PRE2023 3 Group11

Thermal regulation of a fire-rescue robot

Group Members

Name	Student Number	Study
Storm Wilms	1839993	BME
Tessa Groeneveld	1738941	BPT
Abel Galambos	1846647	BCS
Elektra Katsikis	1826654	BCS
Roman Sinickis	1748939	BAP
Tessa Cuijpers	1836927	BME

Introduction & Problem Statement

A fire in a residential building is a common and critical emergency in any big urban area. Apart from the damage it does to the building, there are often people stuck inside whose lives depend on how quickly they are found and rescued by the firefighters. Sometimes, when the emergency services arrive at the scene, the entry to the building is already blocked by fire. The first question that a firefighting crew has, is how many people are inside and where they are. If the entry is blocked, or there are other complications, the search and rescue procedures can only be started after it is safe to enter the building. This delays the rescue and decreases the chances of people trapped inside surviving with every second. Often this leads to either firefighters entering the house even when it is still dangerous, or people not getting rescued in time.

Technology is already being used by fire departments to optimise the process of putting out fires and evacuating survivors. For example, in some situations drones can be used to survey large open areas that are on fire. Technology is also used to allow the head firefighter to track where the fire fighters on site are at all times during fire. We want to provide a tool that can allow firefighting crews to locate survivors even before the fire has died down enough for units to enter the building, decreasing the total time it takes to rescue a person and decreasing the risks for the firefighters themselves.

Our goal is to design a robot that can navigate hazardous high temperature environments, while equipped with sensors and communication technology allowing it to inform firefighters of where possible survivors may be located and how the situation is inside a room where it is still too dangerous for a human to enter. Over the course of this project we shifted from broader design and research to more specific topics to optimise, such as sensors or materials. We investigated multiple possible users and worked in collaboration with both to design a useful and innovative solution. To achieve this we developed a heat shield and gas cooling system that can be applied to a variety of different robots or integrated in new designs allowing robots to navigate more extreme environments.

Objectives

This project will focus on designing a robot that can be used in a fire to find and help rescue people. At the end, the robot should have the following design features:

• The robot must be sturdy and fire resistant to endure the harsh environment during its operation.
• The robot needs a navigation system to find a way through the desired area.
• The robot needs multiple sensors that give intel about the environment in order to find trapped people inside.
• The robot must use a way of transportation that is suited for fires. It should be able to step over or avoid fallen debris that is produced by the fire.
• The robot must be easy for firefighters to use.
• The robot should be as small as possible for it to travel through all locations in a fire.

Given the purpose of the robot as well as its objectives, this project will focus on the design. Additional prototypes could be developed but is not the focus for now.

USE Case

Users

The users are mostly the firefighters using the robots to locate people in a burning building. They need to be able to quickly and easily understand where the robot has found people. The stakes will be high and time is very much of the essence. Another user group is the people who are in need of saving. If they are still conscious, they need to understand the robot is trying to help them, they should remain in the place where the robot found them as long as possible for the firefighter to easily find them. Other helpful tips like stay low to the ground to avoid breathing in smoke can be given to the people in need.

The firefighters require, as mentioned, an easy to understand system. They cannot waste precious time on trying to figure out the cues the robot is giving because this will only interfere with the saving process instead of expedite it. They also need a product that is robust and will not break down in time of crisis, because that would again be wasting time. Another important factor is of course that the robot should not overlook people that can still be saved and should make clear that the firefighters should still keep their eyes peeled for potential victims it might have missed to avoid a mistake that would cost a life.

Furthermore the professionals need to be properly trained in the use of any given robot, but also in the use of robots in general. An idea proposed in the (now quite old) [After Action Report to the Joint Program Office: Center for the Robotic Assisted Search and Rescue (CRASAR) Related Efforts at the World Trade Center, section 4.3] is to provide a prototype that the personnel can train with, thus both giving them a head start in training and granting valuable feedback to the designers of the robot.

According to [Frauke Driewer et al, 2005 TODO] some of the most important jobs of a robot for firefighters are:

·        Exploring and going into dangerous places

·        Detecting the location of people

·        Detecting dangerous areas and hazardous materials

·        Sending information from the scene

And the most desired features to be included were

·        Data transfer

·        Working efficiently at high temperatures

·        Climbing stairs

But Moving and acting without exact instructions and Interacting with the rescue team on the scene were rated as less important features.

In [Harbers et al, 2017 TODO] we can see some of the most important ethical dilemmas that were derived based on conversations/workshops with professionals (in the field of SAR):

1.    “Should SAR robots be employed when they might help saving lives, but their application might also lead to casualties?

2.    Should one develop SAR technology that is intended for peaceful purposes even when it has clear military potential?

3.    Should one replace infield workers by robots if that leads to suboptimal performance?

4.    To what extent should information collected by robots be processed to make it more digestible, at the risk of losing or misrepresenting information?

5.    Should one deploy robots, knowing that this may raise false expectations and runs the risk of degraded performance?

6.    Should one deploy robots that may yield responsibility assignment problems?”

Thus for a robot to be deployed in a live situation it is almost necessary that the developer resolves these dilemmas, either generally or at least for the special case of the robot. Or else the users (specifically firefighters) might not be able to use the robot in good conscience.

The people in need of saving need a robot that does not scare them. It should be immediately be clear the robot is their friend and if instructions are given to these people it should be very clear for them to understand even if they cannot see or hear which is quite likely in a burning building.

Process

Designing our robot was an involved process that involved a lot of research and contact with our users. Initial research was conducted by splitting the robot into various characteristics: transportation, sensors & image recognition, communication method, materials & fire resistance, and navigation & algorithm. This allowed us to get a good state of the art literature review (See Appendix ? TODO) while preparing to consult with our first users, the fire department.

MoSCoW Requirements

Based on the literature review and the interviews with the firefighter and technical expert, as well as our own vision for the product, we have established a number of requirements that the robot we are working towards should meet.

Must Have:

• Ability to detect any human within 10 meters.
  1. Motivation: This is the central idea of the robot, thus it is also a core requirement.

• Be directly controlled by a human operator.
  1. Motivation: this requirement was made crystal clear by the firefighters in the interview.

• A visual feed from the robot to the operator.
  1. Motivation: From both interviews, our own intuition and from literature this requirement seems to go without saying, and our goal is not to break norms.

• Maximum weight of 1.5kg, and maximum dimensions such that it can be carried by a firefighter on their belt.
  1. Motivation: The maximum weight is a clear constraint from the interview with the technical expert, and both interviews suggest that the desired use case is that the robot will be carried by the firefighters on their person.

Should Have:

• Semi-autonomous control.
  1. Motivation: A form of semi-autonomous control could ease the burden on the operator while still complying with the second must-have.

•  Ability to withstand short bursts of heat (namely 600 degrees for about 20 seconds).
  1. Motivation: From the interview with the firefighter we can see that these bursts of heat can  happen, and we need the robot to be resilient so that it can freely go where victims might be.

• Ability to monitor extra information, namely internal temperature, presence of obstacles.
  1. Motivation: This extra information could both be useful to the operator, and it could enhance the robot’s ability to stay operational.

• Ability to display its own location on a map of the building.
  1. Motivation: From the interview with the firefighter we can see that this feature would be appreciated, and having it would also simplify the communication of the location of victims.

Could Have:

• Ability to stay in the fire for extended periods of time.
  1. Motivation: While this could be useful, from the interview with the firefighters it seems that such situation can be avoided quite easily, thus it is nice o have but not necessary.

• Recognize hotter areas and thus the core of the fire.
  1. Motivation: Depending on the sensors used this could be a simple task for our robot, but according to the interview with the firefighters, they can already handle such tasks.

• Ability to navigate stairs.
  1. Motivation: Such a capacity could make the robot a lot more versatile, even if it can only go down or up, as seen in the interview with the technical expert.

• Ability to convey information to the victims inside (i.e. a speaker or display).
  1. Motivation: As seen in our user analysis we want the robots presence to be perceived positively by the victims, this requirement could further aid the rescue by providing reassurance and proper instructions to the victims.

Won’t Have:

• A fully autonomous operational mode:
  1. Motivation: Not only is this a difficult task to achieve, but it is also clear from the interview with the firefighters that this feature is not seen as a positive by the clients.

• Ability to manipulate the environment (move objects/extinguish flames/etc).
  1. Motivation: Implementing such abilities would require a much larger and heavier robot then what previous requirements dictate.

User interface

From the interview, some more insight on how the robot’s user interface should work was obtained. Two things the user was quite adamant about are the easy control, without the use of some type of joystick controller, and the robot not being autonomous at all.

Together with some own research, the elements of the interface were compiled into one design. Firstly, the user requested the control to be through finger movements on a tablet. This mostly signals that a simple way of control is desired, so the person controlling the robot needs no extra training.

We opted for providing two ways of control, the first being reminiscent of Google maps. The user would swipe to or select the place they want to go and the robot would then go to this specified location. However, this might not always work perfectly if for example the robots ‘vision’ is not fully synced with that of the tablet. For this reason a more manual mode of control was added, being similar to that in video games.

In order for the person controlling the robot to actually see where to go or use the Google maps type of control, a camera view is added as the main part of the interface design. However, smoke might obstruct vision and a general overview of where the robot is in the building is always handy. Therefore, a small map showing where the robot is in the building is added into the corner. The firefighters have access to a database of these evacuation maps for larger buildings, so these can be used for this purpose.

A slider to adjust the speed was also added, incase the robot needs to speed up or slow down in specific situations. In order to know how much of an in- or decrease is needed, the current speed is shown in the top left corner.

Above this, the current temperature detected by the robot is visible as well. The user can keep an eye on the temperature in order to avoid areas that will put the electronics of the robot in danger. For example, when above a certain temperature or when it starts increasing rapidly, the firefighter can change routes. However, there is no automatic stop when the temperatures get too high, which links back to the request for no autonomy. This means that the firefighters themselves can decide if the situation is urgent enough to potentially destroy the robot for and can continue going into the hot areas if deemed necessary.

Material Analysis

1)     What are the protection requirements for a robot operating in a firefight?

Just like humans, some parts of a robot are very vulnerable to high temperatures and open flames. As of now the component with the least temperature tolerance is a battery. According to [1]  lithium-ion batteries (LIB) are optimal choice for electric vehicles, including a firefighting robot, due to their high energy density and long life cycle. However, LIB, like other batteries, is very vulnerable to temperatures outside its safe range. This range is from -20 to 55 Celsius degrees, but an ideal working range, to avoid fast degradation of LIBs, is from 20 to 45 Celsius degrees [2]. Therefore, this specific temperature range is set as a goal temperature, at which robot’s electronics must be kept.

Another key characteristic is non-flammability. Some electronic parts, especially the chemicals that batteries consist of, are indeed highly flammable, and as a result exposure of inner components to fire can lead to the immediate loss of the robot. Hence, outer components of the robot must be non-flammable.

Lastly, the robot exterior must be stiff and rigid to withstand potential physical damage by the debris.

2)     How much heat would a robot experience in a firefight? How much heat would robot experience standing directly in a fire?

An accurate estimation was carried out by the article [3] of what temperatures and heat flux firefighters experience on firefighting duty. It is wise to base our robot heat-resistance requirements on this table, depending on the usage of the robot (In particular what level of exposure to fire will the robot experience). The data from the discussed article is in the following table:

3)     How do these requirements translate into components?

First, these requirements directly translate to requirements for the robot protecting cover. This protecting cover must encapsulate all vulnerable parts of the robot, like batteries, to keep them from fire, heat, and physical damage. A composite cover is optimal for this, as it allows the utilize of different properties of different materials simultaneously. Rigid physical strong layer – most likely a metal cover, like galvanized steel, non-flammable isolation layer, heat insulating layer (air/graphite/silicon aerogel).

Second, while direct flames and physical damage can be stopped by exterior protection, there can be no 100% efficient heat insulation, so the heat is going to accumulate within the robot over time. One way to deal with it is to have an active cooling system, which solves the problem, but is hard to achieve.

Another way is to have the rate of inner heating decreased to the point where the time required for heat accumulation to become dangerous would exceed the regular time of operation of the robot. While this seems to be ignoring the problem and requires more complex cover design, this simplifies the whole robot by removing active cooling system. It is obvious that any simplification of a system, especially a system operating in extreme conditions, leads to better reliability. (In other words, failure in cooling system = complete failure of the robot. Find a way to exclude cooling system = one less critical failure possible.)

Experiment

Experiment Plan

The experiment will be carried out in the Netherlands, Technical university of Eindhoven, Innovation Space in

2024, with educational purposes. The goal of the experiment is to determine how much heat the perspective composite material can keep out of the insides of a robot. The materials that the composite will consist of are galvanized steel and ceramic fiber. These materials are sufficient in terms of the material research that had been done and are relatively available and cheap considering the user requirements from the interviews.

During the experiment, the materials will form a shield resembling the heat cover on a robot. Galvanized steel will be positioned on the outside as a protective layer from physical damage and ceramic fiber will be on the inside, as the main heat insulation layer. Around these materials, several temperature sensors will be installed. The materials will then be exposed to a heat source. This can be achieved with a controllable heat source where the heat flux is known. The duration of each segment of the experiment is expected to be between 1-10 minutes. In this time range, the most extreme heat occurs in a burning building and should resemble the conditions close enough. The temperature sensors will monitor how the temperature changes over time in different parts of the composite, which can then be used to see how well the materials can withstand the heat.

The experiment should be conducted with different thicknesses of insulating material, as different thicknesses can help to better understand the insulation capabilities of the composite. The robot heat shield should not be too thick because of size and weight limitations, so finding an optimal thickness that can be used for operation is a must. With the data it is expected to see how high the temperature can get inside a robot protected with such insulation, and how does the heat insulation elongate the time it takes for the robot to heat up. After the experiment it will be possible to make conclusions and a final recommendation will then be formed that says if the composite is a viable solution for robot insulation.

In addition to the heating experiment, as a second part of the experiment the capabilities of a provisional compressed CO2 cooling system are going to be tested in a similar setup. With sensors around the cooling system the CO2 is going to be let out of a cartridge, allowing the gas to rapidly expand and thereby lower the temperature. Two attempts are going to be made with instant and graduate gas release.

Setup of the experiment

The set-up for the heating experiment can be seen on the Image (5.4.1). A galvanized steel sheet (C) 0.75 millimeters thick and a ceramic fiber sheet (D) 10 millimeters thick were combined into a composite heat shield. The size of the steel sheet was 250x500 millimeters and the same size sheets were cut out of ceramic fiber. The steel sheet and the ceramic fiber sheet are intentionally much larger than the area of the interest/the temperature sensor, so that later in the model the heat shield can be considered to have infinite area. The composite was positioned vertically, using the remaining ceramic fiber underneath the composite to protect the table surface. Then a smaller piece of ceramic fiber was attached on top of the big ceramic fiber sheet, to have two layered heat insulation. Three temperature sensors (A), connected to Arduino board, were placed around the setup: first one in front of the steel sheet, to imitate unprotected robot temperature, second one behind 1 layer of insulation and the last sensor behind 2 layers of insulation. Finally, a heat gun (B) from “PARKSIDE” which blows hot air at 350 or 550 degrees Celsius, depending on the setting. On the Image (5.4.1) the other heat gun is visible, but it was swapped for a more powerful one. When the sheets of ceramic fiber were cut, the operators used protective gloves, glasses and respirators, while performing the cutting in a highly ventilated location, as small particles of ceramic fiber are dangerous to inhale. During the experiment the operators used gloves, and care was taken when working with hot air and hot materials.

Image 5.4.1: Photo of the heating experiment setup; A – Arduino temperature sensors; B – heat gun; C – galvanized steel sheet; D – ceramic fiber sheets.

The set-up for the cooling experiment can be seen on the Image (5.4.2). A cooling system consisting of a cartridge with pressurized CO2 (C) and a manual gas valve (B) was covered with ceramic fiber sheets (D) to imitate the inside of a robot. Three temperature sensors (A) were positioned ass followed: one touching the cartridge to know its temperature, another one near the cartridge to know the temperature inside a robot, and the last one was held above the manual valve exit hole, to know the temperature of exiting CO2. Galvanized steel sheets (E) were placed around the ceramic fiber to cover the surface of the table and to put some pressure on top of the ceramic fiber sheets, to secure the position of temperature sensors. Again, the operators of the experiment used gloves, when touching the fiber glass and cold metal cartridges

Image 5.4.2: Photo of the cooling experiment setup; A – Arduino temperature sensors; B – manual gas valve; C – cartridge with pressurized CO2; D – ceramic fiber sheets; E – galvanized steel sheets.

Experiment Results

Image 5.4.3: Graph of the heating experiment with heat gun set to 350 ^oC temperature;

Image 5.4.4: Graph of the heating experiment with heat gun set to 550 ^oC temperature;

By analyzing the linear parts of the slopes of image (5.4.3) and image (5.4.4) it was possible to compare the rate of temperature rise in unprotected scenario, with 1 or 2 layer insulation. The results of these calculations are visible in table (5.4.5)

Table 5.4.5	Heat gun set at 350 oC			Heat gun set at 550 oC
	Unprotected	1 layer insulation	2 layer insulation	Unprotected	1 layer insulation	2 layer insulation
Temperature rise rate +-0.001 [oC/s]	0.326	0.110	0.010	1.076	0.197	0.038
Time it takes for the temperature to rise by 1 Celsius degree [s]	3.0	9.1	100.0	0.9	5.1	26.3

From the table (5.4.5) it is seen that heat shield with 1 layer of insulation decreases the rate of temperature rise 3 times with incoming air at 350 ^oC and 5 times with the incoming air at 550 ^oC. A heat shield with 2 layers of insulation decreases the rate of temperature rise 33 times and 29 times respectively.

Image 5.4.6: Graph of the cooling experiment with the instant gas release

Image 5.4.7: Graph of the cooling experiment with gradient gas release

By analyzing the graphs on the images (5.4.6) and (5.4.7) it is seen that graduate gas release allows for better cooling, but even theoretically it is much harder to implement than instant gas release.

Image 5.4.8: Graph of the heating experiment with heat gun set to 350 ^oC temperature, with different axis range;

Image 5.4.9: Graph of the heating experiment with heat gun set to 550 ^oC temperature, with different axis range;

Finally, by changing the axis range of the original graphs of the heating experiment, it is possible to obtain graphs on images (5.4.8) and (5.4.9), which are much more handy to compare the experiment with the model, which is going to be introduced in the section 6.

Cooling System Analysis

During our experiment a cooling system was also tested. Namely using compressed CO2 cartridges. The cartridges that were used, are usually used to pump up inflated tires on the go. They are small, light (in total 48 grams, which includes 16 grams of CO2), widely available and quite cheap. Therefore they would be quite ideal to use. They are made by pressurizing the CO2 to about 70 bars which increases the temperature to about 400 degrees Celsius. After this the gas is cooled down enough to liquify it (which will be around room temperature) and this liquid is stored in the cartridges. When one opens these cartridges the pressure abruptly changes and the liquid turns to gas and escapes rapidly. This change in pressure causes the gas to expand and cool off, because the gas that was already at room temperature is now taking up a lot more space, making it cooler. So all that needs to be done to rapidly cool the surrounding air is turn the nuzzle of the cartridge.

This system seems to be very ideal, but a big downside is that since it is stored under high pressure it can be dangerous to heat it up. Generally, it is advised not to heat it above 50-55 degrees Celsius, which would be a big problem in the use case of fire robots. There could be a temperature sensitive valve that opens the cartridge (partly) if it gets too hot to keep the temperatures low but since 50 degrees is lower than what most of the electronics can handle (~75 degrees Celsius) it seems quite wasteful to try to keep the temperatures lower than necessary. The negative effects of the cartridge heating up too much are also too big to risk it, since this means the cartridge would explode and/or shoot away at a rapid speed due to the gas coming out very fast.

In conclusion the cartridges do work to cool down the electronics as shown by our experiment, but unless the hazards that come with using them can be worked out, it does not seem like a viable option for a robot that has to deal with very high temperatures.

Experiment Reflection

While the explement allowed to prove that a composite made from the chosen materials indeed can be used as a heat shield and can increase the lifetime of a robot inside a building on fire from 3 to 5 times, some aspects of the experiment could be improved.

First of all, our temperature sensors had a limit of 125 Celsius degrees, above which the sensors could have been damaged. Hence in graphs (5.4.3), (5.4.4) and (5.4.9) the abruption of the "unprotected sensor" temperature is seen, as at that point operators removed the sensor in front of the heat shield for it not to get damaged. The experiment can be repeated with a more heat tolerant temperature sensors, or, for example, a pyrometer for monitoring the temperature of the steel sheet.

Secondly, if you compare the experiment results and the data obtained from the model (section 6), you can see a significant difference in temperature rise rate behind two layers of insultation. The experimental setup was surrounded by room temperature air, which provided heat-loss from the sensors, while the model does not account for heat loss. In that sense the model is mush closer to the reality, as in a building on fire a robot is surrounded by hot air. We tried to minimize the effect of the air, by covering the sensors with yet another piece of insulation from the cold side. However, the temperature sensor "behind 2 layers of insulation" had less insulation from the room temperature air, in comparison to the sensor "behind 1 layer of insulation". Hence, the results from the experiment for 1 layer insulation are much closer to the model and to the true value, while the results for 2 layer insulation can be discarded. Adding a better insulation on the cold side of the set-up would result in more accurate results.

Sensors & Image Recognition

When looking into different ways of detecting survivors, one specific method seems to pop up most frequently. These are UWB sensors, UWB standing for ultra-wideband. It is a form of radar sensing, using a different range of frequencies than normal radar sensors do. RADARS, short for Radio Detection and Ranging Systems, send out electromagnetic pulses. These pulses are reflected by objects and transmitted back to the sensor. This way the system can detect where the objects are [4].

Usually, these operate on a rather narrow frequency range with a high energy output, whereas UWB radars use a much broader range of frequencies and produce a lower energy output. This broader bandwidth and lower output energy lead to more accurate detection and higher resistance to multiple types of interferences. The operating range is rather small though, as the maximum reach of these systems is usually around 10-15 meters [4].

However, this type of sensor has one major advantage over other ways of detecting potential survivors, namely that it is immune to obstacles. When using AI image detection or any other type of visual-based detection method, obstacles or walls will prevent the system from detecting survivors, as it cannot see the survivors. This is either a huge limitation, as it does not find all persons, or a poses a big challenge for the robot to go to each place in the building to make sure no one is missed. As for infrared detection, the heat from the fire obviously makes it almost impossible to detect where humans are. UWB on the other hand, detects movements, even as small as respiration movements. Furthermore, this motion detection can even sense these small movements behind walls or other objects. Through testing it shows that this method is quite accurate (93%<) for presence testing, as well as for no presence testing (89<%) depending on the distance from the object. [4].

There have been multiple other papers concluding that UWB is successful for detecting humans in complex environments [5] or behind walls [6], even with more low-cost, light-weight applications [7].

A possible drawback for this method might be the temperature range in which it can operate, but it is extremely hard to find information on this topic. Therefore, more research should be done on this.

A different method might be the use of search algorithms to find potential survivors. Despite not going into the building, these search strategies show promising results in determining the locations of possible survivors if the map of the site is known [8]. This method might be less accurate than UWB, but it does not have to deal with high temperatures as it does not have to enter the building. The map of the site could be determined by getting footage from a drone.

After discussing with our team, it was concluded that the robot would need multiple sensors for different purposes.

- UWB sensor: uses radar to detect motion, even as small as the breathing of a human. By comparing data on the general breathing pattern of humans with the found signal, a conclusion can be made on if there is a survivor near or not. This sensor can usually go through any material except for metals.

- Camera: as it was concluded that the robot needs to be controlled, a camera can be used for the operator of the robot to get a clearer picture of where to go. Obviously, this is very limited due to things like smoke, but can definitely help at times.

- Ultrasonic sensor: As the UWB sensor is capable of detecting through walls, it cannot detect the walls itself. In order to detect any obstacles/walls for navigation, the camera has been installed. However, when vision is obstructed this obviously does not work. Therefore an ultrasonic sensor can be used to sense where walls or any obstacles are, so that they can be avoided.

- Temperature sensor: As creating a perfect heat shield is impossible and there are limits to what the electronics of the robot can handle, a temperature sensor can come in handy. The current temperature detected will be visible on the interface of the operator. So if the operator sees that the temperature keeps increasing as they proceed to go in a certain direction, the person controlling the robot knows to go a different way to protect the robot from unnecessary damage.

In order to properly implement these sensors, some things need to be known. The temperature range, power consumption, weight, and some other things are important. It’s very hard to find specifics for the general sensors, as the variables change per type/brand of sensor. Some research was done to find some specific sensors that seem suitable and they are presented with their properties below.

UWB sensor:

-       Qorvo DWM1001C: In many papers, UWB sensors are tested in static conditions. However, the robot will move, thus a sensor that has been researched/tested under dynamic conditions is the DWM1001C [7]. It’s a low budget option for about 50 euros. According to the data-sheet, it has a operating temperature range of -40 up until 85 degrees Celsius, needs an input of 2.8-3.6V and has dimensions of 19.1 x 26.2 x 2.6 mm. The exact operating range is hard to find, besides the minimum of 10cm, but it will most likely be similar to that of other UWB sensors (10-15 meters). The weight is not listed in the data-sheet, but will most likely not be an issue when considering it’s size.

-       Novelda X4: This sensor comes in at a slightly higher price of about 80 euros. However, this is at a minimum quantity of a 1000 units, whereas the Qorvo can be ordered singularly. It has an operating range of about 10 meters maximum, dimensions similar to the Qorvo, voltage input of 1.8-3.3V and a temperature range of -40 up until 85 degrees Celsius.

However, these sensors are designed to detect other UWB sensors and do not give a raw signal and are therefore not suited for this purpose. Sensors that are useable are extremely expensive and hard to get, therefore the testing of the sensors was not chosen as the focus of the project.

Navigation & Algorithm

After our interview with the fire department we did more research and made design decisions regarding this aspect of the robot. From the interview conducted it is crystal clear that the robot should not be autonomous, but should instead be controlled by a human operator. As such three concrete methods of manual control were decided on by the team. In order of importance these methods are:

1. Full manual control
2. PointCom aka ‘Google maps street view’
3. Draw on map

Full manual control was deemed the most important control method as it is guaranteed to be an acceptable method for the users, and it is the simplest both to implement and to understand. With this method the operator can directly tell the robot to move forward/backward or to turn etc. (Think FPS control from videogames)

Pros:

• Tried and tested.
• Most everyone is familiar with at least the concept.
• Easy to implement.
• Greatest level of human control (and thus ability and responsibility are inherited from operator).

Cons:

• High mental load for operator.

PointCom was determined as the second most important control method. This method involves the operator clicking on a point (presumably on the ground) in the robot’s FOV and the robot autonomously going to that point. (Think google street view)

Pros:

• Robust against input lag.
• Reduced mental load for operator.
• Preferred method by users^[12].
• Easy to learn^[12].
• Already shown to work well on real outdoor-robot^[13].

Cons:

• Hard to implement (needs good image processing^[13]).
• Impacted by low visibility.

Draw on map was chosen as the final method that could be implemented, it involves the operator drawing the path the robot should take on a map of the environment, after which the robot will follow the path.

Pros:

• Robust against input lag.
• Reduced mental load for operator.
• Can plan ahead.
• Already have been proven to work (though in a much simpler environment)^[11].

Cons:

• Needs a map of the area.

• Needs accurate location of the robot. (though this might be needed anyways for locating victims)

For both options 2 and 3 the robot would need to be aware of potential obstacles in its path. While this is a potentially challenging task, it is also an issue that has already been solved numerous times through numerous means^[14][19][20]. Furthermore it has been demonstrated that such an algorithm can serve as a tool for mapping and determining position too^[14].

In case of low visibility there may be a need to use alternative method of obstacle detection, such as ultrasound sensors, this needs to be further investigated in knowledge of the capabilities of the chosen sensors.

The final challenge that was investigated (mostly before it was determined that the robot should not be autonomous) is the issue of self-guided navigation in an unknown terrain. For both of the chosen semi-autonomous navigation techniques we will need some path planning to be done by the robot. While there are many proposed solutions available for this issue^{[15][16][17][18]}, I recommend using D* or D* Lite^[15] as this is the same algorithm mentioned by reference^[13], and it has a detailed, well analysed and well explained (fairly simple) algorithm in reference^[15].

Communication Method

A firefighting robot needs to be able to communicate the information it is collecting back to firefighters that are either offsite or outside the fire environment. Research needs to be done on how to effectively and safely allow this communication to occur.

The communication between the robot and the firefighters needs to allow data, possibly video and audio data, to be sent from the robot to a computer wirelessly. This one way communication is the bare minimum for our prototype to function, however depending on how autonomous we decide to make our robot it may be necessary to also send commands back to the robot, instructing it what actions to perform and how and where to move.

Below are a variety of researched and analyzed methods and a corresponding article showing an example of how the technology is used in a similar system. One of the following will be chosen as the most appropriate method for our prototype.

TCP/IP: [21]

Bluetooth: [22]

ZigBee: [23]

MQTT: [24]

ROS: [25]

Following our user study consisting of an interview with a firefighter, we narrowed down the scope of our robot considerably. The firefighter had informed us that they would not want the robot to be autonomous, but rather remotely controlled. Additionally, they requested visual feedback from the robot on a tablet operated by a fireman outside of the fire.

With this additional information, we can narrow down the communication systems to be put in place in our prototype.

When considering wireless communication between a drone and a tablet in a highly dangerous environment, communication delay is an important factor. The off-site firefighter will be remotely ensuring the drone not only moves to where they desire, but also ensuring the drone can avoid flames, navigate gusts of steam, and dodge falling debris. For this to go smoothly communication delay must be at a minimum.

Researching how to minimize delay lead to the conclusions that the most important factors are a low latency connection, minimizing data transmission, using real time operating systems, and optimizing control algorithms.

After reading multiple scientific articles I’ve excluded Bluetooth and decided radio transmitter data-links are the preferable method both to control the drone and to stream video data back to the firefighter controlling the drone.

Transportation

The robot that is going to be designed in this project needs a way of transportation. Robots in general have a lot of ways to transport like walking, rolling or a snake like movement. To determine what is best in the case of a fire rescue robot, pros and cons of each transportation method are needed to evaluate the best possible option. It is also important to keep in mind the environment that the robot operates in, since the robot needs to be able to face a challenges in encounters, such as fire and falling debris.

Walking

A robot can achieve a walking motion if it is equipped with legs. Examples of such robots can be seen below and are developed by Boston dynamics. These robots use either two or four legs, depending on the application. But there also exist robot with more than four legs, to increase stability.

Boston dynamics robots

Pros:

• A walking robot can be quite fast and versatile, which enables the robot to navigate efficiently through the burning environment.
• Legs can be made from strong and durable materials that are fire resistant.
• Having a walking motion makes adapting to the environment quite easy. Stepping over falling debris is possible as well as walking over direct burning surfaces.
• The body of the robot is raised above the ground by the legs, protecting the important electronics and equipment on board from burning surfaces.

Cons:

• Creating a walking motion in a robot is quite hard on both a mechanical and electrical level.
• The legs of the robot come with certain dimensions that could make the robot bigger than desired. It will be hard for firefighters to carry such a robot to the scene.
• A leg could get damaged in the action, making the robot potentially completely unable to move.
• Development of a walking robot will be more costly.
• Weight distribution is incredibly important and it could bring risks of falling over. This would need to be perfect in order for the robot to do its job without problems.

Rolling

Other than walking, a lot of robots use wheels to transport. There are mainly two ways to do this, namely with wheels or tracks. Having wheels makes the robot able to move very fast, like a remote control car. It is needless to say that this way of transportation is quite efficient, fast and applicable in many situations. In the figure below a robot can be seen that was already designed to do similar tasks that this project requires, which shows that wheels can be a feasible transportation method. Besides wheels, a robot could have tracks. This is similar to vehicles like a tank. Tracks can support a lot of weight and can travel on a lot of different surfaces. Implementing a rolling like way of transportation has a lot of room for innovation and can be achieved in a lot of different ways.

A two wheel rescue robot

Pros:

• Rolling can be fast and reliable.
• Using the right materials, wheels or tracks could be able to withstand burning surfaces, thus being able to drive through direct fires.
• Rolling is very versatile and can get the robot all over the area very fast.

Cons:

• Having wheels would bring difficulty with clearing big obstacles, like fallen debris. It could get stuck or unable to move a certain direction since it is blocked. Tracks could be better to solve this problem.
• Wheels and tracks are prone to wear and tear, which would increase maintenance.
• Such a system requires a lot of energy, so the longer the robot needs to operate, the bigger the battery it would need on board.

Snake like movement

Another way that a robot can move itself is by recreating the movement of snake or worm. This way, the robot can be designed to reach narrow places. Although this method is less common that the previous two methods, there is still a lot of research involving this transportation system.

Pros:

• It can reach small and narrow places that are potentially obstructed by the environment, which other robots/firefighters would not be able to get to.
• Can easily adapt to the rescue area, going over fallen debris or navigating around obstructed paths.
• The size of the robot can be changed quite easily to fit the given fire hazard, by connecting multiple robots together. (If this is feature that the robot has).

Cons:

• This is hard to design since it is less commonly used in robots.
• It would require a lot energy, posing the same problem with have a rolling system explained earlier.
• The robot will be fully on the ground, therefore, the robot must be well protected from direct fires and should be able to be in a fire for longer periods of time.

Flying

Another method that can be considered is a robot that does not drive or walk, but can fly. A drone could potentially be used to fly through the area searching for people. Drones are becoming more and more popular in all kinds of applications, therefore, a lot of different drones already exist and there is a lot to choose from. Drones are already being developed to assist firefighters. One drone could withstand 200 degrees Celsius for ten minutes without losing any functionality (https://www.advancedsciencenews.com/a-heat-resistant-drone-that-can-fly-into-fires/). This is of course not enough, but shows that heat resistance can be achieved in a drone. However, flying with drones in an enclosed burning area can bring some problems with it, as can be seen in the pros and cons below.

Pros:

• Drones are fast and versatile.
• Drones can avoid flying through direct fires because of its mobility.
• With drones, locating people can be a bit easier since they approach the scene from above.
• A drone can enter a building from many entry points, like doors and windows.
• Drones can also fly over the burning area, assisting firefighters to locate fires all around the perimeter. So a drone is not only for rescue, also for general intel.

Cons:

• Drones are usually not fire resistant, so different materials have to be used, especially for the propellors.
• Flying a drone inside for instance a house can be very hard and precise movements need to be made to avoid hitting something. Either an automated system would need to be developed or it can be remote controlled, but firefighters would need to be trained to fly a drone.
• The effects of the propellors that come from flying with a drone can either be good or bad. The wind could calm down the fire right below the drone or it could enhance the fire. More research/experiments would need to be done for this.

Transporting the robot

Before it was seen in what way a robot can move itself, but that is not the only thing that needs transportation. The actual robot needs to be delivered to the designated area, thus needing transportation. Firefighters bring a lot of equipment to the scene and have big trucks to store al their stuff. It is only the question if there is space left for this robot (This will become more clear after the interview). Therefore, it is important to keep the size of the robot as compact as possible to be easily transported and carried by the fire department.

If we take a look into other applications of robots, like the police department, a lot can be learned from them. Bomb detecting robots are commonly used to protect police officers against dangers. These robots are part of a special division inside the department known as bomb squads. The robots are brought on scene in an extra vehicle with more equipment. It can be seen that this is not particular efficient for the fire department. If they would need an extra vehicle to transport this robot, more people and money is needed for more operations. Thus, a solution is needed that figures out how a robot can be deployed when it is needed.

Conclusion

There are a few different ways a robot can move itself, like walking, rolling, slithering or flying. Each of these methods has its ups and downs and can be considered in the design. It is a matter of discussing which method would be the best fit for our requirements and how this method influences the other components of the project. A solution for deploying the robot by the fire department is also needed.

Week 3: Decide which method is most appropriate for our design

Deliverables: Concrete decision with justification

Multiple transportation methods have been discussed and their pros and cons have been analyzed. It is now time to choose one main method that is going to be implemented in the design of the robot. The way this is done is by looking at what the robot needs to be able to do and what is best for the user. An interview was conducted with the fire department of the TU/e which gave more insides into what is expected from this robot.

Before the findings of the interview are used into the decision of transportation method, it was already decided that the method was either going to be driving or flying. These methods are most commonly used in robots and are thus familiar to a lot of people. Furthermore, these methods are also the most convenient in the environment that they need to operate in, because of their high mobility. The other methods, namely walking or crawling, were disregarded because of their difficulty to design and operate. They could be effective but given the time for this project as well as the other features that need to be developed, it is a little bit outside the scope of this project.

One of the most important things that were found from the interview was in what degree the robot should be autonomous. It became clear that the fire department wants a robot that can be completely controlled by hand. The robot should be fully under control and do the things that the firefighters want, to not waste any time. Thus, the robot will be fully under control by the firefighters. The best transportation method for this would be driving. A drivable robot can  easily be operated by anyone without much training. Many people have had a remote control car in their youth for example. Flying on the other hand, is way more difficult. Properly flying a drone for instance is not a piece of cake. Therefore, firefighters would need special training to be able to fly the robot which requires more work and money.

Another problem with flying is the environment. Flying a drone inside a building has a high risk of bumping into walls and other stuff, disabling the robot during its employment. Such a risk is not desired. Also, during a fire the air quality significantly changes very fast inside a building. A drone will then have a harder time to stay in the air. As operating a drone becomes more challenging, it raises doubts about whether a drone would remain feasible for this robot.

It was also mentioned during the interview that firefighters keep doors and windows closed until the fire is extinguished. If a robot needs to get inside, a door can be opened very fast to get the robot in but that is all. A drivable robot could easily be put inside the building in seconds, but a drone will be much harder to get inside that fast.

All in all, it seems that a drone is way harder to design and use by the fire department. Although a drone has a very high mobility and can fly over fallen debris, a drivable robot is more reliable and easy to use. That is why a drivable robot will be the main focus in this project.

Driving: How?

Driving a robot has a lot of advantages that already have been discussed earlier. It is easy, reliable and effective. The major concern that driving a robot has is what happens when it faces a obstruction. In a building in general, there could be stairs and other height differences. When that building is on fire, more debris can fall down and will obstruct the robot even more. When possible, the robot should drive around the obstructions. But when that is not an option, the robot should definitely drive over it. The transportation method should thus be able to reflect that.

To create a robot that can go over obstacles, a connection was made to other machines that could do that. A military tank came to mind. Its tracks are designed to not only support the weight of the tank, but also enable the vehicle to drive on rough terrain. These tracks are very effective in clearing obstacles. Another advantage about these tracks is that there is a wide material choice for them. They can be made out of steel or rubber. Therefore, these tracks can be made out of material that is fire resistant.

Conclusion

All transportation methods were put under a microscope and it was quickly discovered that the robot would either fly or drive. Flying a robot inside a building is quite a risk, especially if the pilot is not experienced. Because of this risk and complication, driving will be the main transportation method. In more detail, tracks are going to be used to drive. Tracks are durable and able to drive over obstacles. The next step is to realize this transportation method.

Week 4: Change of plans

In the meeting it was discussed that choices have to be made in order to narrow down the scope of this project and actually create some deliverables. Initially it was planned to design a full robot with its software, but that will take too long. Instead, existing robots could be used as a base where additional hardware can be installed on.

In the previous research to what transportation method was best to use, a conclusion was made to design a driving robot. It has a lot of advantages but faces problems when it comes to clearing obstacles. The best way to clear an obstacle with a robot on the ground would be to step over it, not drive over it. However, a walking robot was disregarded before because of its complexity to design. A walking robot is favorable in a lot of ways in comparison to the other transportation methods and would be the chosen method if not for its complexity. But with this new approach to the project, an existing walking robot can actually be used as a base for the robot. This means that the transportation method will be changed to walking and an existing robot has to be found.

What walking robot should be used?

When it comes to commercially available walking robots, there is one model that really stands out: Spot from Boston Dynamics. This quadruped robot is designed to help in all kinds of situations: Monitor and collect data, safety and security, industrial automation and rescue missions. Spot is a highly mobile robot with a top speed of 1.6 m/s. It can climb stairs and walk on very rough terrain without tipping over. If it does happen to lose balance and tip over, Spot can get itself back on its feet. Overall, Spot is highly versatile and a very good candidate as a base for a rescue robot.

Furthermore, Spot is designed to allow add-ons to its body and software. More sensors can be installed on the robot to create more functionalities. That means that a radar sensor could be added for the search of people in burning buildings. Spot is also equipped with mounting rails on its back to carry payloads up to 14 kg. This means that more hardware could be added, as well as fire resistant materials. These fire resistant materials are definitely needed since the operating temperature of Spot is only 20-45 degrees Celsius.

To conclude, the focus of this project has steered into the materials and fire resistance of the system. That means that more research will go into how a robot can be protected in a harsh environment like a burning building. The findings from this research can than be used to improve a robot like Spot in order to perform rescue operation alongside the firefighters.

Users research

Week 2 + 3 interview fire department:

Introduction interview:

Thank you so much for participating in this interview. Have you had time to read and understand the informed consent form? Great, this interview will last for about 15-30 minutes and it will help us get a better understanding of our user group. We can stop the interview at any time and you can quit your participation at any time. All the answers will be analysed completely anonomously and nothing can be traced back to you as an individual. Please feel free to elaborate on your answers and we are open to any suggestions that you have.

With your consent, I would like to record only the audio of this interview, we will transcibe the audio completely anonomously and delete the audio immediately after the transcription is done and no one will hear it except me and my group member to transcribe it, is this okay with you?

The product we are designing is meant to assist firefighters, it will do this by going into a burning building and locating any people still in there after this it will report these locations to the firefighters so they can rescue them more efficiently. The way this will all work is something we are working on now and your input will be very useful in this process. Are there any questions before we start?

Dutch: Heel erg bedankt dat u mee doet aan dit onderzoek. Heeft u de tijd gehad om de consent form te lezen en begrijpen? Fijn, dit interview gaat 15-30 minuten duren en gaat ons helpen om onze gebruikersgroep beter te begrijpen. We kunnen dit interview stoppen op elk moment en u kunt stoppen met meedoen aan dit onderzoek op elk moment. Alle antwoorden zullen compleet anoniem geanalyseerd worden en niets kan naar u terug geleid worden als individu. Voel u alstublieft vrij om uit te breiden op uw antwoorden en we staan open voor alle suggesties.

Met uw toestemming, zou ik graag de audio van dit interview willen opnemen, we zullen deze audio zonder namen te noemen overschrijven en het bestand daarna verwijderen. niemand zal deze audio horen naast ons groepje om het over te schrijven. Geeft u hier toestemming voor?

Het product dat we aan het ontwerpen zijn is bedoeld om brandweermannen en vrouwen te helpen, dat gaat het doen door mensen te vinden in een brandend gebouw en dit rapporteren aan de brandweer zodat ze hen gerichter kunnen redden. De manier waarop dit precies gaat werken, is waar we nu mee bezig zijn en uw input gaat heel belangrijk zijn in dit proces. Zijn er nog vragen voordat we beginnen?

Questions:

1. What is the protocol for finding people in a burning building? Please talk me through how this process works.
   • Wat is het protocol voor mensen vinden in een brandend gebouw? Neem me alsjeblieft mee in hoe dit process werk.
2. What is the most time-consuming process in search and rescue during a fire in a building? ( In the sense like – is it lowering the fire intensity/searching for people/carrying them out of the building that is the most time-consuming? )
   • Welk onderdeel van het proces duurt het langst tijdens het zoeken en redden van mensen in een brandend gebouw?
3. Around what temperatures are you usually dealing with when there is a house on fire? What about if it is a special lab facility or chemical factory?
   • Met welke temperaturen moeten jullie meestal werken in een brandend huis? En welke temperaturen in een lab of chemische fabriek?
4. Do you think a robot that helps firefighters locate people in a burning building could be useful to you?
   • Denkt u dat een robot die u helpt door mensen te zoeken in een brandend gebouw voor u nuttig zou zijn? Waarom?
5. Do you have a vision of what a perfect version of robot like this would look like? Explain?
   • Heeft u een beeld van hoe een perfecte versie van deze robot eruit zou zien? Leg uit?
6. Would you rather have a robot that finds people on its own or that is controlled by someone outside the building? Why
   • Zou u liever een robot zien die uit zichzelf rijdt en mensen zoekt of eentje die door iemand buiten het gebouw wordt bestuurd? Waarom?
7. If there is a robot that can find out the exact location of people trapped inside a house on fire before you enter it, how would you want the robot to tell this information to you – map the people on the floor map, or lead you to the people, or You have another idea?
   • Als er een robot bestaat die zelfstandig mensen voor jullie vindt in een brandend gebouw, hoe zouden jullie de informatie over de locatie van die mensen graag willen ontvangen van een robot? (Denk aan een plattegrond of dat de robot je ernaartoe leidt)
8. Do you already know of products that help you find people in a burning building?
   • Kent u al andere producten die jullie helpen met mensen vinden in een brandend gebouw?
9. Are there any ristrictions in terms of size and shape for the robot?
   1. Zijn er ristricties in de maat en vorm voor de robot?

Closing interview:

Thank you for answering all our questions, this will really help us to design a robot that will be as useful as it can be. If you have any further questions, do not hestitate to contact me. This is the end of the interview. (and I will end the recording now)

Dutch: Dankuwel voor het beantwoorden van onze vragen, dit gaat ons erg helpen met het ontwerpen van een robot die zo nuttig is als hij kan zijn. Als u nog meer vragen heeft, twijfel niet om me te benaderen. Dit is het einde van het interview. (en ik ga de opname nu stopzetten)

Summary/findings:

The main and most important finding from the interview was that the interviewee had a clear preference for a man controlled robot. He said it would also help them understand the situation inside better if it was controlled and the robot would have a camera to show the operator what it can see. This would improve their situational awareness in addition to finding people. He said this was preferred over autonom even if the people outside operating it were already quite busy. Another important finding was that this robot could be very useful since the firefighters currently do not actively look for people while the fire is not completely controlled/gone. He said this is the first priority and once this is done they will start looking, which leaves a gab for the robot to look for people while everyone else is busy inside. He also mentioned they often have the evacuation floorplans you see hanging in big buildings at their disposal before they enter a building. This not include most residential buildings of course, they usually do not have a floor plan or anything like it for houses.

The firedepartment also currently uses heat cameras to assist in the search of people but mostly the core of the fire. In addition to that bigger departments also have a drone team, however these drones mainly focus on flying around the perimeters of the building and do not actually go inside.

Week 5 + 6: Interview with Danny Hameeteman

Introduction interview:

Thank you so much for participating in this interview. Have you had time to read and understand the informed consent form? Great, this interview will last for about 15-30 minutes and it will help us get a better understanding of the technology for our product. We can stop the interview at any time and you can quit your participation at any time. All the answers will be analysed completely anonomously and nothing can be traced back to you as an individual. Please feel free to elaborate on your answers and we are open to any suggestions that you have.

With your consent, I would like to record only the audio of this interview, we will transcibe the audio completely anonomously and delete the audio immediately after the transcription is done and no one will hear it except me and my group member to transcribe it, is this okay with you?

The product we are designing is meant to assist firefighters, it will do this by going into a burning building and locating any people still in there after this it will report these locations to the firefighters so they can rescue them more efficiently. Based on a previous interview with a firefighter we decided that the robot should be remote controlled. They also said to focus on improving their situational awereness via the robot. How to do this and how to make the robot as heat resistance as possible we are working on now. This interview will hopefully help us get a better understanding of how to do this. Are there any questions before we start?

Questions:

1. Which robot is more appropriate for fires, jens or spear for our use case, and why? What are the differences between the two?
2. What are its mobility limits of these robots? And how much load can it carry?
3. How were user requirements implemented?
4. How do user requirements differ per hazardous situation and how does this impact the design?
5. What are the main obstacles for this type of robot? How should these be handled?
6. What sensors are used and why?
   1. It seems the robot mostly gets video footage, are there any other ways that could be interesting to create better environmental awareness?
7. What type of materials are used to make the robot more heat resistant?
   1. What type of coating is used?
   2. How expensive/difficult is this to implement?
   3. Are there other options you would recommend (more suitable for our project)? (What about galvinized steel?)
8. What are the limits in terms of temperatures, potential damage etc?
   1. How is the robot able to reach these limits?
9. Have any of the robots been applied in real life situations so far or just lab testing? How have the results been?
10. In the promotional video it is mentioned you can throw the robot, what are the benefits of this?
11. Are there any other things we did not discuss that you think could be interesting for our project?

Closing interview:

Thank you for answering all our questions, this will really help us to design a robot that will be as useful as it can be. If you have any further questions, do not hesitate to contact me. This is the end of the interview. (and I will end the recording now)

Summary/findings:

It turned out the expert was not yet looking into heat shields or any way of making the robot more heat resistant. This means there is an opening for us to make recommendations about this during this course. The main constraint our solution has to take into account is weight. The expert was very clear that people did not want to use the device if was not very light weight. The second important constraint is of course cost, if the solutions is very expensive the fire department will probably not find it a worthwhile investment or not even be able to afford it at all. There was not a clear limit given in additional costs but there was a limit on weight, namely 300 grams. The robot now is 1.2 kg and it should not be more than 1.5 kg. Therefore we will try to find a solution that is under 300 grams and as cheap as possible. The maximum temperature that the robot can handle is about 85 degrees. Another notable thing that the expert mentioned is that because their robot is so low to the ground, they actually do not encounter big difficulties with visibility in a room with a lot of smoke. Additionally he advised that we either focus on improving situational awereness with a teleoperated robot or we focus on an autonomous robot that looks for people. Since we now mostly narrowed our focus to the heat shield, this is not extremely important for us anymore but it is still worth mentioning.

The Model

The experiment that was conducted gave insides into the heat transfer through the chosen materials. It has shown how the temperature changes and to what extend the temperature increases over time. A theoretical model is also created that will simulate the experiment. The purpose of this model is to see how accurate the temperature can be estimated. This estimation can help to do further predictions with the model, i.e. increasing/decreasing certain parameters, that can help with making design choices in the future.

Creation of the model

The model makes use of a ordinary differential equation which takes different thermal resistances and heat capacities into account. The thermal resistances used are those of convection through the layer of air, conduction through the galvanized steel, and conduction through the insulation. All of these thermal resistances have been calculated separately with the appropriate values for each of the materials, these are then added up to get the total thermal resistance that is to be used in the ODE.

The heat stored in these layers and their temperatures due to this heat have been calculated as well, by multiplying the mass of each layer with the specific heat capacity of their respective materials.

The area that is directly heated by the heatgun is very small, but eventually a bigger area of steel and insulation will heat up as well. This has been implemented in the model by starting of with the small area directly heated as the initial value and then slowly increasing the area used in the ODE over time. This makes the model more accurate than just assuming the heat is distributed evenly over the entire area of the galvanized steel plate and insulation that were used in the experiment. However, this assumption is only added to make the model more accurate when comparing with the experiment done, but not for a realistic case. If the material really would be placed in a fire, the heat would come from all directions, thus making the assumption that the entire material heats up simultaneously would be the accurate choice.

As the heater was turned off after 300 seconds during the experiment, the model also plots the temperature over 300 seconds to make a clear comparison. As there were three temperature sensors place during the experiment, the temperature values for each of these locations were plotted as well. These locations are: before the steel plate, after one layer of insulation, and after a double layer of insulation.

If the model can be deemed reliable enough for the situation replicated during the experiment, other useful predictions can be made as well. By changing the values set at the beginning of the model, the temperatures can be plotted for other circumstances too. For example how the temperature increases when the heat supply is doubled or the layer of steel is made thinner.

Validation of the model

The model as well as the results of the experiment have been gathered, which means that it can now be determined how well the model functions. In the figure below, the model results can be seen which can be compared to the results of the experiment. During the experiment, a continuous heat source was pointed at the materials for 300 seconds. At that time, it can be seen that the temperature behind the steel plate of 0.75 mm and a ceramic fiber layer of one cm is 62.60 degrees in the model and 65.94 degrees in the experiment. If the ceramic fiber layer is doubled, the temperatures become 39.74 and 31.50 degrees, respectively. This is a mean error percentage of about 5% for one layer and 20% for two layers.

Temperature evolution over time of the model.

The reasons for the bigger deviation between one and two layers of ceramic fiber are mainly caused by the assumptions made in the model. The connection between the first layer of ceramic fiber to the steel and the second layer of ceramic fiber to the first layer is not identical. There could be a layer of air in between the layers during the experiment that has not been accounted for in the model. Also, it is assumed that the heat from the first layer will travel in its entirety to the second layer, which is not fully accurate since there is a small amount of heat loss at that moment. This means that the model is most accurate at small deviations of the parameters and will get less accurate when the thicknesses of the materials are increased on a centimeter scale.

Further predictions

One of the main limitations to a fire resistant robot is its weight. This has been established during the interviews. Because one of the materials is galvanized steel, it would be best to keep the layer of steel as thin as possible. The model was used to analyze what would happen with the heat transfer with different thicknesses of steel. As a result shown in the figure below, it can be seen that the final temperature is only slightly lower with a 1 cm steel layer opposed to a layer of 0.75 mm, which can already be seen in the figure above. This means that the steel layer is not much involved in the heat transfer which means that this layer can be as thin as possible without bringing difficulties to the heat resistance. To find out what the best thickness would be will thus only rely on the structural support it will need to give to the robot.

Temperature evolution with increased steel layer, from 0.75 mm to 1 cm.

To finalize, the model is used to simulate the experiment. It was found that the model comes quite close to the real data which means that it can be used to make accurate predictions. For instance, the galvanized steel plate is a good heat conductor which means that the heat can transfer easily through the steel. This means that the steel plate can be as thin as possible to cut some weight. The ceramic fiber insulation is thus the most important component in the heat transfer and should be used extensively in the design of the robot.

Conclusion

We started off this project with the goal of designing a robot that could aid firefighters in detecting trapped survivors in a fire. During the course of the project, we worked with both firefighters and robotics entrepeneurs, realising that the technologies we were researching and designing could be helpful in an even wider scope. After eight weeks we have delivered substantial results in regards to the design of a robot that can enter a fire and detect survivors, as well as successful results in multiple experiments regarding the thermal regulation and cooling of such a robot that accurately correspond with predicted mathematical models.

While the construction of our robot and implementation of our findings were outside of the scope of this eight week project, we are satisfied with our results and would be thrilled to see them be used in future studies, designs or experiments. For the users that we focused on, we may not have a final product to deliver to the firefighters, but we have delivered interesting findings with regards to the fiberglass insulation and gas cooling system that our second user, Danny Hameeteman, co-founder of Sita Robotics, can take into consideration with his plans for future improvements for his situational awareness robots.

We encourage anyone interested in our work and who wants to pick up where we left to do so.

Recommendations

To get better results for the experiment, longer measurements could be taken to get a clearer picture on how the temperature evolves in the long run. Furthermore, the same amount of insulation and other materials present should be on the other side of the sensor as well, to get more realistic heat loss results.

These heatlosses are currently considered to be minor in reality, as in a fire the heat would come from each side. For this reason, the heatlosses have not been implemented in the Matlab model, but this heat flow from all sides should be added in the future to get a more accurate model.

Literature:

References

[1] - V. Etacheri, R. Marom, R. Elazari, G. Salitra, and D. Aurbach, “Challenges in the development of advanced Li-ion batteries: a review,” Energy Environ Sci, vol. 4, no. 9, pp. 3243–3262, Aug. 2011, doi: 10.1039/C1EE01598B.

[2] - Z. Y. Jiang, H. B. Li, Z. G. Qu, and J. F. Zhang, “Recent progress in lithium-ion battery thermal management for a wide range of temperature and abuse conditions,” Int J Hydrogen Energy, vol. 47, no. 15, pp. 9428–9459, Feb. 2022, doi: 10.1016/J.IJHYDENE.2022.01.008.

[3] - Udayraj, Prabal Talukdar, Apurba Das, Ramasamy Alagirusamy, “Heat and mass transfer through thermal protective clothing – A review”, International Journal of Thermal Sciences, Volume 106, 2016, Pages 32-56, ISSN 1290-0729, doi: 10.1016/j.ijthermalsci.2016.03.006.

[4] Human Presence Detection using Ultra Wide Band Signal for Fire Extinguishing Robot | IEEE Conference Publication | IEEE Xplore

[5] Simulation and signal processing of UWB radar for human detection in complex environment | IEEE Conference Publication | IEEE Xplore

[6] Through-Wall Detection of Human Being's Movement by UWB Radar | IEEE Journals & Magazine | IEEE Xplore

[7] Real-Time Human Detection Behind Obstacles Based on a low-cost UWB Radar Sensor | IEEE Conference Publication | IEEE Xplore

[8] A method to accelerate the rescue of fire-stricken victims - ScienceDirect

[9] Implementation of survivor detection strategies using drones - ScienceDirect

[10] Sensors | Free Full-Text | Experimental Evaluation of Sensor Fusion of Low-Cost UWB and IMU for Localization under Indoor Dynamic Testing Conditions (mdpi.com)

[11] X. Hou et al., "A Novel Mobile Robot Navigation Method Based on Hand-Drawn Paths," in IEEE Sensors Journal, vol. 20, no. 19, pp. 11660-11673, 1 Oct.1, 2020, doi: 10.1109/JSEN.2020.2997055.

[12] Baker, Greg, et al. "Towards an immersive user interface for waypoint navigation of a mobile robot." arXiv preprint arXiv:2003.12772 (2020).

[13] Rohde, Mitchell M., et al. "PointCom: semi-autonomous UGV control with intuitive interface." Unmanned Systems Technology X. Vol. 6962. SPIE, 2008.

[14] J. Kim and S. Sukkarieh, "Autonomous airborne navigation in unknown terrain environments," in IEEE Transactions on Aerospace and Electronic Systems, vol. 40, no. 3, pp. 1031-1045, July 2004, doi: 10.1109/TAES.2004.1337472.

[15] S. Koenig and M. Likhachev, "Fast replanning for navigation in unknown terrain," in IEEE Transactions on Robotics, vol. 21, no. 3, pp. 354-363, June 2005, doi: 10.1109/TRO.2004.838026.

[16] Wang, Y., Mulvaney, D., Sillitoe, I. et al. Robot Navigation by Waypoints. J Intell Robot Syst 52, 175–207 (2008). https://doi.org/10.1007/s10846-008-9209-6

[17] F. Niroui, B. Sprenger and G. Nejat, "Robot exploration in unknown cluttered environments when dealing with uncertainty," 2017 IEEE International Symposium on Robotics and Intelligent Sensors (IRIS), Ottawa, ON, Canada, 2017, pp. 224-229, doi: 10.1109/IRIS.2017.8250126.

[18] P. Saeedi, S. -A. Sorensen and S. Hailes, "Performance-aware exploration algorithm for search and rescue robots," 2009 IEEE International Workshop on Safety, Security & Rescue Robotics (SSRR 2009), Denver, CO, USA, 2009, pp. 1-6, doi: 10.1109/SSRR.2009.5424168.

[19] Ball, David, et al. "Vision‐based obstacle detection and navigation for an agricultural robot." Journal of field robotics 33.8 (2016): 1107-1130.

[20] Huh, Kunsoo, et al. "A stereo vision-based obstacle detection system in vehicles." Optics and Lasers in engineering 46.2 (2008): 168-178.

[21] A Multi-robot Topic Communication Method Based on TCP and UDP | IEEE Conference Publication | IEEE Xplore (tue.nl)

[22] Autonomous Firefighting Robot With Optional Bluetooth Control | IEEE Conference Publication | IEEE Xplore (tue.nl)

[23] Architectural design and performance evaluation of a ZigBee technology based adaptive sprinkler irrigation robot - ScienceDirect

[24] Measurement and Analysis of Network Data Based on MQTT Protocol | IEEE Conference Publication | IEEE Xplore (tue.nl)

[25] Mobile Robot Teleoperation via Android Mobile Device with UDP Communication | IEEE Conference Publication | IEEE Xplore (tue.nl)

[26] Unmanned Aerial Vehicles Networking Protocols

[27] Minimizing End‑to‑End Delay on Real‑Time Applications

[28] Evaluation of educational applications in terms of communication delay between tablets with Bluetooth or Wi-Fi Direct

State of the art review

Novel exterior cover design for radiant heat resistance of firefighting robots in large-scale petrochemical complex fires | ROBOMECH Journal | Full Text (springeropen.com)

Summary/Relevance to topic:

A big issue for firefighting robots is the heat radiated by the fire. There are existing ways for increasing the resistance to heat, used for example with water cannon robots. However, the current method requires a lot of water, increasing the weight of the robot by a lot. This obviously reduces the mobility of the robot a lot. This paper aims to find another way to make these robots heat resistant, using much less water, by implementing an exterior cover. This paper goes into the design specifics of this cover. Even though the aim of our robot is not to assist in the firefighting itself, but rather locating potential survivors, this robot obviously still needs to be heat resistant. Therefore, this design proposal might also be valuable for our robot design.

Robot-aided human evacuation optimal path planning for fire drill in buildings - ScienceDirect

Summary/Relevance to topic:

This paper researches algorithms to assist humans with evacuating, by calculating the fastest routes out. In our robot design, we would like to implement not only the locating of potential survivors, but also a fastest route for the firefighters to reach this person. A similar algorithm as that discussed in the paper can be implemented in our design as well.

Fire Fighter Robot with Deep Learning and Machine Vision by Amit Dhiman, Neel Shah, Pranali Adhikari, Sayli Kumbhar, Inderjit Singh Dhanjal, Ninad Mehendale :: SSRN

Summary/Relevance to topic:

Here, rather than the use of for example heat sensors, AI deep learning and machine vision is used top detect fires. This already works with a very high accuracy. Extending this, the machine vision could potentially also be used to differentiate between fire and potential survivors. This might be more accurate than using merely heat and motion sensors to locate people, thus improving how well our design would work.

A Robot Swarm Assisting a Human Fire-Fighter: Advanced Robotics: Vol 25, No 1-2 (tandfonline.com)

Summary/Relevance to topic:

This paper goes into the GUARDIANS robot swarm, which is designed to assist firefighters in searching big warehouses for survivors to save. This is very similar to what our design aims to do, although we would like to apply this in housefires/other smaller fires as well, not just large warehouses. However. a lot of the technologies discussed in this paper, such as the wireless communication system, are very relevant to our design.

The role of robots in firefighting | Emerald Insight

Summary/Relevance to topic:

This paper goes into the state of the art, as robotics in firefighting is a fairly new technology. So far, the most prevalent technologies include: all-terrain vehicles to assist the actual fire-fighters with getting to and operating at dangerous locations, also giving the firefighters a better overview of the situation by using sensors, and drones that are either equipped with fire extinguishing materials, can hold up a hose (both used for high up buildings), or to again create more awareness of the situation for the firefighters. This state of the art review can help assessing what elements of our design already exist, what needs to be improved, and what is still missing.

Automatic Fire Detection System Using Adaptive Fusion Algorithm for Fire Fighting Robot

Summary/Relevance to topic:

In this paper the authors describe a firefighting robot they have created, listing the materials and systems used to make the robot fire-resistant and robust and to allow it to detect fire and navigate the area. From this paper we can see what worked well to help us decide how to build our robot.

Deep learning assisted portable IR active imaging sensor spots and identifies live humans through fire

Summary/Relevance to topic:

In order to identify humans in a burning building we need software and sensors that can recognize human bodies despite the very hot temperatures that may stop traditional infrared detection from working well. This paper provides an alternative system using deep learning.

Internet of Robotic Things Based Autonomous Fire Fighting Mobile Robot

Summary/Relevance to topic:

Prevention is also important in firefighting; robot assistance can be in place before a fire starts to alert firefighters and monitor the situation allowing for early intervention. This paper outlines such a robot, which provides inspiration if we decide our robot should be more preventative.

Design of a cooling system for an all-terrain electric vehicle for firefighting

Summary/Relevance to topic:

A firefighting robot will contain electronic components in order to control the vehicle, and run navigational and fire detection software. The robot must be able to keep these electronics cool under extremely high temperatures to remain function. This article proposes a cooling system to accomplish this.

Present status and problems of fire fighting robots

Summary/Relevance to topic:

This paper summarizes the current state of firefighting and rescue robots, mentioning variables to consider when designing such a robot such as size and weight, and cost and performance.

Design and Fabrication of an Autonomous Fire Fighting Robot with Multi-sensor Fire Detection Using PID Controller.

Summary/Relevance to topic:

The text highlights the development of fire detection and extinguishing robots, their components, and testing procedures. The focus is on locally available materials and Arduino-based control systems. Sensitivity tests for flame sensors and LM35 (Temperature) sensors are conducted at different times and distances from fire sources. The robot is able to detect and extinguish small fires and shows promising results for the future of firefighting. However, the robot functions better in darker places due to sunlight disrupting the output values.

Portable Fire Evacuation Guide Robot System.

Summary/Relevance:

The text describes the development of a portable fire evacuation guide robot system. This system is designed to gather environmental data and locate people. It features a compact, cylindrical design with various sensors, a camera, and a microphone for communication. The robot is lightweight, remotely controlled and designed to withstand high temperatures and impacts. Firefighters are able to carry and throw this robot in various places to assist them during a fire.

Human–Robot Interaction in Rescue Robotics.

Summary/Relevance:

This paper analyzes human-robot interaction that is involved in rescue robotics. It emphasizes that rescue robots complement, rather than replace human efforts, highlighting the importance of teamwork in rescue operations. The current state involves operations with a 2:1 human to robot ratio. The paper identifies key human-robot interaction research questions and emphasizes the need for human-centered advances to ensure effective rescue operations.

The Application of Multi-agent Robotic Systems for Earthquake Rescue.

Summary/Relevance:

Rescue robots are used in a variety of situations, which include earthquakes. In relation to fire rescue robots, a lot can be learned from earthquakes since the environment is very similar. This paper covers various aspects of a rescue robot, such as the structure of multi-agent control systems, methods for searching victims, path planning and search algorithms. Many of these aspects can come in handy for the future of rescue robotics.

Thermal and structural analyses of firefighting robot.

Summary/Relevance:

A robot that has to endure harsh environments as well as rapid environmental changes requires materials that are well suited for these situations. The paper goes over a structural and thermal analysis that evaluates the performance of a robot that can be used in, for instance, a big house fire. The robot was designed with materials like galvanized steel as the main plate, cubic boron nitride coating for non-flammability and silica aerogel for thermal insulation. Results show that that after 1800 seconds, the inside of the robot only had a temperature change of 2 degrees. It can be concluded that these materials are very well suited for its application and can make sure that all systems on board of the robot can operate under harsh conditions.

Flying dragon robot used to help extinguish fires | frontiers

Summary/Relevance:

This paper delves into research about making a remotely controllabe firefighting robot. The idea is of course that less human fire fighters have to go into the dangerous fire and to instead send robots. How to let the robot move, what the optimal nozzle size is for the best water thrust, new waterproofing techniques, and a larger movable range of the nozzle unit are discussed. These things are relevant to our robot especially if we are able to encoorporate a water tank to help locally extinguish fire around a person, which would of course improve the functionality of the robot.

Ethical concerns about search and rescue (SAR) robots

Summary/Relevance:

This paper considers some ethical concerns surrounding SAR robots. Issues like the level of robot autonomy, laws surrounding robot design and behavior, but issues with the human response to the robots and who is responsible for the actions of the robot.

Improving the SAR robots feedback and interface

Summary/Relevance:

This paper summaries four studies done on what type of feedback and interface a SAR robot should give/have to be the most trusted and best understood. This is very important because having a robot that no one understands or trusts is virtually useless and will only add confusion and fear to an already terrifying situation. The main finding is that multi-sensory interfaces (having e.g., visual, olfactory, and audio feedback) can be very beneficial and have minor effects on the cognitive load. Or in other words you should exploit the redundancy gain.

Robot competition (RoboCup) to locate victims

Summary/Relevance:

This paper shows the results of a robot building competition that had the main goal of building a robot that locates victims and determines their health status. It discusses how the different teams tackled this challenge and the outcomes of their strategies. It gives an overview of a lot of different and unique ways to locate victims in a maze situation (which is similar to corridors in for instance a hospital) and how effective it was. We could use this to help inform and get inspiration about our decisions about building a robot that locates people in a building.

Process of human behavior in fires

Summary/Relevance:

This paper aims to give an overview of the behavior people display during a fire. It does this by breaking the process down into phases and describes what factors are relevant for an individuals response. For our robots design it is important to understand how people respond in a fire to antipate the interaction the human robot interaction.

Human Presence Detection using Ultra Wide Band Signal for Fire Extinguishing Robot | IEEE

Summary/Relevance to topic:

This paper describes a remote controlled, 4-wheeled fire extinguishing robot, that is capable of detecting various environmental factors such as temperature and smoke, and it can also detect human presence using something known as “ultra-wide band radar”. This appears to be quite similar to the system we are considering.

Humanoid robots rescuing humans and extinguishing fires for Cooperative Fire Security System using HARMS | IEEE

Summary/Relevance to topic:

This is a paper written as part of a cooperation between multiple universities, and provides some information about a humanoid fire-rescue robot that was designed. The scope of the project seems comparable to ours (though still larger), and thus it may be relevant despite being light on real-world applicability.

Ethical concerns in rescue robotics: a scoping review | Springer

Summary/Relevance to topic:

This is a somewhat fresh (2021) literature review about the ethics surrounding rescue robotics. While this source may not be relevant to any design activities that we would like to perform, it could serve as a great starting point for analysing any ethical aspects.

Exploring the Ethical Landscape of Robot-Assisted Search and Rescue | Springer

Summary/Relevance to topic:

This paper identifies ethical concerns and value conflicts that arises from the use of SAR robots. The paper mainly focuses on Values Assessment Workshops whose participants were professional (Italian) firefighters. The paper thus details concerns and dilemmas regarding SAR robots, it is meant as a ‘conversation starter’ and not as an answer.

Robot–human rescue teams: a user requirements analysis | tandfonline.com

Summary/Relevance to topic:

This paper is about the needs of professionals from the field of SAR. The paper includes the end-user requirements of these professionals, as well as some guidelines for rescue systems. This could help guide our endeavours if we want to design a human-robot interface.

An Indoor Autonomous Inspection and Firefighting Robot Based on SLAM and Flame Image Recognition | MDPI

Summary/Relevance to topic:

This article focuses on indoors firefighting robots. It is valuable for the project, as it discusses in detail the complexity of indoor fire environment and proposes a way for a robot to deal with high temperatures, smoke, and the complex geometry of a building. Moreover, it discusses SLAM (simultaneous localization and mapping) which should be used by our robot as well.

A High-Temperature Resistant Robot for Fixed-Point Firefighting | Springer

Summary/Relevance to topic:

This article is relevant as it has a design of a thermal protection structure which covers the robots and assures the normal operation of internal components. This design might be useful for our project as a ready solution or an inspiration source.

Research on Heat Transfer through a Double-Walled Heat Shield of a Firefighting Robot | MDPI

Summary/Relevance to topic:

This article provides another insight into heat resistance for robots and how it behaves. This article is a good source for preparing a test plan for our robot’s thermal-protective shield/cover. Not only a heat shield is designed, but it is also tested, and these tests are what makes this article so valuable within this project.

RoBoa: Construction and Evaluation of a Steerable Vine Robot for Search and Rescue Applications | IEEE

Summary/Relevance to topic:

The article gives a good insight into Vine Robots being used in search and rescue operations. The design proposed in the article can be used within our project, if we choose to base our robot on Vine Robot model. However, a lot of work still needs to be done to make the design fit for extreme thermal conditions (if it is possible).

An Arduino Uno Controlled Fire Fighting Robot for Fires in Enclosed Spaces | IEEE

Summary/Relevance to topic:

The article contains a basic design of a low-budget firefighting robot. If we decide to make a prototype of our robot, this article will be useful, as the Arduino system is indeed affordable and firefighting-robot mentioned in the article shares a lot of properties with a SOR robot for fires, that we have in mind.

Van Wynsberghe, A. A method for integrating ethics into the design of robots. Ind. Robot. 2013, 40, 433–440.

A paper about how to integrate ethics into robot design. "The approach for including ethics in the design process of care robots used in this paper is called the Care‐Centered Value Sensitive Design (CCVSD) approach. [...] In this paper, this approach's utility and prospective methodology are illustrated by proposing a novel care robot, the “wee‐bot”, for the collection and testing of urine samples in a hospital context."

Appendix

Appendix A (Time spent table)

Week 1
Person	Time spent	Task
Tessa G.	10 hours	Meeting deciding on subject and dividing roles (4 hrs) literature review and adding sources (4 hrs) , wrote on users and what do the users require (2 hrs)
Tessa C.	12 hours	Meeting deciding on subject and dividing roles (4 hrs), literature review/ finding 5 suitable source (4 hrs), dividing subjects of the design (1 hr), milestones for each subject per week (3 hrs)
Storm	9 hours	Meeting deciding on subject and dividing roles (4 hrs), literature review and adding sources (3.5 hrs), writing problem statement and objectives with Roman(1.5 hrs)
Abel	10 hours	Meeting deciding on subject and dividing roles (4 hrs) literature review and adding sources (3 hrs) , wrote on users and what do the users require (3 hrs)
Roman	9 hours	Group meeting for choosing the subject of our project (4 hrs), literature study (3 hrs), writing problem statement and objectives with Storm(1.5 hr), minor edits of the wiki page(0.5 hr)
Elektra	9 hours	Group meeting on subject and dividing roles (4 hrs) literature research (3 hrs), writing about milestones and deliverables (2 hrs)

For weeks 2-7 refer here: https://docs.google.com/spreadsheets/d/1G5tPp-6NsQBCDB8bOLenNYROfXyhWo69ukwcyu0_uLk/edit?usp=sharing

Appendix B (Transcript interview fire department)

Interviewer: The product we are designing is intended to help firefighters and firefighters and it is going to do that by finding people find people in a building that is on fire and report that so they can search more specifically. Exactly how this is going to work we are working on now and for that your answers are very important. Do you have any further questions?

Fireman: No, it's just about searching people so?

Interviewer: Yes, he will go into the building at least that is how it looks now, and find people there and tell them that. The first question is what is currently the protocol for searching people in a burning building?

Fireman: When we arrive with our fire engine, we always do an outside reconnaissance first. So around the building. There are six of us, so one couple goes one way and the other goes the other way. We always keep the doors closed as much as possible, we used to throw up all the doors right away and throw in all the windows but now we don't do that any more.

Interviewer: Is that for the oxygen

Fireman: Yes, and then we look for the shortest route of attack. Then, first of all, we get to the fire faster, because it's important that we put it out quickly, provided we come across casualties, then we will rescue them.

Interviewer: So if I understand correctly, the focus is mainly finding the fire and getting to it?

Fireman: Yes.

Interviewer: So there are not necessarily people who are going to look for victims?

Fireman: Not initially, we always try as much as possible to extinguish first and then rescue. It didn't used to be, then it was always search for people first, but when you put out fire the worst danger is gone. Once the fire is out, the windows and doors are immediately opened so that the smoke can escape and then we also have more visibility, which makes it easier to search for victims. And we also always make a decision first about whether to go inside or stay outside. You can imagine that if it's a spreading fire that the fire is so big that we can't actually go inside because then we put ourselves in danger. So we either have a defensive or an offensive outside deployment. Either we are going to extinguish to make it smaller, or we are going to try to preserve the buildings next to it and that is defensive. And we also have those for an indoor deployment, so also an offensive or a defensive indoor deployment.

Interviewer: And inside is then either extinguishing as much as possible or saving as much of the rest of the building as possible?

Fireman: Yes, so those are the four options we have. Often in the case of a spreading fire or if we know that the building too far gone and there are no victims inside then we will go for an offensive outside deployment and if there are buildings right up against it we will also do something about the defensive side have to do. So that is actually the tactic we have in firefighting. That's a fairly new method, only one and a half/two years or so I think. So that.

Interviewer: Okay that's clear. And which part of this process takes the most effort, or takes the longest, takes the most energy?

Fireman: What takes the longest if it's a very big building or a commercial building you can imagine a lot of time goes into that, if it's a small fire we can go in and put that out. If it's a complicated fire so with toxic substances it often takes the longest.

Interviewer: Okay, and what temperatures do you usually have to work with?

Fireman: That varies, sometimes it's not hot when you extinguish it steam comes off and then it gets hotter, then sometimes it can get up to the 5/600 degrees.

Interviewer: Okay, that's hotter than I would like.

Fireman: Those are really just short moments, so if we go out and we use a lot of water and there's a lot of steam coming off then you do have really hot moment for a second of 10 -20 and then that does cool down again.

Interviewer: Okay, do you guys ever have floor plans before you go into buildings?

Fireman: Yes sometimes we have evacuation plans that we then use or in the Mooi, which is a system of the fire brigade that sometimes has floor plans in it.

Interviewer: Okay, and are there floor plans of houses in there? Or only of large properties?

Fireman: Yes, usually only of large commercial buildings, but sometimes we don't have them and then we have to form our own picture of what a house is like.

Interviewer: Okay, and an evacuation plan is that one of those pictures you sometimes see hanging in buildings with the escape routes?

Fireman: Yes, that is one of those maps, and there is often a FAFS response team walking outside when they have an evacuation and we can make good use of them.

Interviewer: Okay, and do you think a robot that searches for people in a burning building could be useful for you or for your organisation?

Fireman: Yes I think so, because we are already working with thermal imaging camera, where we can see places in a building, so where we can use it to find fire or victims or a fluorescent tank that is overheated. So we are already using that.

Interviewer: Okay, and is that someone walking through the building with that camera?

Fireman: We take that with us as standard, that's just a kind of camera we take with us as standard.

Interviewer: Okay, and that's then something you just use with in hand?

Fireman: Yes.

Interviewer: Okay that already sounds useful at least. And do you have an idea of how this robot could be most useful?

Fireman: I think if it could send those images to someone outside, a commanding officer or an officer on duty, who would then have a tablet in his hands where he could look at it, that would be the easiest.

Interviewer: Do you mean camera images?

Fireman: Yes, because we already have a drone team in the fire service who do that as well.

Interviewer: Drones that film?

Fireman: Yes.

Interviewer: Also those that film inside?

Fireman: No, I don't think so, at gemlot (maybe just a different spelling) they have a drone team like that, you'd have to see what that's like, but I don't think they film from inside, only from outside.

Interviewer: Okay then we will definitely take a look at that. Would you rather see a robot driving or flying around from within itself, or someone controlling it from outside?

Fireman: I think it's easy if the person outside can control the robot. I think it would be smart to do that not with joysticks but with a finger movement over a tablet.

Interviewer: Okay and then is there someone free to do that? Or does that require an extra person?

Fireman: Yes maybe an extra person, or a pump operator.

Interview: Okay so that will be okay then probably?

Fireman: Well the commanding officer outside is usually pretty busy too.

Interviewer: Okay but then still it's nicer for someone to drive it themselves? It's worth it then, isn't it?

Fireman: I think the ones inside are very busy, so I don't think that's smart.

Interviewer: Okay, so the one driving it stays outside?

Fireman: Yes, two people stay outside either the commanding officer or the pump operator.

Interviewer: Okay, so if it were autonomous it would fly around all by itself and so nobody would have to drive it, it would just send images or locations. But so that's less convenient anyway?

Fireman: Yes, but then what do you see? Then you would have to send him voice commands.

Interviewer: Okay so it is also important that he gives a certain picture of the situation?

Fireman: Yes I think so yes, that you do have to let a robot drive somewhere you think is important, so not that it just drives aimlessly through a room.

Interviewer: Okay, if there is a robot that finds people in a building, how would you like to receive that location?

Fireman: On a map. With a little doll in it (jokingly).

Interviewer: Okay, and you mentioned that one of the first things you guys do is find the core of the fire, is that something that goes fairly "easy" or would it be nice if the robot helped with that.

Fireman: Yes usually we find the core pretty quickly, because we also have that thermal imaging camera. So if we are in another room and we shine that space camera around we can see the wall which is hot even if the fire is on the other side. So we can estimate pretty quickly where the core is. Also the flow of smoke always helps us find where the fire is because the smoke always flows away from the fire because of the oxygen migrating towards it. If you are trained a bit well, you can know pretty quickly which corner of the building the fire is in.

Interviewer: Okay, so it's not necessarily worth having the robot help there?

Fireman: No.

Interviewer: Okay, we've talked a bit about this too, but do you know of any other products that do a bit of the same thing?

Fireman: Well a few years ago there was a group of students who wanted a face mask, so one of those masks that we have on, they wanted to make a map in the visor there. I think MSA, which is a supplier of breathing apparatus, they were working on that to make that, but I don't know how that turned out. Well drones we have, those heat cameras every fire engine has at least 1 usually two actually. I think that's about it in terms of materials.

Interviewer: Okay, and is there any particular restriction of how big the robot can be to transport it?

Fireman: Yes it has to be small. As flat and narrow as it can be. It has to be stable because it probably has to drive over uneven ground. Bumps.

Interviewer: And is it more convenient if it's flat and long or more square?

Fireman: I would keep it small, flat and low. The previous group talked about throwing it through the window, well then the window is broken. Which we didn't want. So somewhere we have to open a door quickly, robot in and quickly close the door again. If we have kept a fire small for a long time by smothering the fire, making it smaller by admitting less oxygen, there are dangers for us there too because then we can get a backdraft. These are unburnt smoke particles that can spontaneously catch fire when the temperature gets high.

Interviewer: And how do you prevent that?

Fireman: By keeping the doors closed and by cooling smoke gases with water. We can counteract those smoke gases, that smoke hanging from the ceiling, we can cool it with water and then we won't get a backdraft if all goes well.

Interviewer: Okay that's fine. That was basically all the questions we had prepared. Again thank you very much and I will now turn off the recording.

Appendix C (Transcript interview technical expert)

Interviewer 1: Which robot that is developed by your company is more appropriate for fires? The jens or spear?

Interviewee: So indeed a small background maybe, I am one of the founders of SITA robotics. It is a company we try to develop robotics for accessible solutions for police fire departments those kind of applications. So what we have done so far is two projects, one of which is for the ministry of defensive where we developed the spear robot. That is a two wheeled system that you can easily throw into a building and do some research there, it is teleoperated, it has a controller with that so that you can easily have access to what’s happening inside the building. Aside from that we developed jens, which is a four wheeled system, it has a similar concept the only difference is that you can more easily overcome obstacles. Take some extra weight, workload, with you for extra sensors. That one is more specified for inspection purposes I would say, so there is no time sensitive operation. So have the best operation instead of optimize for the shortest time, so in that sense I would say the spear robot, so the two wheeled system, is the more appropriate one. Aside from that we are now also developing a device which is not a robot but a remote “sensor ball”, which is basically a system, you can roll it into a building which gives you information about gas, a camera, sound stuff like that. Actually today there is an article in the Telegraaf about our robot. So that’s another project that is currently running.

Interviewer 1: What are the mobility limits of this robot? And how much load can it carry?

Interviewee: So in this case I think we should focus on the two wheeled system, because that is really used for time sensitive operations. Its able to drive over relatively flat grounds, small obstacles are doable, up to like 6 cm or so, that is about the limit. What typically happens is they know there is something happening on the other side of the wall, but they cannot go there easily, because they don’t want to air going in there as well, so they would like to have some sort of extra camera sight on that location. Especially robustness and trackability is important in that sense.

Interviewer 1: Okay, and how did you implement user requirements when you developed the first and maybe the second robot?

Interviewee: So when talking to clients, you typically hear they want everything but they have a very limited budget, so in that sense we are trying to peel down what is actually of most importance. One of the first spec sheets we got, was we want a 4K camera system on the robot, then we come with a 720p which is far far lower that the 4K that they expected and we get compliments on the quality of the screen. We really try to understand if its actually needed what they are asking for. So why do you want to have 4K, because they want to see what is 3 meters away and specify towards that instead of implement whatever is said by the user. With money you can probably do everything but then you have really inaccessible expensive system that no one is going to use anyway. So that is a really big trade off.

Interviewer 1: You also mentioned that you not only develop robots for military but also for other emergency services, how do used requirements differ there?

Interviewee: So they are all different, at least they say, so the result that you obtained can be different, because in a military situation typically your life is at stake, but typically in a swat team it is more that they already know what will happen inside, they just want a confirmation. So it seems different but the solution that we proposed fits their situations quite nicely. Sorry what was the question again?

Interviewer 1:  How do the user requirements differ? Or you’re trying to come up with some ultimate solution that fits many?

Interviewee: No not so much, but we try to minimize it so that it fits the basic needs for each of them, so for instance a gas sensor that might be really relevant for a fire department but for a swat team not so much. So its not in the basic product.

Interviewer 1: What are the main obstacles for this type of robots?

Interviewee: Like physical obstacles?

Interviewer 1: No like obstacles that would prevent someone from using it.

Interviewee: Yeah so in the end it is mainly the intuitiveness, so if they are not used to using those systems which you typically see because they are not trained for those systems, or maybe one guy out of twenty that is trained. So for us its really the intuitiveness, we really try to make it so that you pull the pin and it starts and you can already directly have sight on your screen, that makes it easier to implement. Though it is still a peoples game in that sense, its definitely not only the technology.

Interviewer 1: Apart from specific sensors, like co2, which sensors are used? It seems that the robot mostly gets video footage, are there any other ways to get better environmental awareness?

Interviewee: Yeah so, two other elements that we are experimenting with is thermal information and acoustic information. You can see a lot from acoustic noise, sound, and also from temperature difference. So each fireman is trained with a thermal gun, so that they can check okay is the handle too hot to grab or is there something behind it? So I think those two really benefit your “sixth sense”.

Interviewer 2: So you use the acoustic gun for echo’s?

Interviewee: Yes or if there is a gas leak. You can visualize the difference in sound.

Interviewer 1: I also wanted to ask, if the visibility is really low, the room is covered in smoke for example, then you get barely anything for the camera. Would then ultra sound be useful?

Interviewee: So I know there have been researches on radar sensors that work quite well in smoke, although radar sensors in general are still quite clumsy so not suitable for our situation. The way we are handling the smoke part is that we are really low to the ground. Smoke elevates so we have not much interference from this. Usually they will not enter a room that is so filled with smoke but if they do they are trained to go low to the ground to see the exit for instance.

Interviewer 1: Ah that’s really interesting, so are there any materials used for the robot coating that are heat resistance?

Interviewee: No, that is one of the user requirements that we did not take into account. Because we believe that the accessibility part is more important than that is not melting or burning when it is to close to the fire. Also you should think of, how close do you actually need to be to the fire to actually get your information? Of course I would be really interested to hear what you have come up with but that is maybe for later.

Interviewer 1: So it is not high heat resistant. Does it have any cooling systems?

Interviewee: No also not. We have a small fan, that is mainly for trying to create a circulation through our sensors. But they do not really heat up that much, we can do it by just dissipating over the metal.

Interviewer 1: That’s good, what would be the limits of the robot regarding temperature?

Interviewee: I do not know what we said exactly but I would say around 85 degrees it would become a problem. And the lowest I would say around -10 degrees. But it mainly depends on your battery.

Interviewer 1: Have any of these robots you told us about been applied in real life situations? Or just lab testing? And would your clients that have classified information share the performance of the robot with you?

Interviewee: So yes we have a couple of prototypes running. Mainly one that are being used in the police case. Not the fire department case yet, because we are still on the part of the development. And yes we do get information back from them, and how the experience is. Especially the things that could be applied to our next product. So what we did with the jens system, we brought it to a lot of police stations for a month and then obtain the results. So the testing we do for free and then we use the results.

Interviewer 1: And what would be the most important take away message from this test?

Interviewee: Of course you have a lot of technical things, like it would be nice if the camera could rotate. So for us the technical things are a side effect to understand if its actually valuable to their operation because if you here about a lot of features that need to be added, then it’s the questions whether it has the added value that they think and we think. So it is more about that to understand whether we need to pursue things or there’s no problem solution fit. So it is more about that, not so much about the technical results.

Interviewer 1: You also mentioned that you were able to throw the spear, what are the benefits of that?

Interviewee: The four wheel system you just need to put down, so if there is small stairs you already need to get up those which makes you vulnerable. So it is easier to go into places you don’t want to yet. Its also easier if you cleared the ground floor and you want information about the first floor, you can throw the robot up the stairs. So that you don’t expose yourself to early, as well for vulnerability as for element of surprise being gone.

Interviewer 1: Can this robot handle rolling down the stairs?

Interviewee: We did some testing on that, its successful, but of course things can damage.

Interviewer 1: Because I think it would be a cool combination with a drone to throw in on the top floor and as soon as it is done with the top floor it goes down by the stairs. Have you thought about this?

Interviewee: We were at a swat team training and they used a helicopter to drop it down on a building and then it swept down like you just explained, so that would definitely benefit their situation in my opinion.

Interviewer 1: Since your robot is not very heat resistant, would you consider outsourcing the development of a heat shield for your robot?

Interviewee: I am not sure if it is the most critical element of this robot yet. Because I’m not sure whether it will be used in those situations. They are using drones already in the building already, but not when the fire is at its most, cause then they are just focusing on containing the fire. So to be honest I don’t think so, and the main reason for that is you are adding additional costs for a feature that is not used so much yet. Until shown otherwise, so if from the next ten robots we deliver nine will burn in less than a month then we know, but id rather do it like that than double the price of the product when it might not be necessary.

Interviewer 1: But if you imagine for a second that you are in sudden need for a coating, what would be your requirements be in terms of weight, elasticity and thickness for instance.

Interviewee: My first view is indeed some sort of painting coat around it that is able to reduce the time it takes for the heat to go inside. In terms of weight, our two wheeled system is 1.2 kilo now so it cannot be more than 300 grams for this. and that is not because the robot is not strong enough, but it is because the fire brigade has to where the robot itself. The first robot we came in with was 2.5 kilograms and they said that’s a nice idea but it is too clumsy and too heavy to actually put this on our belt and take it with us. They have a lot of stuff on their body already. So in the military they say all the new things should be more light weight that what we now have. Fight light is their main mission right now, and I can only imagine that this is the same for the fire departments.

Interviewer 1: Are there any other things that were not discussed but you think could be interesting for our projects?

Interviewee: In the fire departments operation are two phases and the first phase is okay we come in and barely have an idea of what is happening in there and that is the most crucial phase to get information from the inside, because five minutes is quite long for a dynamic situation like that. And the second part where they do more of a monitoring and sustaining phase, those situations are way more suitable for an autonomous robot, but I wouldn’t try to combine those into one.

Interviewer 2: But if I understand correctly you are saying either focus on remote controlled and improving situational awareness or autonomous and finding people? Or am I not understanding that correctly?

Interviewee: Yes, and about this finding people part, even though they say there are not trained for this anymore if they know there is people in there, they are going to save them. They are trained to not take risks but, especially if there is kids inside, it is that above anything. So in that case there is no time for a teleoperated device anyway.

Interviewer 2: Oh that’s interesting because when I talked to the fire department they indeed said first control the fire and then they swoop.

Interviewee: Yes, but than they know that there’s no human in danger. Because i also spoke to a fireman and they said there was a house fire and a mother and the son were trapped on the ceiling and then they just make sure to get there. But those are only one or two time event for a fire department that does not happen on a weekly basis. So even though the fire department sound like they’re saving lives, their main job is to contain the fire and make sure other buildings are not damaged.

Interviewer 2: And have you heard about differences between houses or buildings like atlas, office buildings?

Interviewee: Yes.

Interviewer 1: We thought because there are a lot of people in atlas they will not evacuate in time, especially knowing they are students and would be sleeping maybe.

Interviewee: Indeed, its quite different because with bigger buildings especially, you can also think of elderly houses because they are stubborn, they do not hear the sound. Typically when such a fire comes in they scale up with more cars already. Typical housefires are not that exiting usually. So the bigger buildings are usually closer to their hero image. Because you also get extra layers of communication there, there is an extra officer of demand, the press needs to be there. By scaling up to three or four cars it already becomes more complicated.

Interviewer 2: Okay thank you for answering all of our questions, this will really help us design the robots as useful as it can be. If you have any further questions please do not hesitate to contact me, and I will stop the recording now.

Appendix E (Matlab script)

clear all

% Constants

q_heatgun = 2000; % Constant heat source (W)

T_ambient = 24; % Ambient temperature (°C)

A = 0.25 * 0.5; % area plate (m^2)

A_heatgun = pi*(0.02^2); % Area of hot air coming out the heatgun (m^2)

% Material properties

c_air = 700; % Specific heat capacity of air (J/(kg*K))

x_air = 0.15; % Thickness of air layer (m)

rho_air = 1.17; % density air

V_air = 3; % taking into account that heat will dissipate over more air than just the air between the heat gun and steel

m_air = rho_air * V_air; % Mass of air (kg)

h_air = 16; % Convective heat transfer coefficient for air to steel (W/(m²*K))

c_steel = 420; % Specific heat capacity of steel (J/(kg*K))

x_steel = 0.01; % Thickness of steel plate (m)

rho_steel = 8000; % Density of steel (kg/m³)

m_steel = rho_steel * x_steel * A; % Mass of steel plate (kg)

k_steel = 60; % Thermal conductivity of steel (W/(m*K))

c_ceramic = 1000; % Specific heat capacity of ceramic (J/(kg*K))

x_ceramic = 0.01; % Thickness of ceramic (m)

rho_ceramic = 800; % Density of ceramic (kg/m³)

m_ceramic = rho_ceramic * x_ceramic * A; % Mass of ceramic (kg)

k_ceramic = 0.12; % Thermal conductivity of ceramic (W/(m*K))

% Thermal resistances

R_conv_air = 1 / (h_air * A); % Convective thermal resistance from air to steel outer surface (K/W)

R_cond_steel = x_steel / (k_steel * A); % Thermal resistance of steel (K/W)

R_cond_ceramic = x_ceramic / (k_ceramic * A); % Thermal resistance of ceramic (K/W)

% Ratio of steel plate actually being heated

ratio = A_heatgun/A;

% Ratio becomes bigger over time because the heat spreads over the plate

ratio_function = @(t) ratio * (1 + 1*t/300); % Ratio increases by 100% over 300 seconds

% Total resistance through insulation

R_total = (R_conv_air + R_cond_steel + R_cond_ceramic);

% Total heat capacity

total_heatcapacity = m_ceramic * c_ceramic + m_air * c_air + m_steel * c_steel;

% ODE Function with updating ratio

%ode_temperature = @(t, T) (q_heatgun - (T - T_ambient) / (R_total * ratio_function(t))) / total_heatcapacity;

ode_temperature = @(t, T) (q_heatgun * (1 - 1 * exp(-0.015 * t)) - (T - T_ambient) / (R_total * ratio_function(t))) / total_heatcapacity;

% 1 - 1 * exp(-0.015 * t), This models the heat transfer more accurately

% and makes it more of a quadratic line

% Time span for simulation

time = linspace(0, 300); % From 0 to 300 seconds

% Solve ODE

[t, T] = ode45(@(t, T) ode_temperature(t, T), time, T_ambient);

% Plotting

figure;

plot(t, T);

xlabel('Time (s)');

ylabel('Temperature (°C)');

title('Temperature of the ceramic fiber cloth over Time');

grid on;