Group 5: AED Drone from TU Delft

1. Thom Konijnenberg 0945007 t.j.c.konijnenberg (At) student.tue.nl
2. Rense Nijenkamp 0960705 r.c.j.nijenkamp (At) student.tue.nl
3. Nikolay Stoyanov 0980910 n.stoyanov (At) student.tue.nl
4. Wessel van der Heijden 0951686 g.w.v.d.heijden (At) student.tue.nl
5. Patrick Shaw 0900654 Pthom.shaw (At) gmail.com

Introduction

Every week, 300 cases of cardiac arrest will happen outside of the hospital in the Netherlands. Because it happens outside of the hospital, the survivability rate is less than 10%. This rate is tremendously low for a country in the EU. These rates are caused by the fact that after a cardiac arrest, only an AED or a defibrillator can revive the heart and with every minute passing, the chance of survival drop with another 10%. incident. This high number of casualties is caused by the relatively slow response time of emergency services (10 minutes). Brain death and permanent death start to occur in just 4 to 6 minutes.

However, the AED Drone from TU Delft is a solution for the slow response time. Because the drone is carrying an AED, it eliminates the need for a person to retrieve an EAD, which is very time costly. The drone is faster as an ambulance as well, due to the high speeds it can reach. Furthermore, as the air is much more open space than the roads have, the drone will not be limited by any form of traffic. The drone can reach the patient faster than any emergency service can and revive a person before the crucial minutes are over. However, the Drone from Delft is just a proto-type and the employment of Drones are blooming in this decade.

It is of upmost importance that the drone lands as fast as possible and to be able to land everywhere. However, the TU Delft only tested the prototype in optimal situations where there were no external factors. The goal of this project is to learn how people react to drones when it tries to land as fast as possible and as close as possible to the patient.

In order to show the assumptions that have been taken, a scenario is made which shows the difference between the implementation of a drone and without the implementation A man, aged 54, is taking a stroll through the centrum of Eindhoven on a hot Saturday. Suddenly, he is feeling unwell and decides to rest on a nearby bench. However, as he is sitting down, he feels a great pain coming from his chest. The pain is so severe that he can not cry out for help and collapses on the bench. In total, thirty seconds have been passed since he collapsed. A nearby bystanders tries to make contact but fails to achieve so and therefore decides to call the emergency number. A crowd starts to form around the collapsed man. The call is made and contact is established with the emergency contact center. On this moment, one minute has passed already while the bystander starts to reanimate the man.

The closest ambulance is contacted and has been giving orders to move to the location of the emergency. However, the traffic is terrible and causes the ambulance to loose precious seconds. With great hurry, the ambulance tries to move as fast as possible through the crowd but struggles a bit due to the amount of people that have to move out of the way. In total the ambulance took 6 minutes to arrive at the scene, making the total amount 7 minutes. They try to revive the man with the use of a defibrillator. The man has a survivability rate of less than 50%.

Another bystander who notices the call to the emergency center tried to help by retrieving the nearest AED and using it on the collapsed man. The bystander pulls out a phone and tries to locate the nearest AED. This takes 1 minute as not every AED on the map is available in the weekend. The bystander starts running in order to reach the AED, located 350 meters from the emergency location, on time. The running, retrieving, and running back will take another 5 minutes. In total 7 minutes have passed and the AED has to be used through the help of a bystander

The emergency center sends the closest drone to the emergency location. Due to the high-speed of the drone, it reaches the location within one minute and starts to land slowly. When contact with the ground is made, less than 3 minutes have passed. The AED has to be used through the help of a bystander.

Problem statement

Speed is one of the most important factor in saving lives. This has been mentioned in the introduction and will affect the problem statement as well. Currently, the drone is used in optimal situations, which are not present in every situation. For this project, the landing of the drone will be tackled. Landing a drone can be done in multiple ways, making great use of the speed the drone has. However, the landing will not be done on a landing site, landing will be done when there is a crowd of people.

There is not much research done on how people feel about drones when they are up-close. If a drone must land as fast as possible, it will try to arrive at a low altitude in order to have a low amount of time needed to land. However, to what extent does this influence the behavior of the people who are close to the drone. A drone is different than an ambulance, as it does not show a predictable course when it is moving towards a location. Furthermore, the downstream of wind is a feeling people are not experienced with and may cause fear.

Because the drone is not an object people are used to, an experiment will be done in order to see the effect the drone has on people when it needs to land as fast as possible. If people will move unpredictable, they may locate themselves in positions that are unfavorable for the landing of the drone. More information about the experiment can be found in the experiment plan.

Approach

Literature research.

While there is not much information available about how people react when a drone is trying to land as fast as possible, a lot of information is available about the TU Delft drone. When the capabilities of the drone are known, ideas can be formed on how the experiment needs to be done and if adjustments are necessary for the future. The state of the art can be researched, which will give a lot of insight on the possibilities the drone may have.

Despite the lack information regarding the relations between human and drone, there is a lot of information and references which show the proper base for the experiment that is going to be conducted. Other groups have conducted experiments, which we will be able to use to create ouw own experiment.

Experiment

An experimentation on how people will react when a drone is close by. As mentioned above, there is not much information regarding how people react when drones are close but this give us a base of approach. The experiment will give insight on how people tend to react when the drone is flying above them. Furthermore, this will give information about the personal space one feels when a drone is flying above a person. These kind of experiments should result in proper data which would allow us to what extend a drone can fly without creating any hindrance for nearby bystanders or be hindered by the reaction of bystanders.

Environment

The project is going to be focused on the environment aspect. The people who are not part of the crisis situation will have an effect on how the drone must reach its location. Therefore it is important to learn and know how the environment, or the crowd in this research, will be affected.

AED Map

Based on the AED map, made by the Red cross, one can pinpoint every location of an AED. However, one can conclude that there are certain area’s which are not covered by AEDs. From every AED, a circle is drawn with a radius of 500 meters. This is based on the distance one could access the situation, locate the nearest AED, retrieve it, and apply the AED to the patient. The 500 meters will take almost 10 minutes, giving the patient a chance of less than 10%. The biggest issue is still the time it takes to return with an AED. In order to get the survivability to reasonable numbers, the time must be made shorter. To get an response time within 3 minutes, which are called the golden minutes, is the main goal, as after 3 minutes brain damage will be caused. (link to Nikolay and picture of Alec mont). Therefore, help needs to arrive quickly in order to make a full recovery. The 500 meter are in reality little to 50 meter if one wants to apply the AED in the golden minutes.

For the current map, an area of 28 square kilometers has been taken, in which 60 AEDs are visible. These 60 cover about 50% of the area. Furthermore, of these 60 AEDs, about 40% are accessible at any given time, while the remainder is limited to the business hours of its location. Placing more AEDs in strategical places is an costly project. Most of the AEDs are over the price range of 1000 Euro and need an outside protector, otherwise it will be limited to the opening hours of its location. At the moment, an AED is placed when the owner feels the need to buy one or when it is a public building. This causes area’s which are used for homes and not for businesses or public buildings to be left without an AED. It can be seen on the map that those areas are usually more than 1 kilometer removed from the nearest AED. This will be almost certain death.

Because the time in which a cardiac arrest happens and the time a person calls the emergency number is already so costly on the patients live, a faster system must be used. The AED Drone has a response time of 1 minute inside its 2 kilometer radius. Because of this response time, the chances of survivability are greatly improved. At the moment, there is no better solution as the AED drone brings help at an unraveled speed, helping the patient in the golden minutes. This is caused by the speed of the drone and lack of traffic it finds on its path. Placing AED Drones around Eindhoven, will not only be the cheaper solution, compared to strategically placing AEDs everywhere, it will bring the needed equipment in the most critical minutes of the patients life.

Objectives

Research on how the crowd reacts to a drone landing from above
Create a protocol people should folliw if the drone is in use
Research on how the drone does not disturb the crowd to a large extend.
Improve the landing speed of the drone

USE

User

The users for the AED drone are different than users normally are, as the users from the AED drone are not prepared and is mostly forced upon them. When one comes across a collapsed person who has cardiac arrest, the user has to dail the emergency number and a drone will be send towards the location using the GPS of the person who called the emergency number. After a couple of minutes, when the AED Drone has arrived, further instructions will be provided, as is the norm with an AED. This is one user for the AED drone, the one helping the patient. However, there are more users involved in this matter. Obviously, there is the patient himself. Even though the patient does not use the product, he has the most interest in the product being used well, since it's his life that is on the line. There is also the operator behind the drone who will send the drone and assist in the process of applying the AED after the drone has arrived. Another important user is the emergency services. The drone itself can be considered as a part of the ambulance system. The government and insurance companies are also indirect users of the drone. Also the drone will need maintenance so the people who provide should also be considered as users.

Primary user

People, who are going to make the phone call and provide the needed assistance to the people described on the next line.
People, suffering from heart attack and on which the AED will be used.
People, who will operate the drone and provide assistance to the user who is making the call.

Secondary user

Hospital - The AED drone in function and usability is similar to an ambulance, therefore it is provided that the drone is connected to the hospital(emergency services).

Tertiary user

Government - The government is a big factor as it will introduce the AED drone as part of the healthcare system.
Insurance - The AED drone is going to save lives and is connected to the health of the people, therefore the insurance has a role.
Maintenance - The AED drone is fundamentally a mechanical tool and will need experts who will ensure that it always performs as needed.

Society

For society, this idea has a lot of impact. When a drone is used to help people survive a heart attack or other urgent healthcare problems, there are a number of issues which need to be dealt with:

Privacy. Should the drone be able to access all locations? For example: entering someone’s garden in order to get to the victim. This is very important to the research because we are trying to save lives. However, this can't be achieved without thinking about the people. They need to specify what is acceptable for a drone. Taking a shortcut through another person's garden might increase the chances of survival for a patient, but the privacy of the bystander is at risk.
Crowd control. What are the effects on a crowd of people when a drones passes by at high speeds? For example: a drone could cause a lot of panic within a crowd. Again, the same principle as before applies here. How far can the drone go in order to save lives? It should always be safe, but it should also handle a group of people with enough caution, but also have enough speed to get to its destination quickly.
Parking. How can a parking space be created? How can you get a group of people to make room in an efficient way?

These three main issues focus on society’s opinions and behaviour. This needs to be analyzed by investigating completed studies on a variety of subjects regarding society’s view towards robotics used for healthcare. It is also important to look at the current rules that apply for emergency services such as the ambulance helicopter.

Enterprise

The main enterprises that will be impacted by this idea are:

Insurance companies
Hospitals and government
Drone manufacturers

These enterprises have different interests when it comes to this idea. Whereas the hospitals, government and drone manufacturers might want there to be as many drones like this as possible, the insurance companies might not want this to happen. At the end of the day, the insurance companies main objective is to make profit. Having some expensive drones save lives might be more costly than deploying a “normal” ambulance, which will be more beneficial for the insurance companies. However, the survival of a patient might be worth a lot more than the costs of the drone. A person's life can't really be expressed in terms of money, so there are conflicting interests.

Issues

The ambulance drone is an unmanned aerial vehicle which means that it will need to use a camera in order for it to be operated. The drone will also move in crowded environments and will eventually interact with people. Therefore there are many issues concerning it’s behaviour. There should be rules concerning the ambulance drone, however there are currently no rules specifically for an unmanned ambulance drone, and therefore the drone will have to abide by the rules of consumer or business drones, which will currently prevent this project from taking off for real use. The good side is that there is a possibility that this project can be approved and have exceptional rules concerning it. However, for it to be approved, the drone must overcome dangers and ethical issues which will be discussed in the following sections.

Privacy

The camera itself brings a lot of issues concerning privacy. It is easy for a drone with a camera to gather different kinds of data. If a person is inputting his email on a computer, a drone can easily hover behind him unnoticed whilst recording it. Then it is needless to say that the person who has access to the video can without trouble learn the credentials of this email. Also people do not want to be seen what they are doing in their homes which can be unnoticably done by a drone. Drones can also be used for gathering economical data such as who are the people shopping on a given street ^[1]. The drone’s purpose however is not to use the camera with such intentions and therefore some rules regarding the drones should be omitted and concern the ambulance drone. Even if such a drone may sometimes endanger invade the privacy of a person, that camera is needed for the drone to deliver life-saving functionality which is a reasonable trade-off.

Criminality

According to an article ^[2] a man has shot down a drone which was flying over his yard. That same man has been sued by the owner of the drone and the judge ruled the man to be innocent. However during the case there has been discussion of the height at which the drone was flying. Which leads to the conclusion that if the drone has been flying high enough the man would not be legally allowed to shoot down the drone. The drone is supposed to fly high above the ground and at reasonable speed which means that people who try to shoot down or steal such a drone will not have done so with concern for their privacy, but with violent intentions and should be brought to justice for destroying/stealing government property.

Safety

There are also many issues concerning the safety of such a drone. This is why there are many rules to prohibit drones in many places. Many of these rules have been stated in the code of conduct for drones, however there are a couple of modifications that are needed for the AED.

The current regulations for drone do not allow drones to be flown over cities, buildings and people. However if all the regulations for regular drones are applied to the AED drone the project will be stopped. More importantly this would also be true for ambulances or police cars, but the society has decided that those privileges are a good trade off for a more secure environment. The AED drone will be considered safe to fly over crowded environment and will be operated in such a manner that it will not provide any privacy issues within the society. The purpose of the drone would be to fly from A to B as fast as possible and will therefore cause no problems, apart from when it is landing, but this is a good trade-off in order to increase the survival rate from heart attacks.

Liability

If AED drones were allowed to fly in cities, we get the following problem of who is responsible for the drones. since these drones are autonomous we have no direct control over where they fly, and thus there is always a possibility of the drone crashing into objects. There are many studies about autonomous robots and their liability [reference]. Currently we do not know who is responsible if such a drone crashes, which will be more clear if there are any regulations in place.

So even though there are a lot of problems when it comes to implementing the AED drone into society, an exception needs to be made. When it comes to saving lives, the law suddenly is not that important anymore. For example: an ambulance on the road is allowed to cross a red light at an intersection if it is in a hurry to get to a patient or a hospital. The same principle can apply to the AED drone. If all laws are taken into account, the AED drone simply would not be able to exist. It would be illegal, meaning the project is dead. Since this is a shame and we believe there is actually a future with the drone in it, we assume that the AED drone is allowed to break the law in cases of emergency. What's left to find out is how far the drone can go with this. The next step is to find out what a group of people think is acceptable behaviour for the AED drone in order to save lives.

Theory

- Coronary Heart Disease -

Heart attacks most often occur as a result of coronary heart disease (CHD), also called coronary artery disease. CHD is a condition in which a waxy substance called plaque builds up inside the coronary arteries. These arteries supply oxygen-rich blood to your heart. When plaque builds up in the arteries, the condition is called atherosclerosis. The buildup of plaque occurs over many years. Eventually, an area of plaque can rupture (break open) inside of an artery. This causes a blood clot to form on the plaque's surface. If the clot becomes large enough, it can mostly or completely block blood flow through a coronary artery. If the blockage isn't treated quickly, the portion of heart muscle fed by the artery begins to die. Healthy heart tissue is replaced with scar tissue. This heart damage may not be obvious, or it may cause severe or long-lasting problems. ^[3]

- Coronary Artery Spasm -

A less common cause of heart attack is a severe spasm (tightening) of a coronary artery. The spasm cuts off blood flow through the artery. Spasms can occur in coronary arteries that aren't affected by atherosclerosis.

Common heart attack signs and symptoms include^[4]:

Chest discomfort, mild pain
Coughing
Nusea
Vomiting
Crushing chest pain
Pressure tightness, pain, squeezing or aching in the chest or arms that spreads to the neck, jaw, or back
Dizziness
Dyspnea (shortness of breath)
Face seems gray
A feeling of terror that your life is coming to its end
Feeling really awful (general feeling)
Restlessness
Feeling clammy and sweaty
Shortness of breath

Heart attack risk factors include:

Age. Men age 45 or older and women age 55 or older are more likely to have a heart attack than are younger men and women.
Tobacco. Smoking and long-term exposure to secondhand smoke increase the risk of a heart attack.
High blood pressure. Over time, high blood pressure can damage arteries that feed your heart by accelerating atherosclerosis. High blood pressure that occurs with obesity, smoking, high cholesterol or diabetes increases your risk even more.
High blood cholesterol or triglyceride levels. A high level of low-density lipoprotein (LDL) cholesterol (the "bad" cholesterol) is most likely to narrow arteries. A high level of triglycerides, a type of blood fat related to your diet, also ups your risk of heart attack. However, a high level of high-density lipoprotein (HDL) cholesterol (the "good" cholesterol) lowers your risk of heart attack.
Diabetes. Insulin, a hormone secreted by your pancreas, allows your body to use glucose, a form of sugar. Having diabetes — not producing enough insulin or not responding to insulin properly — causes your body's blood sugar levels to rise. Diabetes, especially uncontrolled, increases your risk of a heart attack.
Family history of heart attack. If your siblings, parents or grandparents have had early heart attacks (by age 55 for male relatives and by age 65 for female relatives), you may be at increased risk.
Lack of physical activity. An inactive lifestyle contributes to high blood cholesterol levels and obesity. People who get regular aerobic exercise have better cardiovascular fitness, which decreases their overall risk of heart attack. Exercise is also beneficial in lowering high blood pressure.
Obesity. Obesity is associated with high blood cholesterol levels, high triglyceride levels, high blood pressure and diabetes. Losing just 10 percent of your body weight can lower this risk, however.
Stress. You may respond to stress in ways that can increase your risk of a heart attack.
Illegal drug use. Using stimulant drugs, such as cocaine or amphetamines, can trigger a spasm of your coronary arteries that can cause a heart attack.
A history of preeclampsia. This condition causes high blood pressure during pregnancy and increases the lifetime risk of heart disease.
A history of an autoimmune condition, such as rheumatoid arthritis or lupus. Conditions such as rheumatoid arthritis, lupus and other autoimmune conditions can increase your risk of having a heart attack.

The Drone in use

The AED drone’s design has be narrowed down to 2 choices:

The AED drone will respond to phone calls regarding heart attacks nearby. It works as follows: A person (close to the person suffering from heart attack) calls the given number. He gets instructions on where the drone will arrive. Parallel to that an AED drone is dispatched to the calling peron’s location. The drone is supposed to autonomously travel at 200 km/h to the target location. However currently there is no collusion avoidance which works flawlessly at the target speed, so the drones may need to be operated manually until that technology is created. When the drone arrives it is picked up by the calling person who is supposed to bring it close to the person suffering from heart attack. The drone itself is equipped with a camera and microphone so the interaction between the operator and the people at the spot is achieved easily. After the drone is placed next to the patient, an able person is supposed to attach the defibrillator in the following way:

Since the AED drone has a camera and operator he is the one who times the shock pauses needed. After a while, hopefully the person suffering from heart attack has recovered, an ambulance arrives at the site. That is when the AED drone is no longer needed, and is returned to it’s base.

State of the Art

AED

The drone should carry an AED. An AED is not very heavy or big, so it should be easy to put this inside the drone and fly with it without any issues.

Navigation

For the purpose of navigating the drone to a specified location, a gps would be needed. Currently GPSis accurate to about 5 meters in the horizontal direction, however the european Galileo navigation system is accurate to 1 meter, and 1 cm for the army. This is more than accurate enough to fly the drone to a destination. However, what GPS cannot do is to modify the navigation based on obstacles in between the current position and the destination. for this, an internal system will have to be used. A camera can be used to see whether there are obstacles in the way, or additionally, there could be some sort of ultrasonic sensing where you map the area in proximity of the drone, and build an obstacle map from there. Since this is a very vague and complex idea, we will not be focussing on implementing this. We make the assumption that the drone is able to fly to its destination without having to overcome any obstacles. The path from A to B can be completed simply by using GPS

Communication

Since the drone operator would have to communicate with the AED user, the drone would need some sort of communication system. This system would include sound recording and image recording, since the operator needs to be able to see what the drone is doing, and where the drone is at all times, as well as assisting the user with the AED. To enable the operator to have access to the video and sound feeds from the drone, the drone would have to be connected to a network where it can send the feeds to the operator. In the Netherlands, the 4G coverage is 80-99% of the Netherlands depending on which network you are. 4G also has a speed of 150 mbit/s which should be more than enough to transmit full hd (1920x1080 pixels) video. This enables the communication of the drone to run smoothly if necessary. In an ideal situation, the drone will be able to reach its destination without any supervision, but if it cannot, the communication should be adequite for both the operator and the user on the ground to cooperate and make the drone perform its task.

Drone Design

It is important that the drone used has a distinctive color since it is an emergency service just like an ambulance. The drone used by the University of Delft is painted in exactly the same colors as a normal ambulance on the road, and we agree with this design. If the AED drone gets implemented into society, the task it performs will be very similar to the job of an ambulance: save as many lives as possible. Therefore it's not that hard to think that the AED drone will be part of the emergency services regarding healthcare, so giving the drone the same color design as the ambulance is the right choice. Another important part of the drone's appearance is the sound it makes. When it comes to this characteristic, we believe the drone should not make the same sound as an ambulance. This is the case because when one hears the sirens of an ambulance, the person immediately has an idea of what to expect. The person expects a car to appear. However, if one hears the same sound approaching but it comes from a drone instead of an ambulance, the person will still expect to see an ambulance. This has a big impact on the person's behaviour, which is not wat we want. One should react differently to a drone than to an ambulance. Also, when one hears the sound, they should know right away what to expect. Therefore, we want the drone to make a clearly distinguishable sound from an ambulance. A sound which notifies the people in the area that something is going on, which a unique melody and tone to it so that it cannot be confused with the sound of an ambulance.

Drone Specifications

Other specs which matter when it comes to the drone are the actual performace specifications. The acceleration of the drone used in the Delft project is 18 m/s2 and its top speed is 135 km/h. This is just the protoype, however. The final product is destined to have an acceleration of 25 m/s2 and the top speed will be 200 km/h. For now, we will use the specs from the prototype, since this is already built and verified. This means the drone will reach its top speed in just over 2 seconds and it is able to reach its destination a kilometer away in just 26.7 seconds when operating at maximum speed. Victims can be reached very quickly with these specifications already, but as mentioned before, the actual product will have even better performance, so even more lives can be saved. The drone needs to carry an AED with it, so it needs to have enough thrust to be able to carry this extra weight with it. The prototype is able to carry a weight of 6 kg's, but the finalized product will be able to carry a weight of 8 kg's. Since an AED weighs only about 3 kg's, the drone could even carry other items on board such as a regular first aid kit.

Personal Space Regarding Drones

Some research has already been performed on the human interaction with drones. One of these researchers is Jessica R. Cauchard. In her article "Drone & Me: An Exploration Into Natural Human-Drone Interaction" she discusses how people interact with drones. What is interesting is that the majority of people interact with drones in the same way as they would with another human or a pet. So even though a drone is not a living thing, it is not really treated like a machine. This is probably because it is able to fly around you on its own without you being in control. This is vastly different behaviour from for example, a computer. When using this, the user is always in complete control and the machine does nothing if the user doesn't tell it to. It also is not moving, which is a big contributor to the living aspect of a machine. Some machines do actually move, cars for example. However, when it comes to cars, the user is in complete control and similarly to the computer, the car does not do anything without the user telling it to. So the combination of (partial) autonomy and the ability to move around gives many users the idea that the drone is a living thing, when of course, it is not.

Similarly to the interaction of humans with drones, some research has been done to look at the interaction of drones with animals. The article "Approaching birds with drones: first experiments and ethical guidelines" from Elisabeth Vas looks at birds being approached by drones. The results were quite clear. 80% of the birds let the drone come closer to them until the distance was 4 metres. Once this mark had been exceeded, the bird would back off or fly away. A result like this was to be expected, since most birds fly away at about the same distance when a human is walking towards them. It gives them enough time to escape in a safe and quick way. However, when it comes to pigeons in the city, they allow humans to come much closer than the 4 metres achieved in the research. This is probably because of the pigeons growing accustomed to humans being near them. With time, the birds will realize that they are not in danger if a human comes close to them. The same thing can happen with drones and birds. If the birds get used to the drone, they may let it come closer than before. And if this is the case, the same result may present itself when looking at human behaviour. Humans learn from experience, so once they see something more often, they grow accustomed to it. In the case of the drones, it could happen that a human lets a drone fly much closer to him or her after a while.

Experiment Plan

Assumptions

Flying the drone by hand is similar to an autonomous flying drone

first of all, the drone should be autonomous, however since we do not have time and the materials, we will simply control the drone "by hand" in this case that means that we use either software available for the pc, or the software available on the mobile phone to control the drone. We assume that faking the autonomy by hand is realistic.

The drone is able to fly to its destination without obstacles

We assume that there are no obstacles in the way such as trees. This is done to make it safer to test the drone, but shouldnt be of much impact to the tests.

The siren is as loud as an ambulance car siren

We assume that the siren system on the drone is as loud as the normal siren on an ambulance. This is done because there is no specification on how much noise the sirens of the drone make. We take the sound of an ambulance, since its a vehicle which has similar functionality to a drone in the case that someone needs help when a person has a cardiac arrest.

The drone is good enough to represent the real AED drone in experiments

Since the drone we use is roughly the same size, we believe that we will get similar results to the real AED drone. We also believe that the level of control of the drone will be very similar to the real AED drone. We do not think that the weight of the drone will impact the tests in any way, since the drones should be able to hover in the same way.

we can use the indiviual behaviour as a basis for the bahivour of humans in a group

Another assumption we made was that we could roughly predict how a crowd would react, by looking at individual measurements. Extrapolate from the individual behaviour what problems would arise, and what would happen if the individuals were put into a group.

The starting position of the drone

Since we do not have the means to test the drone from a far away position, flying towards the test person, the drone will start above the person we are testing on. We assume this will have no significant influence, since when the AED drone will be flying towards it's destination, it will also stop above the destination before it descends.

Personal space with a drone

Setup

Problem statement: What is the vertical personal space of a person in relation to a drone?

Hypothesis: We suspect that since danger rarely comes from above, people are less used to something coming from above, and therefore the vertical personal space is greater than the horizontal personal space.

Aim: The aim of this experiment is to see at what distance, people are still comfortable of having a drone flying above their head.

Equipment & materials: For this experiment we will be using an AR parrot 2.0 power edition drone. The drone has a built-in height sensor. We don’t have access to the actual AED drone, however, we assume that this drone is similar enough that it will produce the same results. Furthermore, we will use a camera to record and analyse the video later on and finally we need some people to conduct the experiment on.

Drone:

Weight: 420 g
Size (length, width): 57cm by 57cm

Number of participants

Participants are informed about the procedure and what is going to happen. They are instructed to tell the drone to stop when they feel uncomfortable/unsafe for the drone to come any closer (stop distance technique)^[5], at which point the drone operator will stop the drone's descent and register the height of the drone from the ground. Before the experiment, the participant is asked to give some general information which might have an influence on the results:

Information beforehand (Participant)

Current field of study/occupation
Experience with drones
Age

The participants are placed standing up in the center. The participant is requested to minimize their shoulder movement or prevent any significant changes in their posture. They are allowed to move their head around freely. The drone is airborne before the participant stands on their spot and the participant will walk to the starting position underneath the drone.

Method: We will be using a within-subjects design (all subjects are exposed to every experiment) in order to allow a direct comparison of distance data from each participant. Let all participants get a close look at the drone in order to prevent that participants let the drone get closer in order to have a better look at it, since the technology is still novel ^[5]. This means that during the experiment, the participant is fully aware of the drone we’re doing the experiment with. The experiment takes place outside, since no suitable open space was found indoors and it increases the realism of the experiment. The space has to have minimal windy conditions, since it could interfere with the accuracy of the experiment, due to drift of the drone. The participant is asked to be standing up during the experiment, and told to minimize their shoulder movement while conducting the experiment. The participant is free to move their head around. The drone will start 10 meters above the participant and will descend at a speed of around 0.2 m/s. The participant is requested to indicate (using the ‘stop distance technique’ ^[6], i.e. say ‘stop’.) the drone to stop when the participant feels uncomfortable. The drone operator will stop the drone once the participant has requested to stop the drone, and will document the height at which the drone has stopped. This experiment is repeated three times on all participants. After all experiments, the participants are asked to express their feeling towards the experiment in order to evaluate if any factors might have influenced the results.

Approach Height: Experiments with drones and birds ^[7] have shown that from an approach height of 30 meters, birds rarely react to the drone and often are undisturbed by the drone from a distance of 4 meters. Other projects have been seen to start at a distance of 7 meters for a horizontal approach. For this experiment, we start at a height of 10 meters and if results show that this is inadequate, or excessive, we will adjust this height accordingly.

Results

Summary

The experiments were conducted on a group of people consisting of 16 persons, which were a mix between male and female. Furthermore, their age was between 50 and 60. The participants indicated that they had little understanding or experience with drone technology. This age group was used as it would show how different people would react, instead of the students, which were used in an earlier stage.

When the experiments were conducted, a few important remarks could be made. There was a difference in height the drone would be felt uncomfortable between the males and females of the participants. After the interview with some of the volunteers, it became transparent that the height difference was due to the boyish nature most men still have. As the women felt the drone was more dangerous, the men showed much more interest in the drone.

When the second round started, it was noticeable that all participants allowed the drone to be closer than it was before. While some allowed just a little bit, quite a few allowed the drone to be an extra halve meter closer. The conclusion for this phenomenon is the unpredictable nature of the drone. One cannot predict the path the drone is going to follow, as it is not limited to directions in only possible way, like a car which can only move forward when it is already being driven. Therefore, when the second drone flight was operated, people started to thrust the drone more and allowing it to be closer to them.

However, when the third round has started, people were getting a bit to overconfident due to their alcohol tolerance. While the drone was yet again allowed to fly even closer, some people let the drone fly in less than halve meter distance. This resulted in some peaks in the data that was acquired. However, the data still showed a descries in distance, as was predicted.

Therefore, when taking the measurements into account, one could conclude what the best way of operation is for a drone. All of the participants allowed the drone to be in a range of less than 5 meters when the first flight was done. After multiple flights, the distance decreased and allowed the drone to be even closer. Therefore, when implementing this data with the AED Drone, one should say that the AED Drone should always fly at a distance of 5 meters from the ground for the first few years, minimalizing the time it would take for a drone to land. The data showed that if people get used to the flight of a drone, they will ease up and allow the drone to be closer. If the AED Drone is in use for a couple years, people will get used to it and therefore allowing it to fly even closer to the ground to make the fastest possible landing, which will be crucial in the golden minutes of when someone is in dire need of help.

Analysis

The experiment was conducted on 16 individuals. The experiment was limited to just a drone coming from straight above the participant, because of inaccurate controls for the drone. For each person, we conducted three trials. Below is a numeric summary of the data:

Numeric summary
	Mean	Standard deviation	IQR	Min	Median	Max	n
All Trials	4.020833	0.7816944	1	2.3	4	5.6	48
Trial 1	4.79375	0.5384778	0.800	4.0	4.80	5.6	16
Trial 2	4.00000	0.3829708	0.725	3.5	4.05	4.5	16
Trial 3	3.26875	0.4867837	0.725	2.3	3.20	4.0	16

Personal space per person
Personal space per trial
Personal space per gender

First, we construct the confidence interval for the mean distance across all data. The confidence interval is constructed as: x̄±z_α/2*σ/sqrt(n). From the numeric summary we see; x̄=4.020833 σ=0.7816944 and n=48. We take a confidence of 95%, which means α=0.05 and z_α/2=1.960. This results the interval: [3.7997 ; 4.2420]

The first result from this boxplot is the seemingly decreasing height per trial. To confirm this hypothesis, we test if we can assume that the means of the trials are the same, i.e. if the difference between trials is significant. Also, the gender boxplot seems to have a difference in distance between male and female participants. This difference is tested by an ANOVA test to test if the means of the two different genders can be assumed to be the same.

Before we test if the means can be assumed to be the same, we first test if we can assume the data follows a normal distribution. To test this, we plot a Quantile Comparison Plot:

Quantile comparison plot

All points in this plot are along the normal line, which means we can assume a normal distribution. Now, we perform an ANOVA test on the trial number to see if there exists a difference between the trials. H₀ = all trials are the same.

ANOVA (Trials)
	df	Sum Sq	F value	p
Trials	2	18.61	41.45	6.19e-11

From this p value (6.19e-11), we reject H₀ and conclude that there is at least one trial which is different from the others. Now that it is known that at least one of the trials is different from the others, we perform a post-hoc Tukey range test to test if the means of the trials compared with each other are significantly different. The null hypothesis in the Tukey's range test states: H₀= 'The means being compared are from the same population, i.e. μ₀=μ₁'

Tukey
	Estimate	SE	t-value	p-value
t2 - t1 == 0	-0.7938	0.1675	-4.738	< 1e-04
t3 - t1 == 0	-1.5250	0.1675	-9.103	< 1e-04
t3 - t2 == 0	-0.7312	0.1675	-4.365	0.000195

We take a confidence of 95%, which means α=0.05.

t2 - t1 == 0: We reject H₀, because p<α.

t3 - t1 == 0: We reject H₀, because p<α.

t3 - t2 == 0: We reject H₀, because p<α.

From this we conclude that the means of the three trials are not equal, which means there is a significant difference of result between all trials. This however does not give us how the trials differ from each other. to find this difference, we use R to create a regression model with response variable 'Distance' and explanatory value 'Trial'. First, using R, we calculate the correlation between the distance and the trial to analyse if the correlation is significant and how they might be correlated.

Correlation matrix
	Distance	Trial
Distance	1.0000000	-0.8048759
Trial	-0.8048759	1.0000000

Since the value (-0.8048759) is negative, there is a clear negative correlation between distance and trial. Since the value is close to -1, we conclude that a linear model is sufficient for this data.

In order to find the actual relation between distance and trials, we create a linear model for the distance and trials. This model will be of the form Y=β₀ + β₁ * x. Here x represents the trials. For the model values, we used R's linear model function:

Linear model (Distance ~ Trial)
	Estimate	SE	t-value	p-value
Intercept	5.54583	0.17907	30.971	< 2e-16
Trial	-0.76250	0.08289	-9.199	5.41e-12

To get a proper model equation, the correct accuracy needs to be applied:

Linear model (Accuracy)
	Value	SE	Accurate Value
Intercept	5.54583	0.17907	5.55
Trial	--0.76250	0.08289	-0.763

Therefore the model is: Y=5.55-0.763x. This therefore clearly shows a significant negative correlation between the distance and the trials. The distance participants let the drone fly above their heads depends on how often they already performed the test.

We now perform the ANOVA for the genders to test if there is a significant difference.

ANOVA (Gender)
	df	Sum Sq	F value	p
Gender	1	5.467	10.82	0.00193

We take again the standard confidence of 95%, which means α=0.05. We therefore reject H₀, because p<α. This means that there is at least one gender which is different from the others and because there are only two genders, we conclude that there is a significant difference between the two genders in terms of the distance to which they allow the drone to come.