PRE2018 1 Group1

Preface

Usefull Links

via the link below notes from coach meetings and our own meetings can be found

notes

Group members

Name	Study	Student ID
Buijvoets D.C.J.T	Mechanical & Electrical Engineering	0902148
Cornet, N.	Industrial Design	1007926
Van Horssen, C.	Software Science	0885378
Mouw, F.A.	Applied Physics	1005735
Stokbroekx, D.L.M.	Mechanical Engineering	1010326

Initial robotic concepts

After discussing several subjects from different fields in society, we came up with the following list of robotic concepts which have potential to solve problems faced by certain users.

• Fire fighter drone which can be used to aid fire fighters
• Cleaning drone for difficult to reach spots in buildings
• Pavement cleaning drone which can remove dirt from tiles
• Avalanche rescue drone that helps rescuing teams search for victims in an avalanche using already available beacons
• Weeding robot which can differentiate between wanted and unwanted plants in a garden and remove the weeds
• Referee robot using image processing to determine the state of a match in e.g. soccer or tennis
• Medical nanobots for drug delivery
• Fruit harvest robots

Chosen concept: Fast Response Extinguish Drone, F.R.E.D.

Eventually we agreed upon the concept of a fire fighter drone. Fire fighting is one of the most dangerous jobs and every year people are still being killed by fires. Extra preparation and information on the fire site can make the difference between life and death. That is where we thought a drone could be of help.

Problem statement

How can drones be used in combination with smarthomes for fully autonomous fire fighting. In this project the focus will be on the fire extinguishing part of fire fighting.

Objectives

• The smarthome needs to give victims information about what they can do to improve their chances of survival without injury
• It needs to be able to communicate the current state of the fire back to the fire department to inform and prepare firefighters
• The drone has to be able to fly inside a building autonomously
• It has to have the ability to extinguish small fires or reduce the intensity of larger ones

During this project, our focus will especially be on the last objective.

Project setup

Before actually starting the project, a setup is made on how it will be executed.

Approach

• Decide functionality of the drone
• Do research on different subjects concerning the functionality of the drone
• Design drone and functions
• Make prototype
• Test prototype

Milestones

For the duration of this course, the following milestones are selected:

• Week 1: Every member will take the time to do research on robots and their interests, in order to broaden one's horizon on the possible subjects. Afterwards a subject for the project will be chosen.
• Week 2: Literature study and further research will be completed
• Week 3: USE analysis is finished
• Week 4: Design for the first prototype will be finished
• Week 6: First prototype will be finished
• Week 8: Final Design, final prototype and all other deliverables will be finished

Deliverables

The following deliverables will be created during this course:

• Thorough research and literature study
• Design research process and report
• A functional drone design
• (several) Prototypes
• Ethical evaluation
• A wiki page on this domain

Planning: Who's doing what

Week	All	Dirk	Natanya	Chiel	Fabian	Daan	Undecided
1	Finding suitable projects + finding articles on the chosen subject		Basic user requirements			Formulating Problem statement
2			Further elaborating user requirements and USE analysis + making persona’s of the users	State of the art research	State of the art research	State of the art research	Further elaborating user requirements and USE analysis + start making designs for the robot
3	Checking wiki and correcting formulation if necessary						Continuing on design for the robot + finishing USE analysis
4		Work on USE requirements		Starting on control software for the robot	Starting constructing of the prototype	Starting constructing of the prototype	Finishing robot design
5	Checking wiki and correcting formulation if necessary	Create a specific situation for use			Construction of the prototype	Construction of the prototype
6	Testing and evaluating prototype + find and analyse flaws in the design + fixing the design flaws				First prototype finished	First prototype finished
7	Find and analyse flaws in the design + fixing the design flaws + checking wiki and correcting formulation if necessary		Preparing presentation	Preparing presentation
8	Final design and robot finished + Presentation of the result

USE analysis

Relevant users and their requirements

• Fire fighters will require a fast response time to the fire and an accurate analysis of the fire site
• Victims require information on how to reduce harm to themselves and others and a safe rescue
• Homeowners/insurance will want as little damage as possible to property.

This topic revolves around a few different stakeholders, which we will look at down below. Through interviews with the fire department in Eindhoven, the Gezamelijke Brandweer Drone Unit in Rotterdam and the Brandwondencentrum Maasstad Ziekenhuis Rotterdam we’ve gathered lots of insights in the different aspects and needs of the stakeholders involved. An overview of the interview can be found at Expert Interviews

Main users

Fire victims

Being amids a fire, is a very stressful and traumatic experience. Victims during a fire can either be conscious or unconscious. When conscious, the victim needs to be reassured that everything will be alright and that they can in most cases easily get to safety by listening to the advises of the smart home. When asleep it is important that the victim is woken up as soon as possible.

Requirements:

Make victim feel safe The smart home needs to try to keep the victim as calm as possible, by assuring that help is on its way and to explain every action it is taken. The victim should not feel threatened by the smart home or drone.

Give instructions After analysing a situation, the smart home needs to, if possible, give instructions to the victim of possible actions to improve chance of survival of the victim, this can be a possible path of escape to the emergency exit.

Cannot (lethally) harm victim in any way The drone and smart home are not allowed to harm the victim in any way, or do something that brings lethal injury to the victim.

Wake the victim If the victim is sound and asleep, it is very important that they are to be woken up as soon as possible. Preferably though a manner that is both fast but also keeps people from panicking.

Other stakeholders:

Building owners Requirements Building should sustain as less damage as possible

Firefighters/fire department Requirements Alarm The smart home needs to automatically sent a notice of fire to the fire department the moment it is detected Escape of victims Victims should escape the burning building as soon and safe as possible

We discovered several things through the USE-analysis. Firstly, we changed our first design requirements from a drone that needed to be able to extinguish a fire, to a drone that needed to just control the fire as long as possible, and extinguish when possible. We accumulated these insights through our meetings with the Eindhoven Fire Department and especially at the Gezamenlijke Brandweer Rotterdam, where they informed us that buying time by controlling a fire rather than try to extinguish it with the drone would be way more valuable. This way there is more time for victims to escape and for the fire department to arrive. The requirement RSM 4 directly was created after the meeting with the Gezamenlijke Brandweer Rotterdam.

Our first idea was that we created just a drone, not a smart home, that would both try to extinguish the fire and give instructions to bystanders so that they could successfully escape. Through our discussion with Anneke van de Steenhoven at the Brandwondencentrum Rotterdam, we came to the conclusion that there needs to be more attention given to the bystanders, so splitting up the task of informing the victims and controlling the fire would be a more beneficial. We learned that people, although maybe professionals, freeze or panic when in a stressful situation. If there is a cool and calm third party giving them instructions of what to do when an emergency situation surfaces, people will have better faith, stay calmer and thus a better chance of survival. This resulted in the creation of not only the F.R.E.D. but also the smart home application.

Personas

Andrea and Frank are two examples of your average humans that could be living in a smart home with a F.R.E.D.

Scenario

To make a good design for F.R.E.D. we are considering a certain scenario. An illustration of the scenario can be found in the two cartoons below.

Relevance to society

• People living in range of the fire station using this drone will be subject to better rescuing in case of a fire
• If the drone is being used to aid at fighting a fire, the people living in the viscinity of that fire will have a smaller chance that the fire will affect them

Expert Interviews

Fire department TU/e

Currently the fire department is running a project with the innovation space, by accident a group member came in contact with someone involved with this. They had a demonstration/meeting with the TU/e fire department planned and we were invited to come and have a look on Monday the 24th of September. Also we received a contact within the fire department, this person is responsible for innovation and repression specifically in south-east Brabant.

Conclusies: (Interview met Brandweer Eindhoven)

We gaan opzoeken wat veelvoorkomende brandjes zijn en daarop blusstof voor bot baseren.Poeder is effectiever vollume/effectiviteit en gewicht/effectiviteit dan alles, maar doet meer dan alleen de brand uitzetten.

We moeten uitzoeken wat de tijden precies zijn die de robot heeft om een fikkie te blussen, en welke brandmelders dus ook het eerste aangeven.

Niet alles is best case scenario (sunny day). Denk aan slingers tijdens een feestje of verouderde huizen.

Het is prettig als drone vuur contoleerd, probeerd te blussen en gelijk daarna alle mensen ontruimd

Drone Unit Fire Department Rotterdam

A conversation was held with the director of the rotterdam fire department, which actually has a fire fighting drone in operation at the moment. This drone is autonomously navigated by another (controlled) vehicle which makes a 3D map of the inside of a building. We will have the opportunity to talk to the project leader of this drone about some points of improvement.

We will visit the Rotterdam fire department on Friday 28th of September

The “Gezamelijke Brandweer” in Rotterdam has been using their drone for 5 years. They are required, when using it, to always have the drone in line of sight. The drone has a reach of 5km, but the firefighters are restricted to a limitation of 500m and 120m high. It can fly on a single pack of batteries for 35 minutes and can lift 6kg. Most importantly, it flies stable, even above a fire. Which is something not every drone can do.

They are of the opinion that a drone is best used to gather information, while the drone is very mobile. Through a normal camera and a heat camera, they can investigate and assess the situation. That way the firefighters can go into the building more prepared and less casualties will fall. Fast response time is crucial when it comes to a rescue mission. It, however, does take 3 people now to fly the drone. (One keeps track of the location, one steers it, one steers the camera). Furthermore, the heat camera can be of a big help when locating and assessing the state of eventual victims. People who have passed away (colder) will have low priority than people who are still alive. Every operation starts with a headcount of people who are (supposed to be) in the building.

It is forbidden by law to fly for private use. Only people from Brandweer Nederland with a pilot license are allowed to freely fly right now.

Their recommendation was to focus on controlling fire with the drone, and not try to extinguish it. 1L of water produces 17L of steam, which is ideal to control fire with. Our choice of drone can carry up to around 0.5kg, which would produce 8.5L of steam.

Audio recording of conversation: https://storage2.cvhorssen.nl/s/MLJdKtEQ0PKYhcg (download available until 31-12-2018)

Audio can be played from browser without file download or account (click on the big icon in the center).

Brandwonden Centrum Maasstad Ziekenhuis Rotterdam

At the Maasstad Hospital, we spoke with after-care specialist Anneke Steenhoven.

Recently, a lot of home based fires have been happening. If someone is by accident on fire, the first thing one is supposed to do is roll on the ground. This minimizes the available oxigen for the fire to burn. However, in mids of panic people tend to forget this and start running around. Also a third party or experts tend to freeze or even panic in these situations.

If one finds out there is a fire in their home, the best idea is to stay calm and call 911. When burned, immediately cool with cold water. The burn should absolutely not be touched.

State of the art: Literature study

One of the most important facets of this project is to come to an understanding of the current state of technologic advancement that is relevant to the drone and its functionalities. Therefore, literature on different relevant catagories is studied to understand the state of the art. Here follows a list of the scientific research that was studied for this project, divided into the different relevant catagories.

Current smart-homes are mostly focused on creating a seamless living environment for the occupants of the house, or to create an energy efficient management system called “smart energy”. Whereas our project focuses on optimising the safety and mental state of the occupants in case of a fire.

State of the Art Smart Homes

MavHome ((Managing An Intelligent Versatile Home) [3] The MavHome smart home project focuses on the creation of an environment that acts as an intelligent agent, perceiving the state of the home through sensors and acting upon the environment through device controllers. The agent’s goal is a function that maximizes comfort and productivity of its inhabitants and minimizes operation cost. In order to achieve this goals, the house must be able to predict, reason about, and adapt to its inhabitants.

The MavHome combines different technologies to understand it’s user (artificial intelligence, machine learning, databases, mobile computing, robotics, and multimedia computing). MavHome features include collection of activities in a database, prediction of inhabitant actions, identification of inhabitants from observed activities, mobility prediction, robotic assistants, multimedia adaptability, and intelligent control and visualization of home activities.

MavHome predicts the user’s net action, through machine learning. By looking at how the user has acted and reacted in the past, it is able to conclude what is the most probable follow-up action.

The Georgia Tech Aware Home [1] This smart home focuses on Health and Well-being, sustainability, Entertainment and Connected Living. Although it does focus itself on Health and wellbeing, this is mainly focused on aging in place (assisted living for the elderly).

For localisation, they use PowerLine Positioning. It is very cost-effective and it uses tags that don’t alter the room itself to track products or humans. Through a low-resolution ceiling camera (watching the movements) the home is able to determine the “mood” of the room. This means whether people are busy or not.

MIT Intelligent Room [2]

The Intelligent Room project is focused on the vision of creating an easy to install smart home environment. Not much should be changed to the environment to get it to work. They believe that the computer environment should be transported to the “real world”, instead of people into the virtual environment (WIMP/VR).

The Intelligent Room is able to detect people and determine their activities and gestures, and even if they are talking to each other or to the smart home.

Through a fixed wide-angle and steerable narrow angle camera the smart home software is able to detect if there are people present in the room. It does this through movement detection software.

The Neural Network House [4] This smart homes focuses itself on creating an smart environment that adapts itself to the occupants. The home needs to program itself to the user needs, the user has no task in this.

This system is called ACHE (adaptive control of home environments). ACHE both monitors user preferences, and then learns to execute these. Secondly, it focuses on being as energy efficient as possible (e.g. turn of the lights when not needed). ACHE can not visibly track the user. It only keeps track of the preferences in the home (e.g. termostat). It has no clue of the concept “human”. In that sense it also can’t accompany to different users with different preferences and needs.

Why Smart homes?

In our interview with burn after-care specialist Anneke Steenhoven, it became clear that even professionals sometimes fail to act coolly and professionally in situations of stress and panic (such as a fire). Even though they know what to do, they have a black-out or just freeze. Although not tested yet, it is very probable that if a third party would give advice or instructions, people can react faster and more thoughtful in such situations. People may not be able to come up with a solid plan themselves, but they can follow those of others. In our case this third party would be the smart home.

As also seen in the previous examples of the current state-of-the-art, it is already possible to sense where people are in the house or even in the room. In case of a fire, it is our mission to accompany the occupants of the house to the exit in the safest way. For this to happen, it needs to be known where the occupants are located in the house. Then it can be determined which route can be taken to the safest exit. Our definition of the safest exit is always the exit that is closest to the occupant and in another room than the fire.

It is our supposition that people will stay calmer because they are told exactly what to do, in a calm matter. This will create the idea that they are in safe hands, although they still have to execute the advices themselves.The smart home has already assessed the situation and drawn a plan for them.

Conclusion

The user localisation that already exist in current smart homes seems very practical, a component we could also use for our smart home. The action prediction, however promising, is not something we could use. The action prediction works through machine learning. The smart home observes the action and reaction of the users and can through that determine what the possible next action of a user could be. In a panic situation, however, people tend to act unpredictable and different from what they would normally do. In case of a fire the same holds true. That leads to us believing that this technology is not yet advanced for us to implement it in our concept.

None of the current promising smart homes focus itself on providing help in case of an emergency situation. Although research has been done by different parties [5], it has never been implemented until now. This is why our research could prove to be a valuable asset in the current state-of-the-art.

Aware Home Research Initiative. (2018). Retrieved from http://www.awarehome.gatech.edu/drupal/

2. Brooks. R.A. The intelligent room project: cognitive technology, in: Proceedings of the 2nd International Cognitive Technology Conference, Aizu, Wakamatsu, Japan, 1997, pp. 271–278 taken from: http://people.csail.mit.edu/brooks/papers/aizu.pdf

3. Cook. D.J. et al. (2003). MavHome: An agent-based smart home. Retrieved from: https://www.researchgate.net/profile/G_Youngblood/publication/4011271_MavHome_An_agent-based_smart_home/links/5451065f0cf2bf864cba8691/MavHome-An-agent-based-smart-home.pdf

4. Lee, K. (2018). Network-based fire-detection system via controller area network for smart home automation - IEEE Journals & Magazine. Retrieved from https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=1362504&tag=1

5. Robles. R.J., Kim. T.H. (2010). A Review on Security in Smart Home Development. Retrieved from: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.178.1685&rep=rep1&type=pdf

Drones in firefighting

Drones and robots are already extensively used in firefighting but there are few examples of actual autonomous fire extinguishing drones. Most drones in firefighting are used to support humans by providing vital information on the emergency.

Drones provide information on how fires are spreading and developing to firefighters to increases the efficiency of the process by sending units to right places. The information gathered by these drones can also be further analyzed for a better understanding of how fire spreads for optimization of extinguishing techniques. One of the largest fire departments in the world the L.A.F.D. (Los Angeles Fire Department) is currently researching and using these kind of drones . The drones have been used for the first time during the massive wildfires in December 2017 and have been a great help in the firefighting process. These drones do still require a human pilot and a human to analyze the information.

Drones also help emergency services with other non-fire related disasters. For example, during floods or earthquakes these drones can be in the disaster area much faster to collect information on the situation for the emergency services.

A different application for drones in firefighting is the localization of victims in fires that would otherwise never be found. Currently one of the best drones for this undertaking is the Firestorm UAV . This drone can find victims inside burning buildings using a thermal camera also this drone can detect toxic gasses and inform firefighters on the situation in a building. Using bright LED lights, the drone can lead a localized victim along the safest path outside of the building.

The emergency services also use the drones to make emergency deliveries to certain disaster areas. The payloads of the drones can be extremely diverse, from AED machines, medical- and food supplies. One of the drones currently in development for this purpose by the company ZIPLINE . The drone that is currently being tested in the USA and already operational in Rwanda for blood deliveries to rural hard to get areas drops the payload mid-air before it flies back to base to be resupplied and launched again.

Another way in which drones are used by fire departments is to make pre-fire plans for high risk or vital buildings. These drones can map escape routes and localize water supplies and potential problems for these buildings.

There are some drones that can also extinguish fire. But these drones are not autonomous and require a human pilot or firetruck for water supply and further support. An example of drones used for actual fire extinguishing are the drones produced by the company AERONES . Although these drones are capable of combatting fires they cannot operate without the presence of a fire truck supporting the drone with water and electricity. This drone can reach heights of 300 meters which is much higher than the height a traditional fire truck can reach. However due to the hoses connected to the drone it can only be used for combating the fire from the exterior. This drone can be useful for fires in high-rise buildings or at other great heights that were traditionally hard to reach.

General Drone Information

• Remington, Raquel, et al. "Multi-Purpose Aerial Drone for Bridge Inspection and Fire Extinguishing." (Unpublished Thesis). Florida International University. Retrieved April 10 (2014): 2016. (Fabian)
• Suresh, Jayanth. "Fire-fighting robot." Computational Intelligence in Data Science (ICCIDS), 2017 International Conference on. IEEE, 2017.(Fabian)
• Design of a portable robot/device that is able to gather environmental information about the fire and guide victims for evacuation: Kim, Y.-D., Kim, Y.-G., Lee, S.-H., Kang, J.-H., An, J. “Portable fire evacuation guide robot system” (2009) IEEE/RSJ International Conference on Intelligent Robots and Systems, IROS 2009, art. no. 5353970, pp. 2789-2794. (Daan)

Autonomous drone navigation

Outdoor autonomous flight has been researched for some time now, since the use of GPS can be a great tool for this application. For a fire fighting drone however, indoor autonomous flight is the real challenge, since the drone has to fly to fires autonomously indoors. The localization of fires can be done by the smart home, and can be communicated with the drone. The drone itself however has to localize itself in its environment and has to be able to plan a trajectory to the fire while avoiding obstacles. Flying autonomous in a GPS-denied environment requires real-time tracking and mapping of the surroundings of a drone using sensors like 3D cameras or laser scanners.

A fully autonomous navigation controller has been made which uses a 3D laser scanner for omnidirectional environment perception. An egocentric grid map can be made an updated in real-time. This map is merged to an allocentric map (of the environment) to localize the drone. The controller can also generate global trajectories (from start to end) and local obstacle avoidance trajectories which together make for a safe autonomous navigation. It is shown that this multilayered navigation planning enables the controller to cope with dynamically changing environments, such as a house fire.

Another approach to obstacle avoidance in indoor environments using only 2D imaging and on board inertial sensing has been researched. This approach makes use of patterns on the ground for localization (not ideal for unknown environments). It is shown that unknown obstacle avoidance using 2D imaging is feasible.

Furthermore, a navigation controller that estimates location and trajectories based on imperfect sensory input has been developed. This controller contains parameters which allow to change the accuracy or speed of the trajectories. Safer navigation can be achieved using this controller.

Autonomous victim detection

• Detecting injured humans on images taken from aerial vehicles: ANDRILUKA, Mykhaylo, et al. Vision based victim detection from unmanned aerial vehicles. In: Intelligent Robots and Systems (IROS), 2010 IEEE/RSJ International Conference on. IEEE, 2010. p. 1740-1747. (Chiel)
• Building maps and marking victims on those maps using hyperspectral imaging: TRIERSCHEID, Marina, et al. Hyperspectral imaging or victim detection with rescue robots. In: Safety, Security and Rescue Robotics, 2008. SSRR 2008. IEEE International Workshop on. IEEE, 2008. p. 7-12. (Chiel)
• Victim detection using an adapted Viola-Jones algorithm: DE CUBBER, Geert; MARTON, Gabor. Human victim detection. In: Third International Workshop on Robotics for risky interventions and Environmental Surveillance-Maintenance, RISE. 2009. (Chiel)
• Using ad-hoc network with base station (firetruck or fire department?): SUGIYAMA, Hisayoshi; TSUJIOKA, Tetsuo; MURATA, Masashi. Victim Detection System for Urban Search and Rescue Based on Active Network Operation. In: HIS. 2003. p. 1104-1113. (Chiel)
• False positive reduction on victim detection from colored images: KLEINER, Alexander; KUMMERLE, Rainer. Genetic MRF model optimization for real-time victim detection in search and rescue. In: Intelligent Robots and Systems, 2007. IROS 2007. IEEE/RSJ International Conference on. IEEE, 2007. p. 3025-3030. (Chiel)
• Detecting victims using pseudo-noise radars, whose signals scatter from body motions of victims: SACHS, Jürgen, et al. Trapped victim detection by pseudo-noise radar. In: Proceedings of the 1st International Conference on Wireless Technologies for Humanitarian Relief. ACM, 2011. p. 265-272. (Chiel)

Fire detection

• Detecting fire from colored images, distinguishing fire and smoke: CHEN, Thou-Ho; WU, Ping-Hsueh; CHIOU, Yung-Chuen. An early fire-detection method based on image processing. In: Image Processing, 2004. ICIP'04. 2004 International Conference on. IEEE, 2004. p. 1707-1710. (Chiel)
• Detecting fire using space-time fluctuations on colored images: YAMAGISHI, Hideaki; YAMAGUCHI, JUNICHI. Fire flame detection algorithm using a color camera. In: Micromechatronics and Human Science, 1999. MHS'99. Proceedings of 1999 International Symposium on. IEEE, 1999. p. 255-260. (Chiel)
• Detecting fire using Gaussian distributions: CELIK, Turgay, et al. Fire detection using statistical color model in video sequences. Journal of Visual Communication and Image Representation, 2007, 18.2: 176-185. (Chiel)
• Fire detection in tunnels using cameras and infrared: NODA, S.; UEDA, K. Fire detection in tunnels using an image processing method. In: Vehicle Navigation and Information Systems Conference, 1994. Proceedings., 1994. IEEE, 1994. p. 57-62. (Chiel)

Fire suppression

Today there exists a wide variety of methods to suppress fires. Since this project is based around a drone, one of the main concerns is the weight of the fire suppressant. It should be as light as possible per amount of fire that it can put out. Furthermore, our method should be able to extinguish an as wide variety of fires as possible. Especially fires of class A, B, C and F (European standard) seem to be most common in buildings.

The use of a pressurized water mist extinguisher seems to be impractical. Although an extinguisher of this type is able to suppress fires of type F (cooking oils and fats), next to ordinary fires, a lot of water is needed to put out an average fire. Typical amount of 9 litres are often required , which is much more than an average drone can carry next to its own equipment. Furthermore, the suppression of fires using water mist often results in a large fire cloud in the process, due to the increased heat transfer that is caused by the water droplets. This would be impractical as it could harm the drone.

Another method is the use of particulate aerosols. Often, particles are generated from a solid or gel and mix with the air. Particulate aerosols prove to be a very lightweight alternative for water, with results showing the same fire suppressing abilities at a 30 times lower volumetric flow, compared to normal water.

• Experiments of different solid particulate aerosol suppressants in the form of a solid, gel or powder: Kibert, C.J., Dierdorf, D. Solid particulate aerosol fire suppressants (1994) Fire Technology, 30 (4), pp. 387-399. (Daan)
• Testing the effectiveness of using nanocomposites as additive to conventional powder suppressants: Ni, X., Kuang, K., Wang, X., Liao, G. A New Type of BTP/Zeolites Nanocomposites as Mixed-phase Fire Suppressant: Preparation, Characterization, and Extinguishing Mechanism Discussion (2010) Journal of Fire Sciences, 28 (1), pp. 5-25. (Daan)
• Different delivery systems (with cooling) for the stream of aerosol coming from a pyrogenic aerosol fire suppressant Vitale, J., Kibert, C., Akers, l. PYROGENIC AEROSOL FIRE SUPPRESSANTS:

ENGINEERING OF DELIVERY SYSTEMS AND CORROSION ANALYSIS.

• Information on and tests with solid dispersion aerosols Kibert, C., Dierdorf, D. ENCAPSULATED MICRON AEROSOL AGENTS (EMAA)

Fire resistant materials

• Lyon, Richard E., et al. "Fire‐resistant aluminosilicate composites." Fire and materials 21.2 (1997): 67-73. (Fabian)
• Myeong, W. C., Kwang Yik Jung, and Hyun Myung. "Development of FAROS (fire-proof drone) using an aramid fiber armor and air buffer layer." Ubiquitous Robots and Ambient Intelligence (URAI), 2017 14th International Conference on. IEEE, 2017. (Fabian)
• Myeong, Wancheol, Kwang Yik Jung, and Hyun Myung. "Development of a fire-proof aerial robot system for fire disaster." World Congress on Advances in Nano, Bio, Robotics and Energy (ANBRE). IASEM Conferences, 2017.(Fabian)
• Abbott, N. J., M. M. Schoppee, and J. Skelton. Heat Resistant and Nonflammable Materials. FABRIC RESEARCH LABS INC DEDHAM MA, 1976. (Fabian)
• Luo, Qiu-Sheng, Shi-Feng Li, and Hui-Ping Pei. "Progress in titanium fire resistant technology for aero-engine." Journal of Aerospace Power 27.12 (2012): 2763-2768.(Fabian)

Hardware and design

Choices made on design and hardware of F.R.E.D. will be discussed and substantiated here. The required response time and required extinguishing capabilities will determine the choice for a drone

Overall system

F.R.E.D. works in a system, consisting of a drone and a smarthome. These two will work together to combat the fire and ensure the safety of the people in the house.

The smarthome must be able to detect a fire, this will be done with fire alarms as are now installed in most homes or heat camera’s. Since the smart home knows the location of the alarm in the house the system and thus the drone knows where to look for the fire. In order for this to work optimal a fire alarm should be placed in most rooms especially those with a higher risk of fire for example kitchens. The use of heat cameras is more expensive but it will be a few seconds faster and could be used for more tasks than just detecting fire, for example detecting the location of people in the house.

• After detecting the fire the main goal of the system is to safely evacuate the people in the home. The task of the drone is to delay and/or extinguish the fire to give the people more time to evacuate.

Within half a minute after the ignition of the fire an connection will be established with the fire department. In this connection the following information is transmitted to the fire department:

• That there is a fire
• Give a headcount and status of the people (and children) inside the building.
• Providing schematics of the building.
• Actions taken by the house to bring people to safety and delay or extinguish the fire.
• In case of special fire (I.E. pressurised gas or electric) information on materials that are burning or close to the fire.

Communicating with the occupants of the house goes via the Smart Home and has the following goals.

• Informing occupants there is fire and what they should do.
• Keep occupants save/guiding occupants outside
• Keeping occupants calm.
• In case of children, have adults retrieve kids before exiting building as long as this does not endanger the adult.

Response Time

The most common types of fire in and around homes are: kitchen fires, electrical fires, heater fires and smoking related fires. In the end a drone capable of responding to all these fire types is preferable. Most fires get out of control fast, usually within minutes. Therefore, it’s important that the drone can be on side very fast. The response time of the drone depends on several aspects: the location where the drone is stationed in the building, the speed at which it can move through the building and the way the fire is reported to the drone.

Placement in building

When placing F.R.E.D. in a building it is important for a small response time that the drone is located close to rooms or areas with higher risks of fire. In regular homes with homeowners that do not smoke the kitchen has the highest risk of hosting a fire. But when there are people who live in the house that smoke indoors the number of rooms with high risk of fire increases drastically. Every room where a person smokes inside is a potential fire hazard. From our conversation with the TU/e fire department became clear that houses where elderly people live are also more likely to catch fire.

When finding a suitable location to place the drone one very important user requirement must be considered. This requirement is that the user doesn’t constantly want to see or bump into the drone in his home. Therefore, the drone should be placed on a discrete location preferably close to all high-risk areas in the house. From now on the assumption will be made that the homeowners do not smoke or do not smoke inside their house. This means that the highest risk of fire is in the kitchen. Since the most ideal situation for the drone to operate in is a smart home environment this will also be considered for finding a good location, this means the drone knows his way around the house and is able to remotely open doors inside the house for it to move from room to room. When a new smart home is built with F.R.E.D. integrated in it, there are a few good locations for placing the drone are:

• The first possibility is to place the drone in the ceiling, in this way the drone will never hinder the user, but it may cause some difficulties for maintenance and routine checks of the drone. It’s therefore important that the drone will be easily accessible in a safe way for the user. However it's also very important that the user does not always bump into the drone.

• A different possibility is to place the drone inside a wall, this can be in a horizontal or vertical position. A drone may have some difficulties to lift of in a vertical position especially when it is heavy. Therefore, a mechanism is required that first brings the drone horizontal before it can be launched when placed vertical in the wall. A disadvantage for the user is that a piece of wall must remain free and unblocked for the drone. When the drone is placed in the wall on an appropriate height, let’s say around 1.20 meters, it’s easily accessible for maintenance or routine checks. Placement on the wall will however hinder the user sufficiently and it can also be very ugly in your house.

• The drone can also be placed inside the floor, this is comparable to placement in the ceiling but in this situation the drone is better accessible by the user when required. A major disadvantage of this is that some places of the floor can’t be used by the user for placing furniture. Also, it’s required that the drone isn’t located somewhere where there is a door, in a hallway or on the evacuation route because this may hinder the user in getting around in his own house.

• Another way to place the drone is in empty spaces like a garage, storage room or attic. An advantage of these locations is that a drone here won't hinder the user in any way. However the user has to keep these spaces quite organized and tidy. When the surroundings hinder the drone to much it can't fly and be on time on the disaster area.

Considering these options the best spot to place the drone is in the ceiling of a building. However in existing buildings the ceiling may not always be an possibility for placement of the drone. When the ceiling is not an option first empty spaces like garages or attics must be considered. There can also be homes that don't have these spaces for example apartments. In these situations the best way of placing the drone would be in a specially designed box on the wall.

Reporting the fire

When a proper location for F.R.E.D. has been found a way of reporting the fire to the drone must be chosen. This can be done via an automatic fire alarm using smoke and/or heat detectors. Since a smart home environment is considered the drone knows the location of the fire when the alarm is initiated. A different possibility for reporting the fire is manually by using a switch on a wall or an app on your smart phone. Manual switches on walls to trigger the fire alarm can already be found in a lot of larger buildings. The only issue with these is that they don’t look very nice and a user might not want to have these switches in every room of his house. Therefore, an app on the user’s smart phone takes away this issue for the user. When a fire is reported via manual switches on the wall or via an app the drone also knows where the fire is located if the alarm is being sounded in the room where the fire is.

For automatic fire detection there are two main types of alarms: photoelectric- and ionization smoke alarms. For the drone to respond fast there should be smoke detectors and alarms in as many rooms as possible. Ionization smoke alarms are usually more sensitive to flaming fires and photoelectric ones are more sensitive to smoldering fires . Since most fatal home fires are a result of smoldering fires it’s best to use photoelectric detectors in the house. The time until the alarms goes off depends on the amount of smoke produced by the fire, but by analyzing fire development movies it’s safe to say that most alarms are triggered around 15 seconds after ignition.

Travel distance and time

The final step for determining the response time of the drone is the speed at which it can fly. But another very important factor for the speed on the response time is the distance the drone must travel. An average newly build Dutch house is 116 m2 , consider this house has 2 floors and is square this would mean the square has sides of 7.6 m. When the drone is located as far away from the fire as possible this would mean it has to travel to the other side of the building change floors and go back to the other side of the building. This would mean the maximum distance it has to travel in an average home also considering walls and moving up to another floor is around 25 meters. A drone that would be very capable to use as a basis for F.R.E.D. (unfortunately not available for this project) is the DJI Matrice 100, it can carry up to 1.2 kilos and has a top speed of 18 meters per second . So, the top speed of the drone isn’t really an issue. But the drone also must make turns and should not further endanger the people in the house. Therefore, the top speed of the drone should be limited to 1 meter per second because the drone will still be able to reach the fire within 30 seconds when it’s furthest away from a fire in an average home and with the drone traveling at this speed user in the house can anticipate on the drone moving through the house.

When taking the deployment time, fire alarm response time and distance the drone must travel into account the overall maximum response time of F.R.E.D. in an average house will be around 55 seconds. But is this sufficient? From analyzing fire development movies some conclusions can be made . Fire development goes very rapidly and accelerates over time. After around 1 minute most fires are still isolated and easy to extinguish. When the drone isn’t on site extinguishing within 2 minutes after ignition it cannot do much more. Usually in around 3 minutes the entire room ignites, and the home is lost. From this can be concluded that a maximum response time of 55 seconds for F.R.E.D. should be sufficient enough to slow completely extinguish the fire or slow the fire down and buy more time for people to flee the building. From our conversation with the TU/e fire department became clear that it’s vital to act in the first minute of the fire. From the fire department perspective, the most important part is to get all people in the building to safety and not if the entire building burns down or not. Homeowners however also put a lot of value in to their belongings and an effort to extinguish the fire in the early stages of the fire might be able to safe their precious belongings.

Fire suppression

Extinguisher requirements

The main idea of the drone’s ability to suppress fires is that it can react quickly to small fires that have not gone out of control yet. It would be unrealistic to assume that a single drone without a continuous water supply would be able to extinguish large house fires, so the focus lies on preventing large fires by reacting quickly.

One of the most common types are kitchen fires. Amongst these, grease fires occur frequently and can be very hard to control. Other kitchen fires can include oven fires and electrical fires. Since grease fires are among the most common types of fires and one of the most challenging to suppress, we set the drone’s goal at being able to suppress an average grease fire (resulting from heating grease in a pan up to its auto ignition temperature).

If the drone is able to suppress this type of fire (type F), we can almost assume that it will be able to suppress fires of type A and B of the same scale too. Though further experimentation might be needed to validate this assumption.

Before choices on how the design for the extinguish mechanism can be made first the requirements for the mechanism must be formulated. In the table below the 7 main requirements for the extinguish mechanism are given with a short comment and priority. The priority scale is 1 = mechanism won't work without it, 2 = important for mechanism to work properly, 3 = not critical to functioning of the mechanism.

Number	Requirement	Comments	Priority
REM 1	Drone must be able to automatically activate the mechanism without human intervention	When the drone can't initiate the mechanism by itself it's useless to the user	1
REM 2	The drone must be able to extinguish for at least 1 minute		2
REM 3	The extinguish mechanism must use foam	Foam is usefull against most type A, B and F fires. Most fires that occur in and around homes are of these types.	1
REM 4	The foam used may not cause damage to the drone or environment when using it		2
REM 5	The mechanism must be refillable after usage	This means the drone is reusable after use, this very important since the definition of a robot is that it must be reusable	3
REM 6	Weight should not exceed 0.5 kilo	Drones can only carry a certain amount of weight however there exist drones that can carry much more these drones are however way to large to operate inside a normal house.	2
REM 7	The mechanism should not disturb the stability and flying capablilities of the drone	When the mechanism makes it impossible for the drone to fly and/or get to the fire in time it's useless	2

Extinguish mechanism

Since the drone has to be able to extinguish fires unmanned, and preferably autonomous, it must have an electro-mechanical extinguish mechanism of some sort. An actuator attached to the drone should be able to control the stream of suppressant (at least on or off). Furthermore, the stream of suppressant should be ejected to the front (or sides) of the drone through a nozzle. A drone cannot fly directly above fire since the air will be too turbulent and therefore the suppressant cannot just be dropped underneath the drone. There should also be a container to hold all of the suppressant.

The actual mechanism will depend on the suppressant that will be used. It could consist of a solenoid valve, since this type of valves can be controlled electronically. The driving force behind the suppressant can be a preloaded pressure in the container, a pump or a chemical reaction. This also depends on the used type of suppressant and the design of the container.

Extinguishing using a solid aerosol (blusstaaf)

One option we found is a ‘blusstaaf’. This is a complete aerosol extinguisher with an electrical activation. It weighs 250 grams, has 45 grams of suppressant and can extinguish for 30 seconds. They cost 39,95 a piece. Because this price is a bit steep and to maybe have a lighter and more compact system, we also want to look at the possibility of making the aerosol suppressant, container and electrical activation ourselves in a way that is more suitable for the drone.

We are still waiting for a reply from the guys at blusstaaf. They said they could provide us with technical information about the blusstaaf.

Extinguishing using a small pump

Since it's not sure if the blusstaaf mechanism is possible/available in this timescale and/or budget different extinguish mechanisms have been designed. These mechanisms are all water based since this is much easier to realize for the prototype of the drone. There are two possibilities to make an automated extinguish system using water. One using a compressed air and the other one using an electric waterpump. In the first picture below a mechanism using an electric waterpump is shown. There are lots of suitable pumps available (example [1]). This design exists of a water container (blue box), power supply, remote activation mechanism, pump (yellow box), small water hoze (green), nozzle (orange) and air inlet for water container (red). This last piece is required to let the pump work, without it the pressure in the container will go to a vacuum and water cannot come out to extinguish.

[1] https://www.conrad.nl/p/laagspanning-dompelpomp-barwig-0444-600-lh-6-m-539090

Attachment

The drone makes use of an ultrasound sensor, which is located at the center of the bottom, to determine its height. This limits our ability to attach anything to the bottom of the drone in the center. One option would be to attach two extinguishers on the bottom next to the center, as shown in the picture below.

This would increase extinguish capability but also the weight. We would have to test if the drone is capable of lifting over 500 grams.

Another option is to attach the extinguisher on the top of the drone, but far enough to the front that the suppressant will not be sucked into the propellers. The figure below shows how the blusstaaf could be attached to the top of the drone, using two attachment parts (blue). These parts can clamp around the cylindrical shape of the blusstaaf and can be attached to the drone using adhesive velcro tape, to ensure the extinguisher can be removed easily. The attachment part in the back also has space for the electronics that will be used to activate the blusstaaf. Placing the electronics in the back also helps balancing the weight of the drone.

The front part would be attached to the extension of the drone frame that holds the camera:

The attachment parts can be made using a 3D printer. The advantage of having 3D printed parts is that they can be very lightweight and can have custom shape which will fit our drone and extinguisher. If 3D printing is no option, wood could be used as an alternative. Wood can also be relatively lightweight.

Activation

The parrot AR drone, which we chose to use for the demo, is controlled with an app via wifi. We can’t however extend this connection with a separate channel for the extinguisher activation. Therefore, we will have to use a separate transmitter and receiver. The real drone will have this channel integrated into its main receiver, along with the other channels for control and communication. To activate the blusstaaf, a button has to be pressed which makes a connection to the battery (for any solid aerosol extinguishers with electrical ignition, a connection has to be made to a power source). A better way to control the activation is to replace the button with a relay which makes a connection when it receives a voltage. A remotely controlled relay can be bought [1] for just under 5 euros.

A schematic of a system like this is shown below.

Influence on drone

When the extinguisher is activated, the suppressant will naturally exert a force on the drone as it is ejected. It is important to know this force in order to conclude whether the drone will be affected by it significantly. The drone has to retain its ability to move freely in all directions while the extinguisher operates.

The reaction of the ejected matter is determined by the rate of mass flow and the velocity of the suppressant as it leaves the nozzle. The force can be found using the expression:

Using the fact that the rate of mass flow is equal to the density of the matter multiplied with the rate of volume flow:

and that the velocity of the matter right after the nozzle is equal to the rate of volume flow divided by the area of the nozzle:

the force equation can be rewritten to:

In order to calculate the reaction of the blusstaaf, the density of the ejected aerosol, the rate of volume flow and the area of the nozzle need to be specified. These values can't be measured for now. However, an estimation can be made to get an idea of the order of magnitude of the force. Upon seeking contact with the retailer of the Blusstaaf a material safety data sheet had been acquired. In this document, it was stated that the aerosol has a density of 50 g/m^3. Futhermore, the description of the Blusstaaf states that it contains 45 g of suppressant and that it can extinguish for 30 seconds. From this, it can be estimated that the rate of mass flow is equal to 1.5 g/s. The rate of volume flow would then be 0.03 m^3/s. Upon close inspection of images of the Blusstaaf, the area of the nozzle can be estimated to have an order of magnitude of 10 mm^2. Using these values, a reaction of 4.5 N can be found.

To determine whether the drone can overcome this force, the maximum force of the drone in the opposite direction of the suppressant stream needs to be know. Since the suppressant will be ejected horizontally, the maximum horizontal force has to be determined. This force can be estimated using the maximum lift capacity of the drone and its current weight. The Figure below shows a schematic of the forces that are generated by the drone.

The force Fmax will be equal to the maximum lift capacity multiplied with the gravitational acceleration:

The component Fy must be equal to the gravitational force acting on the drone in order to keep the current altitude. This force is determined by the current weight of the drone multiplied with the gravitational acceleration:

The maximum angle of the drone follows from the relation between Fmax and its component Fy:

However, this is angle is limited to 35 degrees for most drones, as they will become unstable at larger angles.

The horizontal force can then be found using the angle and Fmax:

The Dji Matrice 100 is a drone that fits all our requirements. Its maximum take-off weight is stated to be 3.6 kg. The drone itself weights 2.4 kg and a extinguish mechanism with a maximum weight of 0.6 kg will be added. Using these values, a maximum horizontal force of 19.5 N can be derived, when the tilt would be 34 degrees. This force is more than three times higher than the estimated reaction of the Blusstaaf. Therefore, it is concluded that the drone can still move in all directions while extinguishing.

It must be noted that this method is not completely accurate since the maximum lifting capacity might change when the drone is at an angle. Also, the direction of the suppressant flow will change as the drone tilts. However, an estimation can still be made to get an idea of the order of magnitude. Furthermore, the maximum take-off weight that is stated for a drone has to be a weight with which the drone can still move in all directions, since moving is an essential feature of a drone. Therefore, it can be assumed that the drone can generate an even larger horizontal force.

For the demonstration, a water extinguisher is used instead of a Blusstaaf. This extinguisher ejects water at a volumetric flow rate of 0.0052 L/s. The area of the nozzle is about 9 mm^2. Using these values, the density of water and the method that is described above, a reaction of 0.003 N is found. At this scale, it can be assumed that any drone that is capable of lifting an extinguishing mechanism is also capable of handling this reaction.

Drone

Requirements Drone

Before choices on how to design the drone can be made first the requirements for the drone must be formulated. In the table below the main requirements for the extinguish mechanism are given with a short comment and priority. The priority scale is 1 = drone won't work without it, 2 = important for drone to work properly, 3 = not critical to functioning of the drone.

Number	Requirement	Comments	Priority
RD 1	Drone must be able to carry 0.6 KG of weight.	Carry extinguishing mechanism and other equipment.	1
RD2	Drone must fit in a box with dimensions 0.6x0.6x0.3 meters.	The dimensions of the drone matter for the storage and usability of the drone, a drone that is to large won't fit trough the door	1
RD 3	Drone must have it's propellors shielded.	Prevent injury in collisions with users. Prevent damage to drone on impact with anything.	1
RD 4	Drone must be able to fly for at least 7 minutes on one charge.	Fly out, do task, return to charger place.	1
RD 5	Drone must be rechargeable	This means the drone is reusable after use, this very important since the definition of a robot is that it must be reusable [2]	2
RD 6	Drone must be able to connect itself to charging station.		3
RD 7	Drone must have Heat camera on board		1
RD 8	Drone must have RGB camera on board		1
RD 9	Drone must have protection against smoke with particle size >7 µm	7 µm is the size at which smoke is visible to the human eye.	3
RD 10	Electrical components of drone must be protected against dust/water according to at least IP65 rating	Prevent damage of low pressure streams of liquid to prevent damage during extinguishing. High resistance to dust as the drone might not be used for a long time so it might collect quite some dust.	2

Design choices

Some design choices around the drone:

Disposable or not The choice is made to not make a reusable drone. The first is that it is possible to use the drone after the initial extinguishing to either retrieve an extra round of extinguishing agent or keep flying around the area of the fire to monitor the current situation.

the second is that it is expensive to build/design a drone that is able to fly stable around a fire, this due to turbulence and lowered air density

Thirdly a very important design choice is that the drone should not have fully extinguishing the fire as a goal. The main goal of the drone is to slow the fire down and maybe fully extinguish it. An entire room is on fire 3 minutes after ignition, every minute you can achieve by delaying the fire is very useful. Human lives are always more important that just stuff.

Possible candidate

In search for a drone, it was important that the requirements would be met. The Dji Matrice 100 is a drone that is compatible with all requirements. It is able to carry more than a kilo, it can fly continuously with an added weight of 1 kg for 13 minutes and its dimensions are smaller than the box that is described in the requirements. Of course it also possible to build a drone from scratch but since there is a good drone available why don't use it.

Bill of Materials for (prototype) F.R.E.D.

Horizontal underwater pump tiny tronics. https://www.tinytronics.nl/shop/nl/robotica/motoren/motor/onderwaterpomp-horizontaal-3-6v?search=water

Tube for underwater pump. https://www.tinytronics.nl/shop/nl/robotica/toebehoren/waterslang-voor-onderwaterpomp-verticaal-horizontaal-3-6v-transparant-1-meter?search=water

433 MHz RF remote with receiver. https://www.tinytronics.nl/shop/nl/communicatie/rf(id)-wifi-bt/433mhz-rf-afstandsbediening-met-ontvanger?search=433

Demonstration and prototype

Since the project duration is only 8 weeks and there is a limited budget not all requirements or plans can be fully implemented in the prototype of F.R.E.D. In this chapter the prototype will be discussed and why it's different from the requirements for the actual drone.

Drone for prototype

We have borrowed a drone which can be controlled by hand. It is equipped with an ultrasound based altitude sensor and two cameras: one facing down and one facing forward. The drone can carry about 0.5 Kg for a really short time if and only if the battery is fully charged. A structure will be attached to it which can hold and use an extinguisher. The drone is not programmable unfortunately. There was an alternative drone available to use which could carry a bit more and is programmable, however it does not have out of the box control software present, so we would need to write code which controlls each propellor seperately and create our own stabilisation software and so on. The drone we use accepts instructions like "go forward", "rotate left" "lower altitude" ect. We think writing control software for the more advanced drone would take up (way) to much time, therefore we went for the simpler non programable drone.

Extinguisher for prototype

Since the drone available for the prototype is only capable of carrying 0.5 kg requirement REM 6 can't be fulfilled. Due to this limitation in weight it is also harder to fulfill requirement REM 2 for the prototype, the longer the prototype can extinguish the better but an exact time for the prototype is still unknown. Due to budget limitations requirement REM 3 can't be fulfilled so in the demonstration a mechanism based on water will be used. This means that the prototype drone is only capable of dealing with type A fires safely. All other requirements should be possible to realize in the prototype. However, the drone has sensors for height detection on it's belly and these sensors can't be blocked by the extinguish mechanism because this would strongly disable the flying capabilities of the drone (REM 7).

Extinguisher using waterpump

For the demonstration, a mechanism based on water will be used. Therefore, a small extinguisher using a waterpump has been made. This prototype consists of a bottle, a waterpump and a plastic tube. Since the pump that has been bought is an underwater pump, it will be placed inside the bottle. The power wires for the pump and the plastic tube will exit the bottle via a hole cut in the top of the bottle. The entire mechanism will be attached to the protective cover of the drone using double sided adhesive tape and calbe ties.

Activation

As discussed earlier, the extinguisher of the prototype will be activated with a seperate RF-channel. To this end, a small transmitter-receiver operating on 433MHz has been obtained. The receiver (which will be carried by the drone) has an operating-voltage of 3-6v, which corresponds with the small pump that is used. Therefore, the entire extinguisher will be powered by three AA batteries in series, adding up to 4.5v. Since the maximum power output of the receiver channel is too low for the power consumption of the pump, an NPN transistor has been used. This transistor will act as a switch for the pump, controlled by the receiver. This can be seen as the relay that was discussed in the previous chapter. A schematic of the electronic circuit of the extinguisher can be seen below.

Weight

The bottle and pump have a weight of 70 g and the receiver with the battery pack weigh 90 g, which gives a total of 160 g. If the bottle would be filled 2/3rd full, the water would add about 300g. The total weight would then be below 0.5 kg, which the drone should be capable of lifting. However, when testing the drone with the extinguisher attached, without water, the drone would not rise higher than 10 cm above the ground. Therefore, we will use a completely different drone which is capable of lifting even more than 0.5 kg.