only force a fast response time, but provide the care that is needed on the spot. A study from 2010 (Al-Shaqsi, 2010) shows that response time is too often the performance indicator for EMS even though there is no valid evidence for this to be the case, and present evidence is conflicting this statement. One almost seems to forget that the patient’s health is the goal in this scenario and ought to be used as a performance indicator.

To achieve both a respectable response time combined with adequate medical aid, the suit is linked to a drone station (which will be located next to the rescue post) that can dispatch drones when the suit says it is time for medical aid. Drones, as they are airborne, do not care for rugged terrain or icy slopes, and the 600 meters it takes to go from camp A to camp B which takes climbers hours of time, can be overcome by the drone in a matter of seconds. Not unlike an ambulance which has many tools that are specialized for different parts of the body like the brain, heart, muscles, skin, lungs, etc., drones exist in various sizes that can be specialized for various tasks as well. For certain emergencies, it is best to deploy a small drone that provides quick care in the form of a pill or injection. In such an instance, the size of the drone is rather insignificant as long as it can carry said pill or injection, yet speed is essential in this instance. Emergencies that require more equipment can certainly be facilitated with drones of a large size. In this situation, speed is also of importance and certainly realizable.

Size

The sizes of drones vary from insect-size to large drones up to the size of a small aircraft. Very small drones (up to 50 cm) (Flynt, 2017) can be used as small surveillance for the hiker, or an emergency device (wearable size) to be employed whenever distress/danger is present. This emergency drone could be used to alarm other people for potential danger, or find the nearest person that can help the hiker in need of help. Small drones (up to 2 m) (Flynt, 2017) can be used to send in medicins/water/first-aid kit, etc. It will be employed from a first-aid post to the corresponding hiker. As the somewhat bigger drones are known to be quite fast, with the current fastest drone to reach up to 314.14 km/h (L, 2018), it will be able to be on time and also take the speed of a first respondent in the neighbourhood to be a guide. Obviously, this top speed is not needed, which makes it even more plausible to have drones that can guide humans by foot or car. Medium sized drones are usually carried by two people. These drones are not very useful in this project. Large drones go up to the size of a small airplane. A human-carrying drone already exist, but is not able to pick up a person by itself. A person in great danger could be evacuated by this drone, but a person who is heavily injured cannot take the benefit of it. This would be a great follow-up study to do.

Range

The range of a drone is also very important. It can vary from 5 km to 650 km of range (Flynt, 2017). Depending on the use, drones can be specified for its range as well. Typically, bigger drones have a bigger range. However, it is not always useful to employ big drones for a certain task. A trade-off has to be considered. If you want to help someone who only needs some medications but is too far away from a first-aid post for a smaller drone, it would be common sense to send a bigger drone with this larger range span. However, this is not cost-efficient, so a smart implementation would be to, for example send these medications way before (s)he actually needs this. This would have to be sensed in advance with the help of the different sensors on the SmartSuit.

Deployment (Connection with SmartSuit)

Considering the fact that mountaineers expose themselves to extreme conditions, often difficult to reach for humans, drones can be deployed when approaching life-threatening conditions (Examples such as dehydration situations come to mind). In any way, the drone would have to be alarmed in some way through the device. Considering the assumption that everyone owns a working phone with SIM card, distress signals can be send to the (base) stations which would then employ the drone with the desired outcome. Different options should be possible. The user should be at the highest level of authority at all times. This would mean that the user can always send distress signals themselves, cancel (incorrect) signals, etc. However, when the user is not capable of doing this him/herself (e.g. when the user has fainted), the product should be able to detect this and send a signal autonomously. A working network is here assumed to always be present.

In conclusion, a drone would be a valuable surplus for our SmartSuit. However, once again, we have elected to focus on the suit itself. In future studies, the integration of the drone should be elaborated upon.

SmartSuit Design

The following sections discusses the actual design of our final deliverable: the SmartSuit.

General SmartSuit Research

The research below was conducted to find out needs and possibilities for Smartwear in general. More research was conducted, but the pieces presented below is the most relevant for our final deliverable when it comes to options and necessities.

Sustainable Wearables: Wearable Technology for Enhancing the Quality of Human Life

Wearables are to be separated from portable devices. Up until recently, wearable devices have provided data, and have not been purposed for assisting the user with other tasks or provide services that go beyond merely showing numbers on a screen. Technologies for sustainable wearables as listed in the article are “Human Data Tracking”, “Human Big Data Analyzing”, “IoT” and “Middleware for Wearables”. As technology has advanced, wearable devices have become smaller and increasingly wireless, providing the user with less fatigue and skin trouble, and the possibility of embedding wearables in smart clothes. Due to the rising possibilities in wearables, data is collected on a massive scale. This data should not be collected just for the sake of collecting data, yet analysed and presented to the user in a meaningful and useful way… wearables have to become “aware-ables”. Currently, wearable devices do not provide this functionality and only show the user his/her own data, aggregated and crammed into a nice diagram or graph. This is where the Internet of Things comes in. IoT, making use of wireless connections, is capable of analysing multiple data sets simultaneously, cross referencing between different inputs, and provide the user with meaningful information that the user can apply to his/her life. Making use of cloud processing, data can be collected at one specific location and processed remotely, before being sent back to the user to be displayed. There are several factors that improve wearable quality of life, prolonged and continuous usage:

• un-monopolizing
• unrestrictive
• observable
• attentive
• communicative
• cost effective
• low power
• durable
• scalable

(Lee, Kim, Ryoo and Shin, 2016)

Relevance for our objective: A clear indication to what is wrong with the current wearable technologies that are being put forward, with a specific problem description and solution. This review and it’s sources provide meaningful guidelines to develop wearable tech that actually provides the user with useful information.

Flexible and stretchable electronics for wearable healthcare

The market for wearable electronics is growing rapidly, which requires changes in how electronics can be worn around the body. There are several technologies that provide a solution. OLAE, or organic and large area electronics uses thin foils made of PET or PEN onto which conductive “ink” is printed. These foils with printed “ink” form a PCB of no more than 50μm. Layering these separate PCB can produce complex but flexible PCBs. Although, more complex wearable devices might need even thinner chips. Another method of creating these small circuit boards is printing thin-film metals like copper, gold or platinum on polyimide layers. The metallization can be as thin as 1 μm, resulting in the device having a total thickness of about 70 μm (van den Brand et al., 2014).

Not only the thickness and flexibility is an issue when it comes to wearables. Flexibility is another key factor in comfortable wearable design. To provide this attribute, the meander strategy can be applied. Electronic functionality is hereby distributed onto several islands that are connected by meander-shaped interconnects, and embedded onto a stretchable rubber. The meander shaped connectives can stretch -to a certain extent- and provide flexible circuitry options (van den Brand et al., 2014).

In summary: with the ultra-thin processing units that are stretchable -to a certain extent- wearable are able to make real time computations.

Relevance for our objectives: Makes it very easy to integrate processing units into clothes that are comfortable to wear. Wearable sensor can make use of on body computations to provide real time processed information that is of actual use to the wearer.

Sensors Research

The first step of designing the SmartSuit is to analyse a variety of possibilities. Below, the most interesting options for sensors have been given.

Do notice that not all of these sensors have been used in the final design. A selection has been made based on research, survey results and what conditions the sensors could measure.

Non-Invasive Electromagnetic Skin Patch Sensor to Measure Intracranial Fluid–Volume Shifts

Elevated intracranial fluid volume (e.g. a rise in fluids inside your head) can cause intracranial pressure to increase. This is extremely dangerous because this can lead to numerous neurological consequences (i.e. a stroke) or even death. A passive, non-invasive skin patch sensor for the head allows this volume to be measured. The sensor consists of only one baseline component, that is shaped into a rectangular planar spiral. This spiral has a self-resonant frequency response when influenced by external radio frequencies. Any fluid volume change of 10 mL increments can be detected, even through your cranial bone. This has been tested on a dry human skull model, as well as in preliminary human tests. Both have proven successful. In the human tests, two sensors have been used, in order to check the feasibility of using this method in the complex environment that is the human body. For both the dry cranial model and the human tests, the correlation between actual fluid volume changes and the first resonance frequency of the sensor have been determined. Both were high, indicating that the sensor reliably measures any fluid shifts. In short, this electromagnetic resonant sensor might be implemented to prevent strokes, hemorrhages and other neurological consequences (Griffith et al., 2018).

Autonomous smartwatch with flexible sensors for accurate and continuous mapping of skin temperature

Epidermal sensors that are closely contacted with the skin can monitor cardiovascular health, electrophysiology and dermatology with high precision and in a non-invasive manner. This research has proposed a ultra-low power smartwatch connected to flexible solar modules and a row of epidermal heat sensors. This functions wirelessly and energetically autonomous. Preliminary experiments show how this device is perfect for long-term, precise and non-stop monitoring of the skin temperature (Magno, 2016).

Device for generating a detectable signal based upon concentration of at least one substance

This patents proposes a contact device which can be played on the eye, in order to detect physical and chemical parameters in a non-invasive way. Using electromagnetic waves, infrared waves and other, this device can scan the cornea to determine for example the oxygen level in your blood. The blood analysis is performed using eyelid motion and closure of the lid to activate a microminiature radio frequency sensitive transensor. These signals are transmitted to an externally placed receiver, for example on glasses. Some of the parameters that can be monitored are heart rate, respiratory rate, ocular blood flow and blood analysis (Abreu, 2017). This patent is published September 2017, thus SotA.

Rapid rate-estimation for cell phones, smart watches, occupancy, and wearables

Using a phosphor-coated broadband white LED that produces light which may be transmitted with an ambient light to a target (for example your wrist or ear), this patent can monitor your respiratory and metabolic parameters and transmit this data to your mobile device or other wearable devices. The transmitted light is scattered and passes through a spectral filter. Based on the waveband/ wavelength range, the detected light may be analyzed to determine vital body functions (such a body fat, heart rate, respiratory functions, etc.) (Benaron, 2015).

Wearable sensors and systems

Connected health has increasingly become a topic of interest. This refers to the use of sensors to monitor patients health. Hybrid systems integrating wireless and e-textile technologies are becoming the application to go to. For example, movement sensors can be strapped to the patients wrists or chest and gather data, which in turn can be send via GPS to caregivers or relatives. The progress of technology has enabled these sensors to be incorporated into clothing, such that a jogger can monitor its heart rate simply by wearing the appropriate shirt. This can be combined with robotics (rehabilitation robotics for example). The use of sensorized gloves improves the robotic therapy that goes paired with stroke rehabilitation. In short, this paper shows how technology is enough developed to implement sensors in daily devices, and that data collected can be used to drive robotic devices and improve customer service (Bonato, 2010).

Calibration of Minimally Invasive Continuous Glucose Monitoring Sensors: State-of-The-Art and Current Perspectives

350 million people around the world have diabetes. This chronic disorder requires continuous monitoring. Traditionally, this was done by taking a finger prick everytime. Currently, many patients still use this method.

In the recent years, researchers have developed a continuous glucose monitor (CGM), which is able to continuously measure the glucose level in the blood. Also this method is invasive, but does not require the user to give themselves a shot everytime it is needed. The current CGM products need to be replaced after several days, but are able to give the information at any time. This device can be placed in the arm or in the abdomen. It is not bulky and clothes can easily hide it. The CGM measures a current signal generated by the glucose-oxidase reaction, transmitting information on glucose concentration in the interstitial fluid (Acciaroli, Vettoretti, Facchinetti, and Sparacino, 2018). The SotA CGM sensor has some room for improvement in accuracy and reliability. This is due to the fact that the signal only indirectly reflect the glucose concentration. The signal is derived from the glucose oxidase electrochemical reaction.

The SotA CGM have no “smart” aspect. However, attempts have been made by Lee et al, who wanted to personalise the data by capturing the essential cyclic nature by exploiting e.g. data from prior weeks so the calibration time would decrease.

Wearable and Implantable Sensors: The Patient’s Perspective

A study has been done on a target group above 18 years or older regarding their perspective on wearable and implantable sensors. When participants (turned out to be mainly British) were asked if they suffered from any medical condition, the majority mentioned some type of arthritis (52%). The second most common answer given was hypertension (12%), followed by asthma (11%) and diabetes (10%) (Bergmann, Chandaria, McGregor, 2012). Of all responders, 27% had prior knowledge of wearable sensors. However, only 5% have ever experienced with these devices. These experiences related mainly to heart problems (e.g., pacemaker) and diabetes (e.g., insulin pump). Data showed that the responders would prefer a small, discreet and unobtrusive system with many people referring back to everyday objects. The majority (~85%) preferred the sensors to be non-invasive. However, many of this group (~95%) would wear an invasive device when life saving situations come into play. This topic was repeated in the closed-ended section, without fellow-up items and rephrased as implantable sensor. When the participants were asked where they would like to wear the device 85% answered external, 10.5% said internal and 4.5% left it blank. A median annual spend of £50 was found for the biotechnology that related to their own preference. A total of 62% of the people were willing to wear the device for more than 20 h a day. However, 37% expected it to have a battery life of more than 6 months. The placement of the technology on or in the body is expected to take less than 5 min (59% of the overall number of replies) and 35% of the respondents even think it should be less than 1 min (Bergman et al., 2012).

Wearable Sensors for Remote Health Monitoring

Wearable sensors comprise different types of flexible sensors that can be integrated into textile fiber, clothes, and elastic bands or directly attached to the human body. The sensors are capable of measuring physiological signs such as electrocardiogram (ECG), electromyogram (EMG), heart rate (HR), body temperature, electrodermal activity (EDA), arterial oxygen saturation (SpO2), blood pressure (BP) and respiration rate (RR). In addition, micro-electro-mechanical system (MEMS) based miniature motion sensors such as accelerometers, gyroscopes, and magnetic field sensors are widely used for measuring activity related signals. Invasive sensors: rectal thermometer; unsuited for continuous monitoring purposes. Axillary (armpit, thus non-invasive) temperature measurement is more convenient compared to the above-mentioned methods, but more lossy and inaccurate (Majumder, Mondal and Deen, 2017).

Measurement and Geometric Modelling of Human Spine Posture for Medical Rehabilitation Purposes Using a Wearable Monitoring System Based on Inertial Sensors

Inertial sensors have been used to measure spinal motion, making the data intuitive and user-friendly for the clinicians and patients who use the system. The data can be transformed into meaningful parameters such as rotation, flexion-extension and lateral bending. Theobald measured cervical range of motion with inertial sensors. It was proven that they are a viable and objective method for evaluating spine shapes (Voinea, Butnariu and Mogan, 2016).

State-of-the-Art Methods for Skeletal Muscle Glycogen Analysis in Athletes—The Need for Novel Non-Invasive Techniques

Currently, the SotA methods for measuring the muscle glycogen have been mainly invasive by means of needles (Elusive Gold Standard). The latest one has been developed by Bergström and is known to cause as little damage as possible, a high quality in minimal time restraints, can take multiple biopsies from one sample and allows measurement of other outcome variables (e.g. fibre typing, muscle damage, respiration, enzyme activity, etc). There have been no non-invasive techniques, except for histochemical measurement and MRS, developed yet for this problem (Greene, Louis, Korostynska and Mason, 2017).

Novel Wireless-Communicating Textiles Made from Multi-Material and Minimally-Invasive Fibers

Current textile used as clothing are able to sense, react and conduct electricity. The next-generation will be able to perform computational operations, thus getting a dynamical role. Active functionalities in a smart textile may include power generation or storage, human interface elements, bio-sensing devices, radio frequency (RF) emission/reception, various assistive technologies such as personal emergency awareness systems and response communication (Stepan, 2014). This article describes the operation of these textiles and the use of antennas (Gorgutsa et al., 2014).

ePatch

Originally, the ePatch is a concept set out to be researched by students of the “Fontys Hogeschool Techniek & Logistiek”, commissioned by the company Yellow Factory situated in Hilversum. The incentive of the research was that Yellow Factory wanted to produce a product for people who build too much tension in their shoulders and neck, resulting in tension headaches. Their intention is to give the user a patch to wear on his/her shoulder, which detects when the muscle strain is too much and then gives either haptic or visual feedback to the user to relax his/her shoulder. This in order to prevent tension headaches that result from lasting muscle tension in the trapezius muscle (neck & shoulder). In order to measure the tension on the muscles, the ePatch uses an electromyogram (EMG) meter. (Costa, Keulen & van Rijsbergen, 2016))

Sodium sensor

Another important blood parameter to monitor is the patient’s salt level (namely its sodium level). What would be the best fit for such a sensor is the newly developed wearable sweat sensor. This plastic and flexible sensor can be easily worn on the wrist and measures you sweat on a molecular level. This molecular tests are then directly sent to a smartphone, and could in our case be sent to whatever transponder the suit will be equipped with. This sensor has only been developed in 2016 by Ali Javey, of the University of California, and is yet difficult to find on the market (hence no clear price indication).

Unfortunately, after doing some research, this kind of sensor is not useful for this project, since it only measures the sodium level the body expels. No indication is given regarding the level of sodium that resides in the blood. The only other form of sensor capable of doing this is a mouth sensor. However, this would be inconvenient for the user to wear. Keeping this in mind, the salt sensor has not yet been integrated into the smartsuit.

Finger pressure sensor

As some avid climbers may have experienced already, injuries to the hand and fingers are common. This probably sounds logical as climbers put increased amounts of weight on just their hands and fingers for which they are not suited by nature, except experiencing an injury to the hand or fingers is definitely not pleasant, as they are used every day, almost continuously. For extreme climbers, taking good care of hands and fingers is especially important. They might just find themselves in extreme conditions, where once the injury has manifested itself, proper actions can not be taken to prevent the injury from worsening and the mission might even have to be aborted. One of the most common injuries is a tendon pulley injury. To understand this injury, one must be at least acquainted with how the fingers work mechanically.

Figure 3: Finger pulley anatomy

To start off, fingers have no muscles. The mechanic functioning of the fingers is controlled by muscles in the inside of the forearm, that have tendons attached to the middle and end of the finger bone. You can imagine the finger as a brake cable on a bike. In the image above you can roughly see how this looks anatomically. As discussed in a BMC climbing injury sypmosium in 2016 with presenters. Dr. Volker Schoffl (surgeon, climber 8a) and Dr. Isa Schoffl (Paediatrician, climber 7a+), fingers comprise of 52% of all climber injuries, with pulley injuries being 60% of all finger injuries (31,2% of total injuries). This injury is mainly caused by what is known as the “Crimp grip”.

Figure 4: Crimp Grip

The “Crimp grip”, as shown above, is not a natural position for the hand, and places tremendous amounts of force on the pulleys in the fingers, especially on annular pulleys A2 (Jebson & Steyers, 1997) and A4 (Schöffl et al., 2009). This specific grip combined with sudden slipping of the feet causes many injuries. It is therefore that experienced climbers often prefer other grips, due to either a prior injury or just experience.

Treatment of such an injury is important, as is explained above. However, in order to treat this injury, the climber must be aware that this is happening in their fingers. The sooner this injury is treated properly, the sooner it is alleviated. Inexperienced climbers tend to “climb through the pain”, which results in longer recovery time than is necessary had they properly taken care of it when it first appeared.

One of the signs of a ruptured pulley besides pain is pressure inside the finger, resulting in swelling of the fingers (Crowley, 2016). Gloves that warn you when you strain your fingers too much or that pressure is building inside your finger could significantly reduce the time it takes for extreme climbers to notice that they are putting too much stress on their fingers, or that their grip is wrong. Someya Group has researched and built fabric using carbon nanotubes, graphene and an elastic polymer that could act as a pressure sensitive glove. They have tested it on an artificial blood vessel, and were able to detect pressure changes and the speed of pressure propagation (Lee et al., 2016). The sensing range and sensitivity can be adjusted by changing the thickness of the fibre layer, concentration of the conductive filler in the fibre, pad size and width/length ratio of the transistor.

For the suit, this material would be ideal in detecting if fingers have swollen abnormally, and while it is there anyways, detect when climbers are using the wrong grip, or a grip wrong (!), resulting in too much pressure on one of the pulleys of the fingers. This can not only improve recovering times, as early indication significantly reduces recovery time (such is the case for basically every injury), but also help prevent the injury in the first place by detecting too much pressure on certain parts of the finger. Besides pressure on the fingers, blood pressure was also able to be measured, which could provide additional data for measuring the blood pressure besides a regular blood pressure sensor to provide the controlling computational unit with more accurate data.

However, the material should only be used on the hand and fingers of the suit, ideally be worn as some sort of glove that is detachable. This to reduce the cost of the suit, as it would serve no functionality on other surface areas of the body. This would improve the ease of use of the suit, especially when putting it on or taking it off. As the majority of injuries in extreme climbing consist of injuries to the hand and fingers, we believe that such a ‘glove’ would greatly enhance the suit in its entirety.

The Design

Based on the research conducted above, the opinion of the mountaineers we had contact with and our own reasoning, we have defined the following design. Below, an overview is given of all components our design should comply to. The most important parts for the deliverable are the wiki, the presentation, the design itself (including what sensors to use and how to develop the SmartSuit) and the algorithm (which is crucial for the sensor readings to be converted into preventive and meaningful warnings for the user).

Figure 5: Design Plan

Electronics

Now that the research for our SmartSuit has been elaborated upon, the actual choices for the electronics will be discussed.

Processors

Table 1: CPU options

Name:	SimpleLink CC264R2F	ESP32	AMD Sempron Embedded	Kinetis KL28	Arduino Uno/ATmega328
Manufacturer:	Texas Instruments	Espressif Systems	AMD	NXP Qualcomm	Atmel
Link:	http://www.ti.com/product/CC2640R2F/description	https://esp32.net https://www.espressif.com/en/products/hardware/esp32/resources	https://www.amd.com/en-us/products/processors/desktop/sempron-cpu#	https://www.nxp.com/products/processors-and-microcontrollers/arm-based-processors-and-mcus/kinetis-cortex-m-mcus/l-seriesultra-low-powerm0-plus/kinetis-kl8x-72-96-mhz-secure-ultra-low-power-microcontrollers-mcus-based-on-arm-cortex-m0-plus-core:KL8x , https://www.nxp.com/docs/en/data-sheet/MKL28Z512Vxx7.pdf	http://docs-asia.electrocomponents.com/webdocs/0e8b/0900766b80e8ba21.pdf http://ww1.microchip.com/downloads/en/DeviceDoc/doc7799.pdf
Instruction Set:	ARM	Xtensia ISA	x86-64	ARM	RISC
Clock speed:	48 Mhz	240 Mhz	3.2 GHz	96 Mhz	8-16 Mhz
RAM:	2 kb	520kb	Requires external RAM	96 kb	1 kb
Power Consumption	5.9 mA (Active) 1.1 μA (Standby)	20-50mA (Not including Bluetooth)	25w (Not found in amps)	9.6 - 12.8 mA	13-21 mA
Size:	2.7 x 2.7 mm – 7 x 7mm (Larger option has more pins and a lower minimal order size)	7 x 7mm (48 pins, 10-13 GIPO’s)	Not found, likely to be large.	8x8mm – 14x14mm(82 GIPOS)	5 x 5 mm - 7 x 7 mm
Price (Estimate):	€3.57 (up to 50% discount when buying 1000+ units)	€2.62	€207.49	€2.33	€1.82
Features:	Bluetooth, Dedicated subchips for AES and RNG	Bluetooth, wifi, Dedicated subchips for AES, RSA, SHA and RNG	Fast enough to power screen, Will not function below 0 degrees embedded cooling	Dedicated TRNG generator + accelerators for DES, 3DES, AES, MD5, SHA-1 and SHA256.	-
Notes:	-	Currently in a team member's possession.	-	Timer inaccurate in extreme temperatures, flexible input voltage. Very detailed specifications.	Only multiples of 250 available, but an Arduino utilising this chip is currently in a team member's possession.

Based on this table, the Simplelink and ESP32 processors have been chosen. The exact argumentation will follow in the final SmartSuit architecture section.

Sensors

The sensors incorporated in the SmartSuit are a temperature, heartbeat, oximetric, galvanic skin response, GPS and barometric sensor. These have been chosen based on our objective (being able to monitor vital functions such as heart rate, oxygen level and temperature), the most common risks when mountaineering (e.g. for Hypothermia you need to monitor the temperature, heart rate and galvanic skin response of the user) and their affordability. A lot of the sensors we researched were very good, but are fairly unrealistic to incorporate in a suit and keeping that suit affordable. Let alone the fact that those sensors would likely break in harsh weather conditions. Hence the following selection of sensors:

Temperature Sensor

Ideal and State of the Art: The current wearable electronic devices often lack the sensing accuracy due to movement artifacts because of their rigid en bulky size. It is hard to adapt to the soft and deformable skin. They do provide high performance for signal treatment, communication and power management (Magno, 2016). Contrary to those devices, “skin-like” or “epidermal” sensors exist, which have unparalleled accuracy and precision thanks to their thickness and compliant mechanics. The contact with the skin is therefore robust and intimate. The downside of this sensor is that continuous and long time monitoring is currently only possible with wired connections. This would mean that wireless connections only give “on-demand” measurements. Magno et al. (2016) have proposed a wireless transmission and energy harvesting solution working on solar energy. In their conclusion, they found that their setup consumes 2 mW when accurately sampling temperature. This was functional under and self-sustaining under indoor light-conditions, so they proposed another solution with the help of thermal harvesting of the human body itself for further research, to measure the attainability of such a self-sufficient power harvesting method.

Affordable and Realistic for a Prototype: A wireless body temperature sensor system was created, using relatively affordable materials as to make it accessible to financially challenged countries such as Malaysia. It’s main components are an LM35, and uses an XBee and Arduino with WLAN shield to wirelessly connect to the internet. It reportedly has a 0.25 degrees centigrade accuracy when used to measure temperature on a subjects hands. According to the experimenters this is comparable to a commercial thermometer. This type of technology is both wireless and cheap, with an accuracy that could potentially be improved when measuring a more suitable spot instead of hands.

Galvanic Skin Response Sensor

A GSR sensor system was designed to be used for continuous measurements from the back of the body, designed to be flexible and have on-board signal processing. The measurement results showed a high correlation with the utilised reference system of standard fingertip electrodes, and when graphs of the reference signal and the new device’s signal were compared, the characteristics were similar enough to deem the device sensitive enough. The design was made flexible by utilising a “foam-type conductive material” so it would mold itself to the shape of the body it is applied to. The attachment system is designed to be easily attachable and detachable to several types of bodywear. (Kim, Kwon, Seo, and Park, 2014)

Heart Rate Sensor

Ideal and State of the Art: A small scale heart rate monitor was created using a custom made flexible pressure sensor and an analog anti-interference circuit. The sensor’s response was linear, with a high sensitivity of 13.4 /kPa. The interference circuit removes noise caused by body movements, and when also taking the sensor’s high flexibility into account this sensor proves to be a comfortable long term heart-rate monitoring solution. The system was compared against other currently commercially available solutions and proved to have an equivalent capability. (Shu, Li, Wang, Mi, Li, and Ren, 2015)

Affordable and Realistic for a Prototype: AD8232 Single Lead Heart Rate Monitor with electrodes and wires.

Blood Parameters

One option for measuring blood parameters and other parameters, like pulse rate, arterial blood oxygen saturation and volume changes can be done with the optical technique called Photoplethysmography (PGG). It uses a clip which contains a light source and a detector on opposite sides to detect the cardiovascular pulse wave that propagates through the body. The PPG waves can be described as containing a direct current (DC) component due to venous blood and an alternating current (AC) component due to blood volume changes in the arteries. The principle of the estimation of pathophysiological parameters using PGG is based on differential absorption of near-infrared radiation (NIR) by different chromophores present in the blood of capillaries, when the subject’s finger is exposed to NIR radiation.

After conducting extensive research, we have elected to use the current SotA OCM1 sensor from Taiwan Biophotonics Co. This sensor enables a blood oxygen measurement with only a 3% accuracy range and measures the patient’s heart rate with only 4 bpm variation (TBPC, s.d.). The sensor is relatively low-power consuming and can be implemented in the fingertips or wrists of the suit (where it is most accurate). For our application, the fingertips seem the most plausible solution.

Respiration Rate Sensor

As Koch & Dietzel have shown in their research (Koch & Dietzel, 2016), it is entirely possible to produce an array of sensors that can reliably determine respiration rate by measuring body deformations. The sensors in the array are conductive, and deformation of the material changes the conductivity of the material slightly. By placing several sensors in an array, using different directions, a map can be drawn based of the conductivity of each individual sensor. After calibration, this conductivity map can be translated to certain deformation patterns, as the conductivity is determined by the thickness and the deformation radius of the material.

This sensor has not yet been incorporated in the current design. However, this would be extremely useful in a new, more extensive SmartSuit. For that reason, we did include this in the report.

Barometer

Air pressure sensors have developed so far that they can be implemented in smartphones. These sensors measure the atmospheric tides, that are produced by the absorption of solar radiation in water vapor or ozone. The sensors measure this to determine the atmospheric pressure. (Price et al., 2018) There is little holding us back from implementing the same sensors in our smart suit. Sensors at this point are specific enough to even determine what floor a person is walking on. By using sensors that measure changes in atmospheric pressure, the barometer can accurately predict the height of the person (since pressure and height are strongly correlated). (Ichikari et al., 2015) The sensor we will implement in our suit is the BMP280 Barometric Pressure Sensor, manufactured by Bosch Sensortech. (Bosch SensorTech, 2018)This sensor is only a few millimeters large and extremely low-powered (2,7μA power consumption on average). The sensor is accurate up to 0.12 hPa and functions within a range of -40 degrees to 85 degrees (which should suffice for the mountaineering climate).

GPS

Global Positioning System is a method of tracking one’s movement via satelite localization. This space-based system uses the connection between a transmitter (e.g. a smartphone) and a satellite to determine the approximate location and time. This technology proves effective as long as the line of sight to GPS satellites is unobstructed (e.g. not underground). Therefore, GPS tracking is suitable to implement in our suit, as mountains are high and thus clearly detectable by GPS satellites. (Persson, M., Karlsson, D., 2018). The sensor we will implement in our suit is the ZOE-M8B GNSS tracking chip. This sensor, developed by uBlox is precise up to 4 meters and can track movement up to 0,4 m/s and 2 degrees (for the direction) accuracy. It has a special ultra-low power mode called Super-E, which reduces the power consumption to12 mW. (uBlox, 2018) This solves the largest issue of most GPS devices: their massive power consumption.

All sensors have been summarised in table 2.

Sensor details overview

Table 2: Sensor Options

Name/Model:	Website:	Pins (Voltage Range)	Size	Accuracy and/or Precision	Power Consumption	Price	Needs an extra Micro Controller Unit?
Texas Instruments LM35A (Temperature Sensor)	https://www.hackerstore.nl/Artikel/18	V input = 4 V to 30 V Ground V out = 10 mV per degree Centigrade, linear by approximation	4.699 mm × 4.699 mm	0.5°C Ensured Accuracy (at 25°C) typical accuracies of ±¼°C at room temperature and ±¾°C over a full −55°C to 150°C temperature range	60 μA	€1,70	Yes
BMP280 Barometric Pressure Sensor	fhttps://ae-bst.resource.bosch.com/media/_tech/media/datasheets/BST-BMP280-DS001-19.pdf	I2C protocol (8 pin) 8 pins (elaborated upon on page 35 of the data sheet) VDD between 1.71V and 3.6V VDDIO between 1.2 and 3.6V	2.0 mm x 2.5 mm x 0.95 mm	±0.12 hPa range from -40 to 85 degrees	2.74 μA average 4.16 μA max	€2,86	No
AD8232 module (heartbeat sensor)	http://www.analog.com/media/en/technical-documentation/data-sheets/AD8232.pdf	Operating voltage: 3.3V Single supply operation: 2.0 - 3.5V	4 mm x 4 mm	See datasheet	170 μA	From €12,60	No
ZOE-M8B GNSS tracking chip	https://www.u-blox.com/sites/default/files/ZOE-M8B_DataSheet_%28UBX-17035164%29.pdf	I2C Compliant 51 pins: specifics on page 16/17 Table with voltage range on pagina 18	4.5 mm x 4.5 mm x 1.0 mm	In power saving mode: Speed: 0.4 m/s Heading: 2 degrees Location: 4.0 m	12 mW in Super-E mode	€10,90	No
Galvanic skin response sensor	https://ieeexplore.ieee.org/document/6945148/	"Designed to measure skin conductance levels from 0 to 50 µS. A 3 Hz low pass filter is included in the analog signal processing part. The filtered analog signal is followed by analog-to-digital conversion in a microcontroller (MCU). A digitized signal of 10-bit resolution is wirelessly transmitted through a Bluetooth interface." Only communicates via bluetooth, but we can probably strip out the microcontroller	38.5mm x 42.5mm	Figure 6: Galvanic Skin Response sensors Comparison	Designed to measure skin conductance levels from 0 to 50 µS. A 3 Hz low pass filter is included in the analog signal processing part. The filtered analog signal is followed by analog-to-digital conversion in a microcontroller (MCU). A digitized signal of 10-bit resolution is wirelessly transmitted through a Bluetooth interface.	Unreported	Yes (Has a built in one, but it’s not good enough)
OCM1 oximetric sensor	http://www.tbpchc.com/eng/medical.php?gid=13	1,8 to 3,6V, I2C Compatible	8.0mm x 11.4mm x 2.6mm	Heart rate: 3bpm variation for a heartrate between 40 en 240 bpm Oxygen: 3% off for an oxygenpercentage between 70 en 100%	60 mW	Unreported	No
A Pressure Sensing System for Heart Rate Monitoring (SotA, yet not included in the current prototype design)	http://www.mdpi.com/1424-8220/15/2/3224/htm	The device responds linearly to the external pressure with the sensitivity around 13.4 kPa−1. Includes built in processor, built in processor is very big (based on pictures) and we should probably replace it Figure 7: Heartrate sensor input and output	15 mm × 30 mm (size of the sensing part)	The device responds linearly to the external pressure with the sensitivity around 13.4 kPa−1 Table 3: Heartrate sensor and commercial electronic sphygmanometer comparison	Unreported	Artery location (arm, neck, leg)	Yes (built in processor too big)

All sensors described above, expect the last one (too SotA and expensive to incorporate in the prototype), have been implemented in the Hiketech SmartSuit. Based on their qualities, their accuracy and the vital functions they monitor, we believe that a combination of said sensors could provide the user with preventive warnings regarding nearly all mountaineering risks.

Wiring

Another important part of the SmartSuit is the wiring. We have opted to use wires (with bluetooth as backup), as has been shown from the survey as well. However, it is imperative that the wires are sturdy and cannot easily get damaged. We have therefore researched several options.

Currently Available

Adafruit, a company founded in 2005 by MIT graduate Limor Fried with the aim of creating the best online electronics education site and best designed products store for all ages and skill levels. This store also caters to wearable electronics, specifically smart textile solutions (Adafruit Industries, n.d.).

Table 4: Adafruit's Wiring and Smart Textiles Options

Name:	Size:	Price	Resistivity	Notes
Stainless Thin Conductive Thread - 2 ply	23 meters	$5.95	16 Ohms per 0.3 meters Drives components using under 50 mA	Doesn’t oxidise, like stiff sewing thread. (Safe to wash)
Stainless Medium Conductive Thread - 3 ply	18 meters	$6.95	10 Ohms per 0.3 meters Drives components using under 100 mA	Doesn’t oxidise, like stiff sewing thread. (Safe to wash)
Stainless Steel Conductive Ribbon	5 mm x 1 meter	$24.95	2.6 Ohms per 0.3 meters Drives components using under 500 mA	Doesn’t oxidise, thin, strong, smooth, flexible like textile ribbon. (Safe to wash)
Stainless Steel Conductive Ribbon	17 mm x 1 meter	$29.95	1.2 Ohms per 0.3 meters Drives components using under 1000 mA	Doesn’t oxidise, thin, strong, smooth, flexible like textile ribbon. (Safe to wash)
Pressure-Sensitive Conductive Sheet (Velostat/Linqstat)	28 cm x 28 cm x 0.1 mm	$3.95	Volume Resistivity : <500 ohm-cm Surface Resistivity : < 31,000 ohms/sq.cm	Pressure sensitive, squeezing reduces resistance. Heat sealable
Conductive Rubber Cord Stretch Sensor	2 mm diameter, 1 meter long	$9.95	350 Ohms unstretched, resistance increases by stretching	Maximum stretch: 50 - 70% longer than resting length. Takes minutes to revert to original length. Not an exactly linear sensor, resistance may vary.
Conductive Hook & Loop Tape	7.5 cm	$7.50	Hook: 1.8 Ohm per sq, Loop: 1.4 Ohm per sq, 0.8 Ohm through mated closure	Silver coating, can oxidise.
Flex PCB Material - Pyralux	15 cm x 15 cm	$8.95	Unreported	Can be cut with scissors or etched with a PCB etching system. Doesn’t crack as easily as copper tape.
Eeonyx Stretchy Variable Resistance Sensor Fabric	33 cm x 30 cm x 0.5 mm	$24.95	129 KOhm / square cm surface resistivity in rest, halves when stretched	Bidirectionally stretchy. Can be washed.
Knit Jersey Conductive Fabric	20 cm x 20 cm	$9.95	46 Ohms per 0.3 m in stretchy direction, 460 Ohms per 0.3 m in less stretchy direction.	Uses silver yarn, can oxidise. Dry clean.
Knit Conductive Fabric	20 cm x 20 cm	$9.95	1 Ohm per 0.3 m	Uses silver yarn, can oxidise. Dry clean. Bit stretchy, not much.
Woven Conductive Fabric	20 cm x 20 cm	$4.95	1 Ohm per 0.3 m	Users copper & Nickel plated nylon. Dry clean. Does not stretch.

State of the Art

As we were considering wires, it dawned on us that extreme sporters in the mountains are prone to wear and tear, especially when it comes to mountain climbers. It is therefore useful for our wiring to be sturdy enough to deal with these conditions, or possibly self-repair? Luckily, Sun et al. (2014) have researched self-healable conducting wires. In Sun et al. (2014), they have tested wrapping carbon nanotubes (or similarly conducting material like Ag-nanowires) around self-healing polymer. The self-healing polymer amends breakage after approximately 20 seconds, making use of Van der Waals forces without greatly impacting the conductive properties of the wrapped carbon nanotubes. Despite Van der Waals forces being very weak, the number-density range of 1010-1011 cm-2 of the carbon nanotubes produces a high force. Tensile strength after breaking was preserved at around 80% (averaged over aligned carbon nanotube sheet, silver-nanowire network and carbon nanotube network), with 92% as a maximum. The resistance of the different wires has also been tested, with the carbon nanotube sheet being the closest to original resistance at an approximately 120% from original. These wires would be perfect for our suit, and solve the breakage issue.

Since the SotA wires mentioned above are at the moment still not widely implemented, the most affordable option for our SmartSuit would be to use regular wires. However, in future designs, when more sensors and more SotA technology is implemented, these wires should be implemented as well. For now, we still focus on the realistic options, while these SotA wires will definitely be implemented in future SmartSuit (after testing the prototype).

Batteries

The last factor regarding electronics is the choice of battery to support and power all sensors and MCU's. Three options are elaborated upon below.

Lithium-polymer batteries

LiPo batteries are flexible and make use of a dry polymer to which an elektrolyt gel has been added (Replace Direct, 2018). These batteries are only 4mm thick and can be charged entirely within an hour. Some important features are the fact that it is flexible, it has a high cell voltage (3,7V) and it has a high energy and power density.Unfortunately, these batteries are not fit to use in extreme weather, so the only viable option here is to keep them close to the human body. At temperatures above 70 degrees, the battey can expand and get damaged, for example. An example of such batteries is the Apple iPhone 5s accu, with a capacity of 1500mAh and a price of 19,50 euro (123Accu, 2018).

Hermes Lithium-Metal battery

This battery, developed by Solid Energy, is the lightest rechargeable battery on earth.The battery is approximately 6 cm high, 4 cm broad and 6 mm thick. It functions under 3.8V and has a capacity of 3,4 Ah (at 25 degrees). It is flexible and extremely high-density. This battery is currently used in for example drone applications and can withstand temperatures up to 45 degrees, with a minimum of 0 degrees while charging and -20 degrees while discharging. (Solid Energy, s.d.)

GMB Low temperature LiPo battery

This battery uses the same technology as the first described. However, this battery can perform at temperatures from -40 to 45 degrees. It has a capacity of 40 Ah and a nominal voltage of 25.9V. This might be the best solution for our design. (GMB, 2015).

As a result, the GMB LiPo battery has been chosen for our design. We first have computed how many were needed to power the SmartSuit long enough. However, we finally decided to build in one battery per separate MCU. In order to ensure that the entire suit functions optimally, and to ensure the suit is being powered for longer than 72 hours (as the users requested), this choice has been made.

Data transmission

For the final component of the SmartSuit, literature research has been conducted. Since we have elected to elaborate on the functionality of the algorithm, this section remains theoretical.

Short- and Medium-Range Standards for Wireless Communication

Four key short-range, low-power, wireless communication standards, namely Bluetooth (IEEE 802.15.1) (Eliasson et al., 2008), UWB (IEEE 802.15.3), ZigBee (IEEE 802.15.4), and Wi-Fi (IEEE 802.11), predominate in the sensor application domain. They all have different data throughputs and ranges (Lee et al., 2007), as can be seen in Figure 11. Bluetooth has been heavily adopted for body-worn applications; Zigbee or 802.15.4 is used for indoor and outdoor multinode network applications; and UWB supports a variety of applications, ranging from implantable sensors (Yuan et al., 2011) to high-precision geolocation determination (Win et al., 2009). Wi-Fi is usually applied in ambient applications where longer ranges, higher data rates, and immunity to signal attenuation in high blockage environments are required. Table 5 shows all advantages and disadvantages of the four mentioned standards.

Since wireless technology is increasing in the health and sport section, the need for a good battery life is required. Therefore, Bluetooth Smart has been developed. It allows Bluetooth Smart users to connect devices and using significantly less battery life. Since Bluetooth has been integrated in many devices, especially in portable devices, an upgrade to Bluetooth Smart can be done easily.

Another known protocol would be 6LOWPAN driven by low power devices, able to join the cause at low data rates and duty cycles. 6LOWPAN is more like a cloud (Olsson, 2014). It is in need of a wireless internet connection (IPv6) in order to work.

In 2012, the IEEE announced the 802.15.6-2012 protocol for short-range, wireless communications designed for in proximity to or inside a human body with data rates up to 10Mbps. Compliant devices use very low power to minimise the specific absorption rate (SAR) into the body and to increase the battery life. The standard supports quality of service (QoS) and strong security, which include EEG, EKG/ECG, temperature, heart rate, oxygen, and blood pressure.

Taking all available choices into consideration, Bluetooth Smart seems to be the most applicable. Zigbee could also be an implementation.

Table 5: Advantages and disadvantages of the four main communication standards (McGrath et al, 2013).

Figure 8: Key protocols for wireless applications (McGrath et al, 2013).

Communication to the base station

As is widely known, communication to the base station (which in turn communicates with other stations, to the destination) can be done in several ways. One could think of phone calls, text messaging and mobile internet. What is needed to have this communication is a SIM (Subscriber Identification Module) card. As the name suggests, communication to a (base) station is not free. Stations are owned by companies which we pay to make use of. Thus, in the proposal, the user has to have access to the base station using a SIM card. As it is common to already own one (103 customers out of 100 people owning cell phones (EC, 2007), thus likely to have a SIM card in it), our proposal will use this fact to be sure that the wearer of our design has access to the stations.

It is becoming a standard to have a working network everywhere, which means that more stations have to be build to supply the increasing demand. This also means that stations are being built at the most difficult places thought of, just in mountains. Still, this has not been fully realised so alternatives are needed. One could think of Walkies, GPS trackers and satellite phones to communicate with emergency lines. This backup should always be present as well. Our prototype will still work with mobile phone communication, but we keep in mind that not every corner has a working network. For now, we assume that the sporter has active connections at all times, thus the use of SIM cards suffice. When further implemented, alternatives can be programmed and developed using the same codes and algorithms to reach the base station.

Cloud Storage

Nowadays, everything is being stored on some sort of cloud. This cloud storage is increasing, so our design should also consider cloud storage. If the product would come to the market, the enterprise would need a storage place. This is out of the scope and will not be considered any further.

The SmartSuit

Based on all research, all processors, wires and sensors selected, the final visual design for our SmartSuit has been created. For the sake of branding, we have named our design the 'Hiketech SmartSuit'.

Figure 9: Wired Visual Design

Figure 10: Unwired Visual Design

Base Architecture

The basic functionality of the Hiketech SmartSuit is given below:

Figure 11: Base Architecture

• Every sensor is at least 10 cm away from the nearest MCU (microcontroller unit) gets its own MCU. If the nearest MCU has insufficient free pins, it will also get its own MCU.
• All MCUs will store data from its sensors, and transmit only when asked for by the master.
• Any non-sensor subsystems, like the long distance communication with the drone, will also have their own MCU’s.
• We will use the CC264R2F since it is extremely small (2x2mm) and can be attached to pretty much any sensor without significantly increasing size. It also has an extremely low power consumption and built-in bluetooth. Exceptions can be made whenever practical.
• Every MCU gets a ID, starting from 1 to 128. When the main controller loses connection to a MCU, the MCU with the lowest ID still connected to the same sub-network becomes the master of the sub-network.
  • If a MCU has not received a signal from the master before some time out (even signals addressed to other MCU’s), it will conclude the connection to the master must have broken.
  • After signalling to all other sensors on the network that it has taken over, the new master will open a Bluetooth connection to check if the master is actually dead or that a wire is simply broken. If a Bluetooth connection can be made, all data from other sensors on the same network (if any) will be sent to the master.
  • A master or submaster, upon detecting a child-MCU is not responding, will switch on bluetooth for some fixed time to try and attempt a bluetooth connection in case of wire failure.
  • If a child-MCU is already in a sub-network before it loses connection, will not attempt to reconnect with the master of the subnetwork, but rather whatever master has the lowest ID.
• Any sensors with a built-in MCU which is not programmable will be given the highest IDs, as they can not take over the network.
  • These sensors will simply shut down when the connection breaks and no other master is in the same network.
  • We could possibly attach an extra MCU even to sensors with a built-in MCU if we want even more robustness.
• Since we have elected to use wires to connect all MCU's and sensors (which is less power-consuming and more efficient), we should make sure these wires are SotA and are not easily breakable.
• We will make use of I2C connection.

Advantages of this setup:

• We use less pins, so we can use sensors that require a lot of pins (I2C requires only one input and one output pin per node).
  • We use less wires as well, the I2C can run as a single wire to some hub of many sensors, then split.
• A lot of the sensors on our current sensor list have built-in MCU’s that are already I2C compatible, so we can connect them to the network for free.
• Allows transfer of data digitally, so no analog inaccuracies.
• Allows every chip to record data and send it faster in batches, thus freeing up processing power on the main chip to do other computations.
• Allow use of separate power supplies, so we can use the cool ambient power systems for the relevant sensors only.
• The I2C multi-master system allows a different microcontroller to take over as “master” when the main one breaks, so it can still function and record data when any one MCU is functional.
• Since the SimpleLink CC264R2F has Bluetooth, we can switch communication to Bluetooth in case a wire breaks.

Disadvantages of this setup:

• More power consumption, every MCU needs to be powered.
  • If we switch to Bluetooth, even more power needs to be used.
• More work: Every chip has to be programmed separately.
• Since the main controller can only poll one sensor at a time, it may react slower.
• I2C supports a maximum of 128 MCU’s on a network.
• Even the small chip is still a solid, fragile, hot block that makes the suit less comfortable, especially if we add some padding for protection.
• In order to use switch the master node, the I2C connection must be to pins that can be switched to be either input or output. This limits the choice for pins.

Alternative Proposal (might be interesting to investigate in future research):

• Only use I2C for sensors that require too many pins otherwise.
• Connect every sensor via long wires to the master MCU with analog signals.
  • This does cause noise on long analog signals, but should not be much on the scale of the suit.
  • Every sensor needs at least two wires directly to the master MCU, so many wires to areas far from the master MCU that have a lot of sensors.

Reasoning behind processor choice:

For the end-product central chips: ESP32 has been chosen because it has the highest processing-power to cost ratio, while having useful features like Bluetooth and an “ultra low power” mode. Furthermore, it contains 48 pins which should be enough. The main disadvantage is the high peak-power consumption and its size.

For the end-product per-sensor chips: SimpleLink has been chosen. Its main advantage is that it has extremely small variants, allowing the user to easily glue it to sensors without taking significant extra space. The power consumption is also extremely low. It also contains Bluetooth.

For the prototype chips: The ESP32 has been chosen again. This is due to the fact that good development boards are available, and they are cheap. It is also useful to use the same chip twice to ensure maximum code reuse.

Reasoning behind the choice for I2C

There are several reasons why SPI (most evident alternative) was not chosen. The main advantage of I2C is that it is natively supported by the greatest number of our sensors. Moreover, SPI needs four (or three when using microwire) wires while I2C only needs two. It is thus easier to implement compared to SPI. As we want to use as less wires and space for the prototype, fitting everything onto one PCB and processor would be the most ideal case. Furthermore, I2C facilitates low-bandwidth serial communication on the same PCB, which means that more sensors can be soldered to one single PCB, which makes the prototype less bulky. Also, SPI is loosely defined, meaning no standard version is available, thus making firewall cracking easier. The main disadvantage is that it is slow: 400kb/s, over than 10x slower than SPI. This speed needs to be divided over all sensors since they are all connected to the same communication line. What would make SPI a better choice as well as its ability to synchronously send and receive (full duplex mode). Because of these reasons, when actually implementing the design, it could be possible to change to SPI because of its speed, synchronisation and greater range of use. But as has been motivated, I2C has been chosen for the current design.

The Algorithm

Health Warnings

Hyper- and Hypothermia are some of the most common conditions during mountaineering. For that reason, we have elected to focus our preliminary algorithm on this condition. Specifically, sensor readings for the temperature, heart rate and galvanic skin response are needed to make an appropriate estimation of this condition. When taken separately they can give an approximation of someone’s well being, mostly using averages and estimations. However, as everyone is different, for our application it is important to get a more accurate estimation of someone’s body’s state to base our warnings and actions on.

Considering the encountered time-constraints, only the before mentioned most important three sensors were implemented in the prototype. Any other sensors researched below have not been implemented for the purposes of the prototype. (yet are still valuable for the final SmartSuit)

To get more accurate readings, the sensors will start with baseline measurements of the person in rest. This is to determine their average body temperature and galvanic skin response. For body temperature this is done because the range of possible base temperatures ranges on average from 36.1 to 37.2 degrees centigrade (MedLine Plus team, 2018), and a shift of one degree centigrade in either direction is already enough for a warning; if a person’s core temperature drops by 2 degree centigrade they are already experiencing a mild hypothermia (after a degree drop they experience a severe hypothermia and need medical attention). If a person’s core temperature increases by 1 to 1.5 degrees they either have a fever, or a mild case of hyperthermia. If a person’s body temperature increases by 1.5 to 3 degrees they are experiencing serious hyperthermia, and are in need of medical attention (Calvin, 2018).

Figure 12: Accidental hypothermia in severe sepsis (Pathological Society, 2014).

The graph suggests the relationship between inflammation and temperature in severe sepsis. As inflammation increases, temperature accordingly increases. After the point of ‘critical inflammation’ core body temperature drops, this may be an adaptive response to prevent further damage mediated through inflammation.

“Accidental hypothermia” is described as the sudden drop of temperature due to environmental causes. In this figure, one can see that there is a relationship between inflammation and temperature in severe sepsis. At the intersection of the two graphs, “critical inflammation” has been reached, causing hyperthermia to turn into hypothermia. This point can be seen as the point at which the sporter is in great danger and needs medical assistance.

This would mean that a person is “safe” regarding sepsis when (s)he is between 36.1 to 40 degrees Celsius (hyperthermia). When a person hits the 40 degrees and drops down to 32 (hypothermia), one can conclude that the person is under severe sepsis (Pathological Society, 2014).

Even these measures are estimates based on averages however, and need to be combined with more sensor data for a more accurate warning system. People experiencing dehydration, hypothermia, or hyperthermia have a reduced galvanic skin response, because they stop sweating and their skin dries out. Galvanic skin response tends to rise when sweating, as sweat increases conductivity, and tends to drop when the skin dries up. Activities and movements such as sports can also influence galvanic skin response readings, but these can be attributed accordingly when other sensor data is taken into account (Shariff, Hingorani, Albadawi, 2015), such as heart rate data, which during exercise should be between 50 to 85 % of 220-age (Gholipour, 2018) (This heart rate data will also be used to alarm professional help in case of cardiovascular events). Under hypothermia conditions a person’s heart rate will drop, under hyperthermia conditions a person’s heart rate will initially rise, and eventually drop as the person goes into shock.When most or all spikes in galvanic skin response have been attributed to other causes such as physical activity, what is left is the base galvanic skin response. When this is base higher on average than the baseline measurement, this is indicative of a sweaty skin, when this base is lower on average than the baseline measurement, this means the skin is drying out. There is no such data found on the exact values of when things get dangerous, but what could be done is looking at the derivatives of the increase/decrease of the heart hate associated with hyper/hypothermia.

Figure 13: The body temperature change IN VEINS: oesophageal, tympanic, arterial and jugular venous blood temperature. As can be seen, a linear increase/decrease to cool down/heat up the body, trying to keep the body temperature constant. (Nybo, 2012).

f(t) = 0.9e-3*t+T_base

This figure is a comparison of how quickly the body temperature can change in veins. One can calculate that for this particular case, the temperature increase is approximately 0.9e-3 degrees Celsius per second. Recovery has a faster decrease: 0.08 degrees Celsius per second (Nybo, 2012).

Figure 14: Heart rate and body temperature measurements for walking into running into sitting (Iwasaki, 2015).

In this figure, the heart rate and body temperature have been measured for a person who went from walking (at time 0), to running (at time 5 min), to sitting (at time 35). The different curves show different speeds. The blue curve corresponds to running at 6km/h, red to 8km/h and green to 10km/h. Error bars have been included in this figure meaning that several people participated in this research. So note: Two of the seven subjects gave up 15 and 20 min after running started at 10 km/h because of difficulty maintaining the pace. Therefore, seven datasets collected at each time point, except for 20 min at 10 km/h (n= 6) and from 25 min to 34 min at 10 km/h (n=5) (Iwasaki, 2015). For the algorithm, hiking is more of a “walking-pace” intensity level with an increasing intensity when going on slopes or when actually climbing. Therefore, we will compare this intensity with the red curve (also because the green curve is incomplete). For the heart rate, the increase coincides most with a logarithmic function. Thus, when starting to increase the intensity at time = 5 min: HR = HR_base + 25 + log_1.05425(t+1) per minute. As one can see, the body temperature increases exponentially with 1.00054586 per minute or T = T_base*1.000272893^t, where t is in 30 seconds. As not a lot of data points have been taken, the temperature decrease while resting seems to linearly decrease with T = T_base-0.00333t per 30 sec (Iwasaki et al., 2015).

Figure 15: Effects of dehydration (Kleiner Nutrition, 2015)

As can be seen from the figure 15, it is getting dangerous when someone loses more than 2 kg of fluid. This will result in impairing. However, these are all absolute losses, which can actually differ from person to person. Figure 16 and 17 (also) show relative weight losses.

Figure 16: Relative dehydration risks (Tuned into Cycling Team, 2012)

A galvanic skin response detector is able to measure a person’s hydration level. It is expressed in gain (dB). Figure 17 shows simulated sensitivity. One can see that after drinking, the gain will increase accordingly. This is just a proof how galvanic skin responses work. No further useful conclusions with respect to this project can be made with this information.

Figure 17: Simulated galvanic skin response sensor sensitivity (Asogwa, 2016).

Table 6: Concequences of dehydration (Montain, 1992)

"We found that the magnitude of increase in core temperature and heart rate and the decline in stroke volume were directly related to the body weight loss (and thus dehydration accrued) during exercise. Thus, when subjects exercise at 62% to 67% Vo2max under the present environmental conditions (33°C dry bulb, 50% relative humidity, wind speed 2.5 m/sec), the optimal rate of fluid ingestion to attenuate hyperthermia and cardiovascular drift is the rate that most closely matches fluid loss through sweating, at least until the rate of fluid ingestion replaces 81% of sweat loss."' (Montain, 1992).

When the suit detects the onset of hypothermia, the drone can be dispatched to send warmer clothes, heat elements, and perhaps a light-weight tent to serve as shelter and protection. When the situation worsens the person will be warned to not actively heat up their body, and a medical team will be sent to offer professional help. When the situation worsens the person will be warned to no longer ingest any food or drink, and try to cool down actively as much as possible and a team of medical professionnals will be sent. In conclusion, this means there are several options to detect hypothermia. To detect hypothermia, a galvanic skin response sensor, a heart rate sensor, and a temperature sensor will be used.

To combine these sensor readings into a more personal advice, acceptable, risky, and dangerous ranges for each sensor will be compared and combined. When at least two sensors are indicating a risk, a warning will be issued to the wearer, and the wearer will be given the option to decline the dispatch of a drone to send first aid material. When the wearer declines, another warning and drone request will be sent and offered for declination each time another sensor enters the risky range, or a sensor enters the dangerous range. When two or more sensors reach the dangerous range, a request for professional medical attention will be issued and offered for declination.

For the temperature sensor, values in the between baseline - 1 and baseline + 0.5 are ok; values between baseline -1.5 and baseline -1, and baseline +1 and baseline + 1.5 are risky; and values between baseline -2 and baseline -1.5, and baseline +1.5 and +3 are dangerous.

For the galvanic skin response sensor, provided the person is still active which is determined using the heart rate sensor, the ok range is between the midpoint of the baseline measurement and the base of when the person is active and sweating, and the base of when the person is active and sweating. The risky point is between this midpoint and the baseline, and the dangerous range is below the baseline.

Figure 18: Normal and abnormal heart-rate responses (Chaudhry et al., 2017)

This figure shows the relation between heart rate and intensity for normal and abnormal heart rate responses. This curve has to be taken into account when also considering people with a cardiac dysfunction. For this project, when the heart rate increases with time, one can conclude that the intensity of the sport is increasing as well (Chaudhry et al., 2017).

Nes et al. (2013) have developed an equation to calculate the maximum heart rate of a person. This is: HRmax = 211 − (0.64 × age). This is not the only way to calculate this. There are a lot of other equations which have almost similar results as Nes et al. For this project, we have used this one as it is the most general one. If time would have permitted, more accurate equations would have been used which are, e.g, dependent on the gender and age as well.

Height related issues

When people ascend too fast, their blood will expand too fast, sometimes even creating air pockets. This can lead to cerebral and pulmonary oedema, where fluid starts to build up in the brain or lungs. This can be predicted by use of an altimeter to monitor their speed of ascension. This condition is very dangerous, and first aid can’t properly help them. Because of this, when the rate of ascension goes above 305 m a day (Ultimate Kilimanjaro team, 2018), a request for professional help will be offered for declination.

Another problem encounterable at great heights is hypoxia, where the blood oxygen levels are below 60 mm Hg rather than between 75 to 100 mm Hg (Mayo Clinic Staff, 2018). Considering hypoxia can speed up hypothermia, it is important to strictly monitor the blood oxygen levels using the oxygen sensors mentioned in earlier weeks. This too is a condition in which professional help is a necessity.

For the altimeter, the ok range of values is between 0 and 250 m per day. The risky area is between 250 and 285 m per day, the dangerous area is 285 m per day and above, with a warning sent at 250 and 285 m per day. Considering everyone is different, and this indicator uses only one sensor, the user will be able to set the dangerous rate of ascension from the standard 305 m per day to a different level as most people attempting dangerous climbs have practiced and trained often and know their body well.

For the blood oxygen levels, between 80 and 95 mm Hg is the ok range, between 65 and 80 mm Hg is the risky range, and warrants a warning concerning hypoxia and hypothermia. Levels below 65 warrant offering a declination option for professional help.

Not all data is known, so sometimes, only an educated guess is available. However, other tricks can be implemented, like taking the derivative can help “guessing” the problem. For example, heart rates keep changing, this does not mean that when the heart rate increased with 30 beats per minute is necessarily dangerous. There is a factor that is being forgotten which is time. If your heartbeat increases with 30 beats per minute in 10 minutes, this is not a coincidence. However, when this changes within a 10 seconds, one should ring a bell. Therefore, the slope of the increase is a very important characteristic.

Functionality

The proof of concept is divided into 3 main parts: The basic interface, the prediction/regression, and the warning detection. Each of these parts has been placed in a seperate folder. The project can be found at https://gitlab.com/mousetail/social-robots-1-poc-2/blob/master, clone to your disk somewhere and run index.html. There are no dependencies.

The interface (in script.js)

For the interface, we decided to use HTML/Javascript since we had experience using it, and since it allows for rapid prototyping. There is also extensive library support. We decided not to use a framework, since there was no framework both Maurits and Jelte knew and the interface was simple enough that a framework was not needed.

The layout was designed to show everything concisely on one page without scrolling. Every graph got a even share of the screen space. The design is not responsive. Graphs are drawn using the canvas API, specifically the 2d context. It provides easy function for drawing lines in various styles, and we did not have to worry about this. It was also easy to draw the legend and other visual effects.

In the real prototype, sensor data will come from actual sensors. However, since we have no sensors, and also because we want to quickly show how our detection works without having to do complex physical tasks, we needed a way to enter historic sensor data. At first, we tried using sliders and a record button that would store data that came from the sliders. However, it proved to be quite difficult to control precisely, plus the difficulty of moving multiple sliders at the same time. We also tried a system where you could enter a “from” and a “to” input for each sensor value and we would automatically generate noisy data. However this did not allow us to input the more complex quadratic curves some sensors typically showed. Thus, we settled for a system where you could “draw” the historic data straight on the graph. The main advantage is that you can easily draw complex curves, and remember what curves with a certain effect look like. The disadvantage is getting exact values is difficult, and that this would need to be implemented from scratch, since most graphing libraries don’t support this.

The regression (script.js and regression.js)

Regression algorithms can get quite complex. Lucklily, Maurits found a quite advanced Javascript regression library that implemented the algorithms we wanted at the time. We used “Regression.js” by a Tom Alexander. This library provided n-th order sum-of-least-squares regression. We already thought we wanted least-squared, and the ability to decide what order to use later was a nice bonus.

The library turned out to have quite a number of bugs, or undocumented features. One of which was that every function needed a “precision” passed in, which was used to round the output. That this argument even existed was not documented, and everything returned “NaN” when not passed. There were several other issues and the final version we used is heavily modified from the original. The actual algorithm implementation is unchanged.

The data used as input from the regression is data drawn on the left half of the graph. The regression is used to extrapolate a second line that covers the right half of the graph. The X value used is the time relative to “Now” or the center line and when the point was drawn. The Y value is the graph Y transformed into the relevant units for the specific sensor. The results are scaled back before drawing them on the graph.

In order to make the regression fit real-world data better, we used two techniques. When looking at real world temperature data for temperature, we noticed a tendency to “bounce back” after reaching a extreme value, though we found no such evidence for other sensor measures. We decided to simulate this process for all sensors. When a sensor value seems to be approaching a unsafe value at a low velocity, we simulate the bodies homeostasis “pushing” the value back to normal. A significant velocity is needed before we decide the sensor value will reach a extreme value. This correctly modeled the most extreme points, however after the curve “bounced” back, the rest of the curve became inaccurate, resembling a sine wave. In our real world graphs, the body temperature very slowly returned to a normal value after the extreme and never bounced over to the opposite extreme. However, since health warnings are most likely at the extreme point and we only show the first occuring warning, the rest of the graph does not really matter.

A second chance addresses a problem with least-squares regression: That all points are weighted equally. In our application, more recent data says a lot more about the shape the curve is likely to follow than old data. In order to still use out algorithm, we have a preprocessor that removes some values (deterministically) from the data set. The more old the data, the more points are removed. This turned out to have a very subtle effect on the output curve, but definitely an improvement. A much stronger effect is needed to match actual real-world graphs.

Overall our regression works reasonably for body temperature values. Heart rate changes very chaotically and does not even approximately match a parabola at our scales, so our method will fundamentally not work for heart rate. It is very difficult to find data about skin conductivity, but predicting the value is also probably a lot more complex than a quadratic function.

The warning system (warnings.js)

The actual warnings can be found in the section Health Warnings. The implementation of the warnings is in a separate “warnings.js” file. It contains a function that takes a list of values and returns a list of warnings. The main interface repeatedly calls the function for every time stored, using actual data for times that have already happened and predicted values for future dates. It also sorts through the warnings and choses which to display. Since a warning will probably be generated again and again for every moment that a sensor stays inside the danger zone, the interface displays only the first occurrence of a given warning. It will also show only the most severe warning in every category.

The warnings are shown in the following form: “[Time] Warning (severity)”. Times are given in the form “t - x” for past events, and “t + x” for future events. Severity goes from 0 to 2, severity levels have different meanings for different categories of warnings, they only serve to order which they should be displayed and the absolute value is not relevant.

Problems/Future improvements

This is a list of features that would have been nice but we were not able to implement due to time constraints.

• Denoising of data, for better derivatives. This could be a simple gaussian denoise.
• Choosing to load some actual medical datasets to test the algorithm on actual data.
• Ability to choose different regression settings for different sensor measures, since some benefit from a different regression power.
• Smoother interface
  • Less ugly button
  • Animation when zooming
  • Responsive
• Include the other sensors we wanted, like the oxygen or altitude sensor.
• Switch the entire regression method to a probability based one so we can predict the accuracy of your predictions.
• Fix the bug where moving the mouse too fast while drawing will randomly not move some points on the curve.

Uncertainty

Note: We were not able to implement the uncertainty model in the script. This is due to the fact that all of these implementation are done on the raw signal coming in. We have no such data and we are working with software, no hardware attached. As every person has different baseline measurements as well, no concrete calculations can be done from other available data. For these reasons, the further explained model will only be described, with incomplete data, but not implemented in the actual script.

As of now, the prediction model is a simple regression of the n-th order (user can decide). This gives a fast real time prediction of values based on prior data. However, a simple regression is not very reliable when it comes to prediction, and even though in the very near future (t + 60s) it might do its job very well and extremely fast compared to other methods, the use case for such a near future accuracy is practically non-existent as by that time, it is too late to prevent an injury or further harm to the user. We realized that we needed something better for the algorithm to be actually useful for our users. Also, as this project is working with different sensors, all having different sensitivities and errors, a predictive filter and probability model will be needed. The suit becomes very cumbersome once the false positives begin to flood in, and false negatives are even worse as this can be a life or death decision factor. These problems hinders the user, but also the rescue post. Furthermore, preventive care is favoured over acute care. Therefore, the algorithm should be able to predict what could happen at least 15 minutes beforehand so the right actions can be taken.

We have found promising research, with a particular study that set out to predict core body temperature based on earlier data of core body temperature only (Gibrok, Buller, Reed, Hoyt & Reifman, 2010). Their incentive was the fact that military personnel deployed in tropical or desert climates suffer from heat illnesses, and for this reason, the U.S. Army is developing a system to detect such non-combat -and combat- related injuries. It started as a prediction model for core temprature (Gibrok, Mckenna, Reifman, 2006) and then was expanded to incorporate prediction errors to indicate reliability (Gibrok, Buller, Hoyt & Reifman, 2007). Initially, a Butterworth filter is used to predict temperature offline, based on any individuals’ data. It provided accurate predictions because of the ability to incorporate future data points as it was implemented offline. It is however crucial, for the suit as well as the U.S. Army prediction system, to be able to predict in real time. Therefore, Gibrok et al. have implemented a Butterworth zero-phase, low-pass filter. As the only difference of using this algorithm for other sensors would be the coefficients and vectors, we can use the same algorithm to predict the other sensor readings.

To have a real time core temperature prediction algorithm, data filtering and a predictive model are needed. Filtering can be simply done by putting a low pass filter in the circuitry. Here, a Butterworth zero-phase low pass filter will be used at a cutoff frequency of 42.5 mHz (Gribok, Buller, Hoyt & Reifman, 2010). The cutoff frequency is based on the analysis of the power spectrum of the core temperature, such that approximately 99% of the variance was contained in the range below the cutoff frequency. A Butterworth filter has the best smoothing results in the passband in the offline algorithm (Gribok, Buller, Hoyt & Reifman, 2010). In this article, a Butterworth filter with order 5 has been used. Order five means that the filter contains more circuitry, thus more capacitors and inductors to filter better. The transfer function of this filter (Electronics Tutorials Team, n.d.). is as follows:

[math]\displaystyle{ H(j\omega) = \dfrac{1}{\sqrt{1+\epsilon^2(\dfrac{\omega}{\omega_p})^{2n}}}, }[/math]

where [math]\displaystyle{ H(j\omega) }[/math] is the transfer function, [math]\displaystyle{ \epsilon }[/math] is the maximum pass band gain, also known as [math]\displaystyle{ A_{max} }[/math], [math]\displaystyle{ \omega }[/math] is the cutoff frequency, [math]\displaystyle{ \omega_p }[/math] is the pass band frequency and n is the filter order, set to 5.

The phase shift can be cleverly eliminated by using a forward-backward technique. This means filtering the raw signal forward in time and then taking the same filter to filter the signal backward in time again.

The sampling frequency has been taken to be one sample per minute. This is due to the fact that the body takes a rather long time to change the core temperature. Because of the long sampling time, high-frequency noise gets introduced to the sample. This is fixed by downsampling it to 5-min. intervals, keeping only every fifth-sample signal before applying the Butterworth filter (Gribok, Buller, Hoyt & Reifman, 2010).

Now, a Butterworth filter will be used by taking the previously filtered and unfiltered signals to produce a filtered signal:

[math]\displaystyle{ f_t = \sum_{i=0}^{n}\phi_i y_{t-i}-\sum_{j=1}^{n}\phi_j f_{t-j}, }[/math] where [math]\displaystyle{ y_{t-i} }[/math] is the raw signal at time t-i, [math]\displaystyle{ f_t }[/math] is the filtered signal at time t in minutes, n is the order of filter (here, we used n=5), [math]\displaystyle{ \theta }[/math] and [math]\displaystyle{ \phi }[/math] are vectors of filter coefficients (which are unknown). These coefficients are different for every sensor model. An autoregressive model of order m is begin used to make predictions. It uses its own previous values and a stochastic term, linearly, to predict the outcome. Assume that the filtered signals are known until m, then

[math]\displaystyle{ p_{t+1}=\sum_{i=1}^{m}b_i f_{t-i+1}+\mu_{t+1}, }[/math] where [math]\displaystyle{ p_{t+1} }[/math] is the filtered prediction at time t+1, b is the vector of m unknown AR coefficients and [math]\displaystyle{ \mu_{t+1} }[/math] represents white noise, also with unknown parameter [math]\displaystyle{ \sigma }[/math] (variance). Note that increasing m also increases the waiting time before collecting real-time predictions. Repeating this equation M times gives you a prediction for the time at M, which has been chosen to be 20 in this article, as it gives accurate results. To get vector b, one has to determine some “training” data. This can be done by configuring the baseline measurements and knowing that body temperature has some “universal” model as every human has the same anthropomorphic characteristics. This training data has again been filtered with the same Butterworth filter [2]. To determine coefficient b, the least squares method can be used, but b will become ill-conditioned (large variance). Gribok et al proposed to extent the LS method by adding a penalty function. This comes with other pros and cons, which will not be mentioned now.

It is important to know how accurate the prediction is. The prediction accuracy has been calculated with the root mean squared error (RMSE).

[math]\displaystyle{ RMSE = \sqrt{\dfrac{1}{N}\sum_{i=1}^{N}(p_{i}-f_{i})^2}. }[/math] Here, N is set to 20 min, just as all previous data and calculations. This RMSE value now shows how much the prediction will be different to the actual data at time t = 20.

Single-point predictions are never sufficient, so a reliability test must be done. This will give the uncertainty of the prediction. The prediction interval (PI) for the predicted value at t+1 can be estimated with

[math]\displaystyle{ PI =p_{t+1} \pm Z_{\alpha/2}\sigma(pred) }[/math] where [math]\displaystyle{ Z_{\alpha/2} }[/math] is a prediction factor associated with a [math]\displaystyle{ \alpha% }[/math] false positive finding and [math]\displaystyle{ \sigma }[/math] is the standard deviation. Gribok et al have chosen [math]\displaystyle{ Z_{\alpha/2} }[/math] to be 2.98.

Results When applying all known variables, the RMSE for real-time predictions (20 min) was found to be [math]\displaystyle{ 0.33^o C }[/math] for eight individuals, [math]\displaystyle{ 0.34^o C }[/math] for a cadet and [math]\displaystyle{ 0.22^o C }[/math] for a soldier. When looking at the prediction at time 10 and 20, 20 showed to be twice as inaccurate as 10 min (RMSE of 20 min to be 2x RSME of 10 min). Therefore, a further prediction should not be made as this will become more inaccurate. In the offline trials, the algorithm produced better results as it was able to incorporate future data points, however, for our suit this was just not useful as we want to predict hypo- and hyperthermia as accurate as possible in order to prevent these illnesses. As this algorithm seems to give good results, we intend to implement this in our design. Again, we have not succeeded in doing so. Hopefully, this research conduct would be enough to explain the uncertainty model.

It is however important to remark that besides prior core body temperature to predict future hypo- or hyperthermia, other variables that the suit measures are also an indication of one’s core body temperature, or provide a bodily environment that increases the likelihood of core temperature to go up or down. Besides core body temperature, other predictions can be made regarding the health of the user, as External environmental factors are also very much a determining factor in what the body does, as the body works hard to keep homeostasis regardless of the environment. Then lastly, person specific data such as age, length, body mass index and weight can be influential in predicting how the body reacts, and are very easy to obtain. For future research, incorporating these measurements can improve the suit and its reliability, and should have been done from the get go.

User Alerts

Based on the results from the surveys some more research was done in terms of a notification system. Unfortunately, due to a lack of time, this has not yet appropriately been developed. Some preliminary research is provided below.

Notification system

An application was designed that gathers information from other apps on one’s phone, and places the user on a waiting list. It shares personal information the user allows the app to share, and receives notifications with information or waiting list data (West, 2016).

Visual and auditory user notification methods for smart-home hazard detector

This device detects developing hazardous levels of smoke or carbon monoxide and gives warning signals using lights (three options), sounds/speech (using a speaker), and alarm horns. The hazard level is accessed and compared to preset threshold levels, based upon which alarms can be activated. The device is designed to tie in with smart-home solutions (Fadell et al., 2014).

Future Research

As we have indicated multiple times throughout the report, there are still a lot of points to improve in and conduct further research in. The reason for this is the fact that our research has remained fairly broad and theoretical. As you might have noticed, we struggled the first few weeks with finding a clear goal. By extension, we researched a lot of different things, and only went in-depth in the algorithm. Therefore, the following points are still valuable to conduct further research in:

The Algorithm

This is without any doubt the most important point for future research. We have been focusing on developing an algorithm that can triangulate sensor data and use this to provide the user with warnings. However, this is still a preliminary version and should be fine-tuned. The product now only contains three worked out sensors. Of course, for maximum usefulness, a suit with all kinds of sensors would give the most accurate and reliable results. The algorithm is only a fraction of the initial plan. It contains the most basic functions and the uncertainty model could not be implemented. Issues such as the regression being only an approximation of the real sensor data should be improved. Furthermore, some combinations of sensor readings could not be found, thus have no (reliable) warning system. The algorithm measures baseline values, but does not go into great depth. Take for example extreme sporters with some kind of disease. They do not necessarily follow the same e.g. heartbeat trend, see Chaudhry et al., 2017. The algorithm only contains coding without the uncertainty model. This model has to be worked out further and implemented in the code as well.

The SmartSuit Design

In this project, we have developed a prototype for the Hiketech SmartSuit. For this prototype, we have elected to use sensors and electronics that are affordable for the user, in order to keep costs low and to first test the functionality of the suit. Should this concept work, it would be valuable to develop a Hiketech SmartSuit 2.0, in which all SotA sensors are incorporated (namely the finger pressure sensor), in which the SotA wiring is used (self-healing cables) and in which the connection to the drone is incorporated. This might still take some time, and might be research for in the far future, yet this would be valuable at some point. Furthermore, the wiring is sort of random at the moment. Where should the wires be placed exactly? How flexible should the wires be? Which points are the most damaging places? etc.

Two other important aspects regarding the design are:

• Researching the alternative options for the SmartSuit base architecture. In the report, we have opted for a certain architecture. However, alternatives have been given as well. These have not yet fully been researched, and it would be valuable to investigate the trade-off between several options. Perhaps, for a future design, an other architecture could be more valuable.

• Researching the actual trade-off between I2C and SPI connection. For the prototype design, we have used I2C. However, SPI has its advantages as well. This could be valuable to research.

Data Transmission

Communication is a very important part for the end product. It is needed in various ways, such as communication between the sensors, the sensors and the main CPU, the CPU and the device used to get notifications, the main CPU and some base stations, base stations and drones, etc. Up to now, we have provided a theoretical overview of the options for data transmission. This should still be conceptualised and tested. How to actually ensure all data is properly transmitted, both to the user and the rescue post when necessary?

User

Up to now, we have based our assumptions on both literature research and the opinion of a few mountaineers. However, in order to increase the reliability of these assumptions, more participants should be found. Furthermore, while writing the algorithm, we noticed that the warning system was actually not user friendly. Now, the system seems to be friendlier, but more attention should be paid to the user in a potential follow-up project. For example, how does the user want the notifications? How often does the user want the notification? In what way does the user want the notifications? etc. Another follow-up project could be the communication of the user to the rescue post. Does the user want a microphone to be able to communicate at any time? This could result in a privacy violation, so the question would be if (s)he wants this and/or what other means are there to communicate? It is hard to communicate when you are disabled in any way, so talking would seem to be the easiest way. Again, this is often unwanted because of privacy. Furthermore, this suit was designed for hikers/mountaineers. This does not mean that a look-a-like suit cannot be made for other extreme sporters.

Drone

As indicated in the report, we have given some basic ideas and the groundwork for incorporating te drone in the SmartSuit. However, this is still one of the main points of further research. How to connect the drone specifically, how to program it that it know what care to provide, how to connect it to the rescue station and communicate with the rescue squad, etc.? This can be elaborated upon and could be considered as a research on its own.

Feedback on the Process

One last section we have chosen to include in the report concerns the reflection on our process and progress. As has become clear in the weekly meetings, our group as been struggling with defining a clear goal and going in-depth in one specific subject. To be entirely honest, we are quite disappointed by this fact. As has become clear throughout the process, we have done a lot of research and have mainly developed a theoretical framework. However, except for the algorithm, we have little actual conceptual proof of our project. We would have liked to see this otherwise, and that is the reason why we will reflect on our process, in order to prevent this from happening in the future.

At the start of the project, we had little idea what we wanted to develop. We opted for a SmartSuit to be implemented in extreme sports, but this was still very vague and a broad topic. One massive mistake we made here was assuming that the teachers would provide us with a direction. Since we had no clear goal, we hoped the teachers would give us one. When it became clear this was not the objective of this course, we had no idea how to proceed and were stagnating.

The following weeks had been filled with as much research as possible on all topics related to the SmartSuit, since we did not know what we wanted as end deliverable. This too was a huge mistake. By doing this we researched a lot of things (some of which we did not even use), we lost a lot of time and we forced ourselves to remain working on a superficial level. In the future, we should make sure that we first define a clear goal, and only when we have an objective that is realistic (SMART), then we can start researching and elaborating on our subject.

Near the end of the course, it dawned on us we really had to make some progress in order to provide the teachers with an actual deliverable. We have the teachers to thank for this, since in week 6 (maybe 7) we had a long and fruitful conversation regarding our stagnation and how to solve this issue. We then quickly discovered that what the user wants most is a warning that prevents them from approaching critical conditions. In order to ensure the SmartSuit could provide such preventive warnings, we needed an algorithm. This had become our deliverable.

The last few weeks, we have been working extremely hard in order to get a preliminary version of the algorithm working. By the time we realized what we wanted as end deliverable, we only had three weeks left to get this working. This led to an extremely high workload and stress within the team, but we are actually proud of the result we managed to achieve within this period. This is also why for the peer review, all grades lie between 6,5 and 8 (we all contributed a lot and put in a lot of effort to get the algorithm working). In this regard, we are reflecting happily on the process, even though it should never have come to that situation in the first place.

In short, we are not entirely happy with the results (too theoretical and broad) but on the other hand we are proud of what we managed to achieve the last few weeks. In future projects, our main point of attention would be to first identify our end goal and then start working (instead of the other way around).

What we have learned

During the project, we have learned many things. The most important aspects are given below:

• Knowledge regarding mountaineering and its risks. During the project, we have conducted extensive research on mountaineering, what risks are associated with it and what are points that should be improved.
• Adopting a User-centered approach. While we first set out to define our own goal, we quickly learned that the benefits for the user should be at the center of our design. That is why, from then on, we based every choice on what would be best for the user.
• Understanding the current SotA technology and its functionality. Since we conducted extensive research for our SmartSuit, we have learned a lot regarding the technology that currently exists, how it works, what is important to take into account and what properties are of interest. During this research, our understanding of technology increased exponentially.
• Learning to work in a multidisciplinary team and dividing the tasks efficiently. In this project, every member came from a different study and had a different field of expertise. It was very insightful to see who excelled in what aspects and how tasks could be divided optimally. We learned a lot from each other and especially near the end (when the workload was extremely high) we learned to delegate an build upon each others skills.

Unused Literature Research

Space Exploration

At the start of the project, we have researched wether or not a SmartSuit could be valuable for space applications as well. However, we did not elaborate on this. The research we conducted is given below.

Space psychology and psychiatry

Common health problems in space with some degree of relevance to our subject include:

• Visual Illusions
  • Microgravity causes astronauts often mistake stationary objects as moving, as well as being unable to judge their own movement accurately. This is probably of no relevance on mars considering mars has some gravity. This effect usually lasts between a couple hours and a month at max.
• Error proneness
  • A significant factor is difficulty with hand-eye coordination due to low gravity. We have never been to mars, but this may also be an issue on mars with some, but unusual gravity
  • A different issue is sleep deprivation, caused possibly by a upset cartesian rhythm. Sleep medication is the second most used type of medication on the ISS, after motion sickness medication.
  • The consistent high-stress of the work of the space-explorers may also be a factor
• Somatoform disorders
• (Possibly psychological) Muscle weakness

(Kanas and Manzey, 2008).

Could this be the first mars airplane?

This article details a SOTA airplane, the “Prandtl”, optimized for autonomous flight on Mars. It is a fixed-wing vehicle and not a helicopter-like drone. While the project is still in development, the goal is to make it able to fly for 5 hours, giving it a range of 165 km. However, the next planned test model has a max flight time of only 10 minutes, with a range of 32km. The plane is planned to fly to mars around 2020-2022. The entire design is mostly optimized for gliding and for maximal range needs to be launched from a high altitude, though there is no reason why a more powerful engine could not be attached to a similar design (Levine and Conner, 2017; Gibbs, 2016).

Helicopter could be ‘Scout’ for mars rover This article has some information about a test for a helicopter on mars. It would be used to scout an area for rovers to explore later. There is a functional proof-of-concept. The device would weigh 1 kg and with a have a distance of 1.1 meter from one blade tip to the other (Landau, 2015).

Safe passage: Astronaut care for exploration missions

This article claims safety and health issues are the greatest problem preventing exploration of deep space. Major space-related problems include:

• Bone loss
  • also, astronauts break bones or sprain joints sometimes due to heavy objects moving through 0 gravity, slamming themselves into walls, or other problems related to low gravity. A average of 0.44 bone breaks or joint sprains are recorded per 14 days on space shuttles.
  • Second most common on the russian MIR, with 32 incidents
• Cardiovascular Alterations
  • 0.02 incidents every 14 days during space shuttle program
  • The most common problem on the russian MIR, with 32 incidents total
• Reduced performance
• Neurological and sensing problems
  • One incident every 14 days on average on the space shuttle
  • “headache” is the third most common incident on the russian MIR, with 17 incidents.
• Immune system problems
  • probably links to the various other problems, but not directly measurable
• Insomnia
  • 13 incidents on the russian MIR
• Muscle problems
• Neuro Vestibular Adaptation (Motion Sickness)
  • By far the most common problem on the space shuttle, with a average of 2.4 incidents every 14 days.
• Radiation effects
• Hearing problems
• Clinical Capability
  • Exposure to toxins
  • Altered drug reactions
  • Illness
  • Decompression

Most common problems are motion sickness, nasal congestion, and sleep disorders. A interesting observation is that astronauts die disproportionally often in car crashes. The most common medical procedures performed on submarines are:

• Wound care
• Suturing
• Cleansing
• Nail removal
• Fluorescein eye examination
• Incision and drainage of abscess
• Tooth restoration

(Ball and Evans, 2001).

There is a lot more date in this article about procedures on submarines and antarctic bases. A very useful source for later.

Smartwear Restrictions and Necessities

More research into the restrictions and options for a smartsuit was conducted than was actually used in our final design. The unused research can be seen below.

A review of wearable sensors and systems with application in rehabilitation

Because health care system in the US is not equally distributed across the areas of interest in which health care is needed the most, it becomes increasingly attractive to bring the health care inside the homes of patients. A combination of both ambient and wearable sensor technology would provide the means necessary to bring health care inside the homes of people across the globe. The application of the sensors can be categorized as follows:

• health and wellness monitoring
• safety monitoring
• home rehabilitation
• assessment of treatment efficacy
• early detection of disorders

Concluding, the step towards home monitoring and home rehabilitation creates interest into designing and developing wearable sensors and system that provide data and assist remotely (Patel, Park, Bonato, Chan and Rodgers, 2012).

Relevance for our objective: Several wearable sensors mentioned can be applied to extreme sporting wearables in order to facilitate monitoring of the condition of extreme sporters.

Non-invasive wearable electrochemical sensors: a review

As chemical sensor used to be invasive, wearable chemical sensor were absent in on-body sensing technology departments. As electrochemical sensor are becoming increasingly non-invasive, they find their place in wearable devices and applications such as health-care, sport and military. The specific type of sensors make use of either sweat, tears or saliva, and require (obviously) direct contact with one of the aforementioned fluids. A benefit of these sensors that make use of bodily fluids is that they can continuously provide information about the status of the actor wearing the device. It should be noted that the recognition element is not similar for the types of bodily fluids tested, and therefore, to have the full range of elements recognised, all three bodily fluids should optimally be used while sensing (Bandodkar and Wang, 2014).

Relevance for our objectives: Non invasive way of sensing chemical homeostatic properties of a subject. Can be worn without obstructing extreme sport athletes.

Hybrid System of Electro-Textile based wearable Microstrip Patch Antenna with Tuning Holes

A patch that can be sewn into textile which can emit radio frequencies (Ullah and Baghel, 2015).

Relevance for our objectives: Data collected can be transmitted using this patch. Possibly to phone to relay the data to a specific server.

Micro-Drone for Gas Measurement in Hazardous Scenarios via Remote Sensing

As gas releases, whether they are volitional, meant to do harm, or mother nature, become more prevalent, devices to detect such releases are no unwanted luxury. By combining a drone with gas measurement equipment, gases can be detected from a safe distance. In order to achieve this functionality, the drone makes use of techniques common in nature known as”plume tracking” or “odor-source localization”. The measuring techniques have been tested in both a test chamber, with two standard measurement devices as reference, as well as a field test in a volcanic crater on Lanzarote. The test chamber experiment provided near perfect results compared to the reference devices. The test inside the Lanzarote volcanic crater provided successful results as well, showing a significant low concentration of SO2, a typical volcanic gas. To summarize: the combination of drones and gas measurement devices provide safe possibilities to detect airborne hazards (Bartholmai and Neumann, 2010).

Relevance for our objectives: Detection of extraterrestrial airborne hazards, as well as safe areas and measurement tactics for extreme sporting areas such as high altitudes or remote locations.

Chronological Workdocument

Since the process has taken several weeks, we have elected to show the chronological progress of our project as well. On this page, you can see what we did each week and how much progress we made.

Do notice that the chronological workdocument is outdated! The choices and changes in the project have not been edited for each week. Furthermore, our objective has changed several times during the project, as well as our final deliverable. Therefore, the chronological document contains inconsistencies and is incomplete with regards to the actual report. The chronological workdocument is thus merely to view our process, the contents are in no way representative for the final deliverable!

References

123Accu.nl. (2018). Apple iPhone 5s accu (1500 mAh, 123accu huismerk). Retrieved from https://www.123accu.nl/Mobiele-telefoon-accu/Apple/616-0652-p275360.html

Abegg, S. (2006). A Mostly-Graphical Presentation of Mountaineering Accident Statistics. Stephabegg.com. Retrieved on June 17th 2018, from http://www.stephabegg.com/home/projects/accidentstats

Abreu, M. M., (2017). U.S. Patent No. 15/602523. Tortola: Geelux Holdings Ltd.

Acciaroli, G., Vettoretti, M., Facchinetti, A., & Sparacino, G. (2018). Calibration of Minimally Invasive Continuous Glucose Monitoring Sensors: State-of-The-Art and Current Perspectives. Biosensors, 8(1), 24.

Al-Shaqsi, S. Z. K. (2010). Response time as a sole performance indicator in EMS: Pitfalls and solutions. Open Access Emergency Medicine : OAEM, 2, 1–6.

Asogwa, C., Collins, S., Mclaughlin, P., & Lai, D. (2016). A Galvanic Coupling Method for Assessing Hydration Rates. Electronics, 5(4), 39. (Asogwa, Collins, McLaughlin & Lai, 2016)

Backe, S., Ericson, L., Janson, S., & Timpka, T. (2009). Rock climbing injury rates and associated risk factors in a general climbing population. Scandinavian journal of medicine & science in sports, 19(6), 850-856.

Ball, JR & Evans, CH (2001) Safe Passage: Astronaut Care for Exploration Missions. Washington DC, NATIONAL ACADEMY PRESS

Bandodkar, A. J., & Wang, J. (2014). Non-invasive wearable electrochemical sensors: a review. Trends in biotechnology, 32(7), 363-371.

Bartholmai, M., & Neumann, P. (2010). Micro-drone for gas measurement in hazardous scenarios via remote sensing. In Proceedings of.

Bergmann, J. H., Chandaria, V., & McGregor, A. (2012). Wearable and implantable sensors: the patient’s perspective. Sensors, 12(12), 16695-16709.

Benaron, D. A., (2015). U.S. Patent No. 14/864,860. San Francisco: AliphCom

BMC Climbing Injury Symposium 2016 (Facebook Event). (n.d.). Retrieved from https://www.facebook.com/events/1729266083994236/

Bonato, P., (2010). Wearable Sensors and systems. 25-36

Bosch SensorTech. (2018). BMP280 Digital Pressure Sensor. Retrieved on 20 May 2018, from https://ae-bst.resource.bosch.com/media/_tech/media/datasheets/BST-BMP280-DS001-19.pdf.

Brand, van den, J., de Kok, M., Sridhar, A., Cauwe, M., Verplancke, R., Bossuyt, F., ... & Vanfleteren, J. (2014, September). Flexible and stretchable electronics for wearable healthcare. In Solid State Device Research Conference (ESSDERC), 2014 44th European (pp. 206-209). IEEE.

Brocherie, F., Girard, O., & Millet, G. P. (2015). Emerging environmental and weather challenges in outdoor sports. Climate, 3(3), 492-521.

Brymer, E. (2009). The extreme sports experience: a research report. IFPRA world, 6-7.

Brymer, E., Downey, G., & Gray, T. (2009). Extreme sports as a precursor to environmental sustainability. Journal of Sport & Tourism, 14(2-3), 193-204.

Calvin, K. (2018). Hyperthermia: too hot for your health. National Institutes of Health (NIH). Retrieved May 25, 2018, from https://www.nih.gov/news-events/news-releases/hyperthermia-too-hot-your-health-1

Carlsen, K. H. (2012). Sports in extreme conditions: the impact of exercise in cold temperatures on asthma and bronchial hyper-responsiveness in athletes. Br J Sports Med, 46(11), 796-799.

Chaudhry, S., Kumar, N., Behbahani, H., Bagai, A., Singh, B. K., Menasco, N., … Myers, J. (2017). Abnormal heart-rate response during cardiopulmonary exercise testing identifies cardiac dysfunction in symptomatic patients with non-obstructive coronary artery disease. International Journal of Cardiology, 228, 114–121. https://doi.org/10.1016/J.IJCARD.2016.11.235

Climbing finger pulley injury....to crimp or not to crimp? (2017, December 22). Retrieved from https://www.mapleclinic.ie/bloghome/fingerpulleyinjury/

Crowley, T. (2016). The Flexor Tendon Pulley System and Rock Climbing. Journal of Hand and Microsurgery, 04(01), 25-29. doi:10.1007/s12593-012-0061-3

Costa, G., Keulen, V., van Rijsbergen, R. (2016), “ePatch, The Electronic Band-Aid”, research by Fontys Hogeschool Techniek & Logistiek, commissioning company: Yellow Factory

EC (2007) 12th Implementation Report. COM(2007) 155. Brussels: European Commission http://www.splitboardmag.com/article/communication-devices-mountains/

EE Herald, “SPI Bus interface”, http://www.eeherald.com/section/design-guide/esmod12.html

Electronics Tutorials Team (n.d.). Butterworth Filter Design with a Low Pass Butterworth. Retrieved June 20, 2018, from https://www.electronics-tutorials.ws/filter/filter_8.html

Eliasson, Jens, Per Lindgren, and Jerker Delsing, “A Bluetooth-based Sensor Node for Low-Power Ad Hoc Networks,” Journal of Computers, vol. 3 (5), pp. 1–10, 2008.

Fadell, A. M., Rogers, M. L., Sloo, D., Veron, M., Guen, S., Le, & Webb, N. (2014). U.S. Patent No. US20180137745A1. Washington, DC: U.S. Patent and Trademark Office.

Figure 1, Finger pulley anatomy retrieved from http://www.turntillburn.ch/research/BIOMECHPAP.htm

Figure 2, Crimp Grip retrieved from https://web.archive.org/web/20110930151110/http://www.nicros.com/training/articles/finger-tendon-pulley-injury/

Finger Tendon Pulley Injury. (n.d.). Retrieved from https://web.archive.org/web/20110930151110/http://www.nicros.com/training/articles/finger-tendon-pulley-injury/

Gardner, T., (2015), “Popularity and economic benefit of mountaineering: instant expert”, retrieved from https://www.thebmc.co.uk/participation-in-climbing-mountaineering

Geddes, L. (2016). Wearable sweat sensor paves way for real-time analysis of body chemistry. Nature. Retrieved on 29 May 2018 from https://www.nature.com/news/wearable-sweat-sensor-paves-way-for-real-time-analysis-of-body-chemistry-1.19254

Gholipour, B. (2018). What Is a Normal Heart Rate?. Live Science. Retrieved May 25, 2018, from https://www.livescience.com/42081-normal-heart-rate.html

Gibbs, Y. (2017, May 11). Prandtl-D Aircraft. Retrieved from https://www.nasa.gov/centers/armstrong/news/FactSheets/FS-106-AFRC.html

GMB. (2015). Low temperature LiPo. Retrieved from http://www.gmbattery.com/product/showproduct.php?lang=en&id=33

Gorgutsa, S., Bélanger-Garnier, V., Ung, B., Viens, J., Gosselin, B., LaRochelle, S., & Messaddeq, Y. (2014). Novel wireless-communicating textiles made from multi-material and minimally-invasive fibers. Sensors, 14(10), 19260-19274.

Greene, J., Louis, J., Korostynska, O., & Mason, A. (2017). State-of-the-Art Methods for Skeletal Muscle Glycogen Analysis in Athletes—The Need for Novel Non-Invasive Techniques. Biosensors, 7(1), 11.

Gribok, A. V, Buller, M. J., Hoyt, R. W., & Reifman, J. (2010). A Real-Time Algorithm for Predicting Core Temperature in Humans. IEEE Transactions on Information Technology in Biomedicine, 14(4), 1039–1045. https://doi.org/10.1109/TITB.2010.2043956

Gribok, A., Mckenna, T., & Reifman, J. (2006). Regularization of Body Core Temperature Prediction during Physical Activity. 2006 International Conference of the IEEE Engineering in Medicine and Biology Society. doi:10.1109/iembs.2006.4397436

Gribok, A. V., Buller, M. J., Hoyt, R. W., & Reifman, J. (2007). Providing Statistical Measures of Reliability for Body Core Temperature Predictions. 2007 29th Annual International Conference of the IEEE Engineering in Medicine and Biology Society. doi:10.1109/iembs.2007.4352348

Gribok, A., Buller, M., & Reifman, J. (2008). Individualized Short-Term Core Temperature Prediction in Humans Using Biomathematical Models. IEEE Transactions on Biomedical Engineering, 55(5), 1477-1487. doi:10.1109/tbme.2007.913990

Griffith, J., Cluff, K., Eckerman, B., Aldrich, J., Becker, R., Moore-Jansen, P., & Patterson, J. (2018). Non-Invasive Electromagnetic Skin Patch Sensor to Measure Intracranial Fluid–Volume Shifts. Sensors, 18(4), 1022.

Himalayan Medics. (n.d.). Retrieved June 2, 2018, from http://www.himalayanmedics.com/services.php#services

Huey, R. B., & Eguskitza, X. (2001). Limits to human performance: elevated risks on high mountains. Journal of Experimental Biology, 204(18), 3115-3119.

Ichikari, R., Ruiz, L.C.M., Kourogi, M. (2015). Indoor floor-level detection by collectively decomposing factors of atmospheric pressure. IEEE Xplore. https://ieeexplore.ieee.org/abstract/document/7346776/

International Federation of Sport Climbing, “Key Figures”, retrieved from https://www.ifsc-climbing.org/index.php/media-centre/key-figures-2

Iwasaki, Wataru & Nogami, Hirofumi & Takeuchi, Satoshi & Furue, Masutaka & Higurashi, Eiji & Sawada, Renshi. (2015). Detection of Site-Specific Blood Flow Variation in Humans during Running by a Wearable Laser Doppler Flowmeter. Sensors. 15. 25507-25519. 10.3390/s151025507.

Jebson, P. J., & Steyers, C. M. (1997). Hand Injuries in Rock Climbing. The Physician and Sportsmedicine, 25(5), 54-63. doi:10.3810/psm.1997.05.1341

Jones, L., (2008), “Extreme sport growing in popularity”, retrieved from http://xtremesport4u.com/extreme-land-sports/extreme-sport-growing-in-popularity/

JST ERATO Someya Bio-harmonized Electronics Project. (n.d.). Retrieved from http://www.ntech.t.u-tokyo.ac.jp/en/project/goal_index.html

Kanas, N, & Manzey, D (2008). Space Psychology and Psychiatry. Springer Dordrecht.

Kim, J., Kwon, S., Seo, S., & Park, K. (2014, August). Highly wearable galvanic skin response sensor using flexible and conductive polymer foam. In Engineering in Medicine and Biology Society (EMBC), 2014 36th Annual International Conference of the IEEE (pp. 6631-6634). IEEE.

Kleiner Nutrition. (2015, August 28). Dehydration and Fluids in Running. Retrieved from http://kilcooleac.com/dehydration-and-fluids-in-running/

Koch, E., & Dietzel, A. (2016). Skin attachable flexible sensor array for respiratory monitoring. Sensors and Actuators A: Physical, 250, 138-144. doi:10.1016/j.sna.2016.09.020

L, R. (2018). World's fastest drone: 200MPH 6s XLR. Youtube. Retrieved on June 5th 2018, from https://www.youtube.com/watch?v=GlELt3KM_OU.

Landau, E. (2015, January 22). Helicopter Could Be 'Scout' for Mars Rovers. Retrieved from https://www.jpl.nasa.gov/news/news.php?feature=4457

Lee, Kang, “A Smart Transducer Interface Standard for Sensors and Actuators,” in The Industrial Information Technology Handbook, Zurawski, R., Ed., Boca Raton, FL, CRC Press, 2005, pp. 1–16.

Lee, J., Kim, D., Ryoo, H. Y., & Shin, B. S. (2016). Sustainable wearables: wearable technology for enhancing the quality of human life. Sustainability, 8(5), 466.

Lee, Jin-Shyan, Yu-Wei Su, and Chung-Chou Shen, “A Comparative Study of Wireless Protocols: Bluetooth, UWB, ZigBee, and Wi-Fi,” presented at the 33rd Annual Conference of the IEEE Industrial Electronics Society (IECON), Taipei, Taiwan, 2007.

Lee, S., Reuveny, A., Reeder, J., Lee, S., Jin, H., Liu, Q., . . . Someya, T. (2016). A transparent bending-insensitive pressure sensor. Nature Nanotechnology, 11(5), 472-478. doi:10.1038/nnano.2015.324

Levine, J, & Conner, M. (2017, June 29). Could This Become The First Mars Airplane? Retrieved from https://www.nasa.gov/centers/armstrong/features/mars_airplane.html

Linke, B. (2008). Overview of 1-Wire Technology and Its Use. Retrieved from https://pdfserv.maximintegrated.com/en/an/AN1796.pdf

Ma, J. L. G., & Dutch, M. J. (2013). Extreme sports: Extreme physiology. Exercise‐induced pulmonary oedema. Emergency Medicine Australasia, 25(4), 368-371.

Magno, M., Salvatore, G. A., Mutter, S., Farrukh, W., Troester, G., & Benini, L. (2016, May). Autonomous smartwatch with flexible sensors for accurate and continuous mapping of skin temperature. In Circuits and Systems (ISCAS), 2016 IEEE International Symposium on (pp. 337-340). IEEE.

Majumder, S., Mondal, T., & Deen, M. J. (2017). Wearable sensors for remote health monitoring. Sensors, 17(1), 130.

Malashenkova, M. (2016, October). Extreme sports in Extreme conditions. Paper presented at ITP Sport, Exercise & Health Research Symposium, Institute of Sport & Adventure (ISA), Otago Polytechnic (OP).

Mansor H., Shukor M.H.A., Meskam S.S., Rusli N.Q.A.M., Zamery N.S. Body temperature measurement for remote health monitoring system; Proceedings of the 2013 IEEE International Conference on Smart Instrumentation, Measurement and Applications (ICSIMA); Kual

Mayo Clinic Staff. (2018, May 08). Hyponatremia. Retrieved May 25, 2018, from https://www.mayoclinic.org/diseases-conditions/hyponatremia/symptoms-causes/syc-20373711

Mayo Clinic Staff (2018, Januari 11). Hypoxemia (low blood oxygen). Retrieved May 25, 2018, from https://www.mayoclinic.org/symptoms/hypoxemia/basics/definition/sym-20050930

McGrath, M. J., & Scanaill, C. N. (2013). Key Sensor Technology Components: Hardware and Software Overview. In Sensor Technologies (pp. 51–77). Berkeley, CA: Apress. https://doi.org/10.1007/978-1-4302-6014-1_3

MedLine Plus team (2018). Body temperature norms: MedlinePlus Medical Encyclopedia. [online] Medlineplus.gov. Retrieved May 25, 2018, from https://medlineplus.gov/ency/article/001982.htm

Melin, B., & Savourey, G. (2001). Sports and extreme conditions. Cardiovascular incidence in long term exertion and extreme temperatures (heat, cold). La Revue du praticien, 51(12 Suppl), S28-30.

Mell HK, Mumma SN, Hiestand B, Carr BG, Holland T, Stopyra J. Emergency Medical Services Response Times in Rural, Suburban, and Urban Areas. JAMA Surg. 2017;152(10):983–984. doi:10.1001/jamasurg.2017.2230

Montain SJ, Coyle EF. The influence of graded dehydration on hyperthermia and cardiovascular drift during exercise. J Appl Physiol. 1992;73: 1340-1350.

Mounteverest.net. (n.d.). Retrieved June 2, 2018, from http://www.mounteverest.net/expguide/route.htm

Mukhopadhyay, S. C., & Islam, T. (2018). Wearable Sensors:Applications, design and implementation. IOP Publishing. https://doi.org/10.1088/978-0-7503-1505-0

Murray, R. (1996). Dehydration, Hyperthermia, and Athletes: Science and Practice. Retrieved from https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1318513/pdf/jathtrain00019-0058.pdf

Nes, B. M., Janszky, I., Wisløff, U., Støylen, A., & Karlsen, T. (2013). Age-predicted maximal heart rate in healthy subjects: The HUNT Fitness Study. Scandinavian Journal of Medicine & Science in Sports, 23(6), 697–704. https://doi.org/10.1111/j.1600-0838.2012.01445.x

Nybo, L. (2012). Brain temperature and exercise performance. Experimental Physiology, 97(3), 333–339. https://doi.org/10.1113/expphysiol.2011.062273

Olsson, J. (2014). 6LoWPAN demystified. Retrieved from http://www.ti.com/lit/wp/swry013/swry013.pdf

Patel, S., Park, H., Bonato, P., Chan, L., & Rodgers, M. (2012). A review of wearable sensors and systems with application in rehabilitation. Journal of neuroengineering and rehabilitation, 9(1), 21.

Pathological Society. (2014, January 16). PYROGENS, CRYOGENS, AND BEARS OH MY! Retrieved from https://ugpathsoc.wordpress.com/2014/01/15/pyrogens-cryogens-and-bears-oh-my/

Persson, M., & Karlsson, D. (2018). U.S. Patent No. US9921293B2. Washington, DC: U.S. Patent and Trademark Office. https://patents.google.com/patent/US9921293B2/en

Price, C., Maor, R., Shachaf, H. (september 2018). Using smartphones for monitoring atmospheric tides. Elsevier Journal of Atmospheric and Solar-Terrestrial Physics. https://www.sciencedirect.com/science/article/pii/S1364682618301937

Replace Direct. (2018). Lithium-polymeer batterijen. Retrieved from https://www.replacedirect.be/nl/batterij-technieken/lithium-polymeer-batterijen.html

Richmond, V. L., Davey, S., Griggs, K., & Havenith, G. (2015). Prediction of Core Body Temperature from Multiple Variables. Annals of Occupational Hygiene, 59(9), 1168-1178. doi:10.1093/annhyg/mev054

Rouse, Margaret. “Serial Peripheral Interface (SPI)”, Last Update: March 2011, http://whatis.techtarget.com/definition/serial-peripheral-interface-SPI

ROYAL DUTCH CLIMBING AND MOUNTAINEERING FEDERATION (NKBV), (2015), retrieved from http://bergsportdag.nkbv.nl/wp-content/uploads/2016/09/Presentatie-NKBV-EN-BSD17.pdf

Schöffl, I., Oppelt, K., Jüngert, J., Schweizer, A., Neuhuber, W., & Schöffl, V. (2009). The influence of the crimp and slope grip position on the finger pulley system. Journal of Biomechanics,42(13), 2183-2187. doi:10.1016/j.jbiomech.2009.04.049

Schweizer, A. (2001). Biomechanical properties of the crimp grip position in rock climbers. Journal of Biomechanics,34(2), 217-223. doi:10.1016/s0021-9290(00)00184-6

SFIA. (n.d.). “Number of participants in climbing (traditional/ice/mountaineering) in the United States from 2011 to 2017 (in millions)”, retrieved from https://www.statista.com/statistics/763799/climbing-traditional-ice-mountaineering-participants-us/

SFIA. (n.d.). “Number of people who went hiking/backpacking within the last 12 months in the United States from spring 2008 to spring 2017 (in millions)”, retrieved from https://www.statista.com/statistics/763799/climbing-traditional-ice-mountaineering-participants-us/

s.n. (2016). Khumbu Icefall: Jason Crossing a Crevasse. Retrieved June 2, 2018, from https://www.youtube.com/watch?v=qHX8UXTWFqI

Solid Energy. (s.d.). Hermes, High Energy Rechargeable Metal Cells for Space. Retrieved from http://assets.solidenergysystems.com/wp-content/uploads/2017/09/08171937/Hermes_Spec_Sheet1.pdf

Sun, H., You, X., Jiang, Y., Guan, G., Fang, X., Deng, J., . . . Peng, H. (2014). Self-Healable Electrically Conducting Wires for Wearable Microelectronics. Angewandte Chemie, 126(36), 9680-9685. doi:10.1002/ange.201405145

Shariff, F., Hingorani, V., L., Albadawi, H., (2015). Monitoring hydration based on galvanic skin response U.S. Patent No. 20160374588A1. Microsoft Technology Licensing LLC

Shu, Y., Li, C., Wang, Z., Mi, W., Li, Y., & Ren, T. L. (2015). A Pressure sensing system for heart rate monitoring with polymer-based pressure sensors and an anti-interference post processing circuit. Sensors, 15(2), 3224-3235.

TBPC Healthcare. (s.d.). Wearable Oximetry module: Module Overview. Retrieved on 20 May 2018, from http://www.tbpchc.com/eng/medical.php?gid=13.

Trofimencoff, T. (Jan 6, 2016). A Flexible and Transparent Pressure Sensor. Retrieved from https://www.engineering.com/DesignerEdge/DesignerEdgeArticles/ArticleID/11409/A-Flexible-and-Transparent-Pressure-Sensor.aspx?e_src=relart

Tuned into Cycling Team (2012, July). Dehydration and Over Hydration (Hyponatremia) for the Cyclist. Retrieved from https://tunedintocycling.com/2012/07/

uBlox. (2018). ZOE-M8B Ultra-small, super low power u-blox M8 GNSS SiP module: Data Sheet . Retrieved on 20 May 2018, from https://www.u-blox.com/sites/default/files/ZOE-M8B_DataSheet_%28UBX-17035164%29.pdf

Ullah, S. U., & Baghel, R. K. (2015, October). Hybrid system of electro-textile based wearable microstrip patch antenna with tuning holes. In Soft Computing Techniques and Implementations (ICSCTI), 2015 International Conference on (pp. 135-139). IEEE.

Ultimate Kilimanjaro team, (2018). How Do I Prepare to Climb Kilimanjaro? Kilimanjaro Training, Gear, Visas and Vaccinations. Retrieved May 25, 2018, from https://www.ultimatekilimanjaro.com/preparation.htm

Voinea, G. D., Butnariu, S., & Mogan, G. (2016). Measurement and geometric modelling of human spine posture for medical rehabilitation purposes using a wearable monitoring system based on inertial sensors. Sensors, 17(1), 0003.

West, A. (2016). U.S. Patent No. US20180144307A1. Washington, DC: U.S. Patent and Trademark Office.

Young, C. C. (2002). Extreme sports: injuries and medical coverage. Current sports medicine reports, 1(5), 306-311.

Yuan, Gao, et al., “Low-Power Ultrawideband Wireless Telemetry Transceiver for Medical Sensor Applications,” Biomedical Engineering, IEEE Transactions on, vol. 58 (3), pp. 768–772, 2011.