It looks like you're having connection troubles. Reconnect now.
01/25/2022 22:12:57
30c3-talk-5406 Latest text of pad 30c3-talk-5406 Saved Jan 25, 2022 |
Hallo Du!
Bevor du loslegst den Talk zu transkribieren, sieh dir bitte noch einmal unseren Style Guide an: https://wiki.c3subtitles.de/de:styleguide. Solltest du Fragen haben, dann kannst du uns gerne direkt fragen oder unter https://webirc.hackint.org/#irc://hackint.org/#subtitles oder https://rocket.events.ccc.de/channel/subtitles erreichen.
Bitte vergiss nicht deinen Fortschritt im Fortschrittsbalken auf der Seite des Talks einzutragen.
Vielen Dank für dein Engagement!
Hey you!
Prior to transcribing, please look at your style guide: https://wiki.c3subtitles.de/en:styleguide. If you have some questions you can either ask us personally or write us at https://webirc.hackint.org/#irc://hackint.org/#subtitles or https://rocket.events.ccc.de/channel/subtitles .
Please don't forget to mark your progress in the progress bar at the talk's website.
Thank you very much for your commitment!
======================================================================
So good morning, and it's nice to see so many people at this early hour, so welcome everyone. I'm incredibly happy to be speaking with you. My name is President Muskie and I have been developing drones for quite a while now. I'm one of the paparazzi UAV system developers. I actively contributes software and hardware designs to the project, help maintain the websites and the server infrastructure. So I might be a little bit biased, but I will try to be as objective as I can. My company, one bit square performs open source micro UAV, consulting, development and manufacturing, I have been working in that field of autonomous unmanned aerial vehicles for almost seven years now, and I'm also the developer of Lisa and Lisa s Autopilot's that are part of the paparazzi platform. When we read about drones in the news, we perceive them as big, scary, expensive and very likely lethal. But this is not what we would be talking about today. We will be talking about this small, the affordable, the sometimes odd and sometimes also quite big. I am also only interested in drones that adhere to the same idea of freedom as freedom of speech. And some of those that everyone could build if they had a minute to spare, I know this is quite a problem for many of us. Also, I think a drone needs to be able to fly by itself as most at the most simple level. The only thing that one has to tell the aircraft is where to go and what route it should take to get there. And this flight instructions can be greatly more sophisticated than that now that we have set up some borders for the things that we will be talking about. Let's start with some history. I am pretty sure there are projects that are older than what I could find, but this is as far back as I could uncover. The project is simply called autopilot or it has some DIY thing in it too, and can still be found on SourceForge. The project was created in 2001, and it's still being worked on. That's the funny part. The Autopilot has an open source
stabilization and navigation system for helicopters, and they documented their first fully hands off of the remote control flights in June 2003. The next project was the development of paparazzi UAV. I know what you might think paparazzi is a strange name. But I just attribute it to the French humor. Around the time when the previously mentioned Autopilot project was performing their first autonomous flights, Pascal Bresee and Antoine DuPont from Inuk Local Nazionale de la Vision Seville in Toulouse in France decided to compete in the NCAA Super IRA autonomous flight competition. The original design used an off the shelf commercial FM copilots stabilization system, which was mounted on their twinjet remote controlled airplane. They were feeding control signals to it from a laptop on the ground that, in turn, was running the actual control instead of a human pilot. This can be rather unreliable set up. It is difficult to predict what the black box commercial hardware and software will do and how it will react to the computer generated commands. This resulted in unavoidable oscillations and end crashes. Antoine and Pascal developed a board very similar to the one it created for the Autopilot project and reusing some of their remote control parsing codes and algorithms, and also adding good deal of their own code. They were able to win the competition, what with that system? This inspired them to continue working on that project and because they wished to share and their ideas and collaborate with other enthusiasts, they released the software and hardware designs to the public or under GPL. In the years that followed, more teams used paparazzi system to compete in competitions and win a lot of prizes involving fixed wing aircraft and autonomous flight. Around the same time, when paparazzi was started to smart, German high school students developed a rarely before seen flying object for and. You can forecast, is a national science compared to competition for high school
students, which recognizes and rewards inventions and innovations. You might recognize the device they invented since it's quite ubiquitous by now. Definitely here. Based on their invention, the company Silver Lit started producing a Quadro copter toy called X UFO. I'm departing here from the open source premise, but stay with me because the innovation is an important one. As you might know, a Quadro copter is inherently unstable. That means it will fall out of the sky just like that if you don't feed signals to the motors. The rule of thumb is that it needs at least 100 hertz getting updates to the motors so that it stays afloat. A human is only able to deliver at most 10 updates per second. If they are really well trained, thus the platform needs some kind of a mechanical or electronic stabilization system to make it playable by a human. The X UFO was price optimized as a toy. It used a mechanical gyroscope with whole sensor pickups. This is quite a mind blowing, considering it was cheaper to build and manufacture such a device than it was to purchase an off the shelf piezoelectric or micro mechanical sensor. The mechanical gyroscope had many issues. It suffered from progression and as well as limited error angle before it started hitting the brackets, resulting in the so-called flip of death. These shortcomings and the fact that the rest of the UFO electronics were quite simple, created quite a big community of people that were modifying and improving the electronics and sensors for the platform. The exciting development and knowledge exchange that was happening in the UFO forum was mostly German speaking, but the forum provided a great meeting place for all the UAV community at that time. Sadly, the forum was disabled without a warning earlier this year, but there are efforts to get it back online. Most of the modifications that were based on the use of two piezoelectric gyroscopes and a lot of hot glue. But even though these modified versions did not, did not
look pretty. They flew better than the original since they did not suffer from progression or the flip of death. No, no. Nonetheless, the sensors were difficult to get in single units. The piezo electric sensors suffered from temperature drift and mounting them perfectly perpendicular to the airframe was quite a tricky, quite a trick. Thanks to those two sensors, it was possible to stabilize the most unstable two axis, namely the if you say the front is here to pitch and the roll axis. So these are the two that this would stabilize. This is when the inventors of the platform, Daniel Gordon and Klaus Michel Dot, came back into the picture together with two of their college friends, Young Stone and Michel acrylic. They found it ascending technologies. They produced their 3D puzzle gyroscope board that was a drop in replacement for the mechanical gyroscope. That was quite an important part and quite an improvement. The 3D board added one more axis of measurement, making your so the vertical rotation all stabilized to. While it was a commercial product, the development inspired a few other people to work on their own incarnations of quadcopter electronics and solutions. Among them was micro copter. Micro copter was mentioned in there for the first time in a UFO forum post in October 2006. Many of you might know of them already. They provided access to their source code and had quite a well documented circuitry. This was great, even though they were not technically licensed, licensed as open source, but something completely proprietary with very questionable status. The Micro Copter Control Board came with seven degrees of measurement. Adding a three axis accelerometer, so a linear motion in all three axis that measured the Earth's gravitational acceleration, providing the absolute down vector. This allowed for the platform to stay horizontal with a smaller amount of drift. This feature allowed one to release the remote control stakes a bit before the quattro copter drif
ted into something horizontally. It also came with a barometer providing an absolute altitude information and making it possible for the vehicle to remain at an set altitude by itself. Another platform that I would like to mention was the Welfare, a board also known as the unmanned aerial video platform you AVP project. As far as I know, it was open since it was possible to access the source code. I am not precisely sure about the exact history of the project and whether it was actually inspired by the UFO, but at the time during which they were developed are very close. The platform used more sophisticated macro mechanical sensors. This made it quite a bit more expensive, but it also much more. It was much more stable in flight. This stability made it much better as a video camera platform out of and out of that particular project you, AVP Nji emerged to composed of a new group of people and its own separate philosophy. This is and this new group mainly targeted Quadro copters high quality sensors, and it was meant for those who wanted to assemble the electronics themselves. They regularly released a stable source code under open source licenses. And as far as I can tell, their goal was a very high level of software stability and best sensors available on the market and an advanced feature set. But since I'm here and the creator of this project is likely sitting here in the audience, he can tell you the details of that history better himself. They have an assembly here, so you should go and visit them. I bet they will be very happy to answer all of your questions. In the midst of all that development, a three axis magnetometer was added, resulting in 10 degrees of measurement and absolute information about the heading of the aircraft. The addition and the addition to the platforms that I already mentioned there, there were many more other platforms to come to our mock up to our copilot with varying levels of openness and availability of the hardware. Most of the pr
oject originated here in Germany, and it took quite a while before it finally caught on in the US. For example, Jordi Munoz started working on his own iyamu for Arduino that later became the pilot. Around January 2008, this over time resulted in thriving commercial companies like 3D robotics. As most of the solutions already contained a controller of some kind and a lot of sensors, it was an obvious next step to add a GPS and start teaching quadcopters to fly by themselves. So this development was happening from both directions. The autonomy focused competition and research driven projects like paparazzi from one side and hobby multi copter driven projects from the other, the paparazzi group developed its own controller board dedicated to quadcopters, called Vous, together with its algorithmic flight plans and flexible ground station software originally developed for fixed wing airplanes. It started to morph into an airborne robotics framework rather than a pure airplane autopilot. Paparazzi was adopted by many universities, as well as a bunch of commercial companies. The universities and companies help to improve the platform and add features for the benefit of everyone involved. Thanks to the people involved, too, we are able to push the boundaries of what we can do with civilian drones every day. Other projects that came from the quadcopter and hobby side of things started adding Position Holt and then Waypoint Flight plans to their existing stabilization, hardware and software. So where are we now? Almost all open source autopilot systems of today, like our two pilot, open pilot and paparazzi. Support fixed wing airplanes in many configurations, a single engine, twin engine flaps, a layer on elevators, flaperon 11s and so on. Multi copter not only quadcopters by a laundry list of different configurations similar to airplanes, but even more so. Three four eight 12 motors in X plus and over and under configurations. It is quite overwhelming how many possibilities
this approach provides. Basically, attach a few motors with with propellers to a bunch of struts, and you will probably be able to make it fly. Helicopters started the whole open source controller story, but to be honest, since the dawn of the multi copters, they do not seem to be the most common choice for conversion into a drone. The main reason is the cost of operation. The helicopter has a lot of very intricate mechanical parts that are expensive to manufacture. And if your software screws up and you crash, the fund becomes quite expensive very quickly. They are still being used in situations where you need a long flight time, it is easier to put together a turbine or a combustion engine driven helicopter than a multi copter. Recently, though, very cheap, very small helicopters have become available from China, making the helicopter an interesting platform again. More on that later. A new category of vehicles called transitioning vehicles is emerging. This is a combination between a fixed wing aircraft and a multi rotor transitioning vehicle could cause take off vertically can hover and on a flip of a switch or a software comment can transition into a forward flight, making it quite more efficient. And this can be also achieved by tilting the engines themselves and not the whole vehicle, as far as I know, paparazzi is the only platform so far that supports this new aircraft class in the vanilla codebase base, including autonomous flight, but other projects seems to seem to be catching up slowly. Many projects used atmo or pick microcontrollers, eight bit architecture and 16 megahertz clock, or, even worse, Arduino. But now most of the projects are using ARM 30 32 bit processors that provide enough processing power to do some fun stuff without the programmers having to cope to decode extortionists. This enables us to think about adding real time OS for managing processes, for example, and encryption to the telemetry links run more sophisticated estimation for mor
e accurate control or run intelligent path planning algorithms. The sensors have come quite a long way as well for the last few years with the thanks to the ongoing development of cell phone technology. The prices of micro mechanical sensors have dropped significantly so that they have become affordable. They might not be the greatest, but they are good enough for most applications. And instead of paying $30 per degree of measurement, we pay six dollars for six degrees of measurement, or $13 for nine degrees of measurement measurement in a single chip. Even though there are many different boards at the court, they have converged quite a bit. What they differ in most is the size and what integrated features they have, like radio communications, a video overlay circuitry server connections and communication channels to sensors and actuators. Here's a sample of a project listed here is only an excerpt out of paparazzi supported autopilot comparison table. As you can see, there are plenty to choose from. On the software side of things, the feature that even the simplest systems with a GPS support provide is position and altitude hold, sometimes also called loiter for hovering vehicles, this feature is quite obvious the aircraft stays on a spot at a designated altitude in case of an airplane. This means flying circles around a set spot. A simple extension to that feature is homing when powering on the aircraft, remembers the coordinates as its home position during the flight. At any time, it is possible to flip a switch or press a button that will make the aircraft fly straight back to that spot. It is a great safety feature and an aircraft saver for beginners, as well as experienced pilots. For Follow Me, an operator is carrying a GPS receiver, and the aircraft then follows the position of the GPS receiver like a virtual leash. I have the feeling this feature is not fully explored yet. I can't imagine using such a feature to convert the aircraft into my personal cameram
an. When flying an aircraft equipped with a gimbal camera, it is possible to tell the autopilot to point the camera at a specific point on the ground. This feature is used very often by reporters and filmmakers. Waypoint navigation is, in my opinion, the minimal feature that needs to be provided before a project can claim that they support autonomous flight. This means it is you don't steer the aircraft yourself using a remote control. Instead, you mark GPS coordinates on a map and the aircraft can navigate between those points. An extension of the waypoint navigation our survey pattern flights. In this case, the operator makes an area, marks an area on the map and tells the aircraft to fly a pattern that will cover the whole area. That is a very useful that is very useful when you are generating maps, surveying crop fields, searching or studying tree canopy in the rainforest, for example. Almost all systems support some kind of geofencing. This means that if the aircraft leaves a virtually fenced off area, it is possible to perform some kind of an emergency maneuver depending on the regulations or emission requirements. It can then try to fly back home, switch off the engine and drop to the ground or stay put and gain altitude instead of continuing the mission. There is a lot of more mission actions that almost all mature project support by now. But these are mostly static flight plans and the conditions change. And if the conditions change, one has to have connection to the aircraft to perform its in-flight alterations to the mission plan. And improvement on that is using dynamic flight plans. One possible dynamic flight plan solution is implemented in the paparazzi platform. The first class citizen of a flight plan is not a waypoint, but a block. Each such block, consisting of multiple multiple stages, represents an action. The actions can be setting system variables, altering waypoints, executing autopilot functions or performing a jump to other blocks based on
the system variables. This variables represent either sensors or counters. This solution represents a dynamic, domain specific language in which one can implement a dynamic flight plan. Each block is statically compiled see routine on the grounds as part of the Autopilot firmware, so the flight plan implementation is safe from accidental change in flight, but still can react based on conditions providing flexibility without compromising code size, restrictions and safety. A very similar system is implemented in the glue pilot, too. Here, I need to give them some props because they are saying on their website that it is exactly the way paparazzi did it. The open pilot team is working on another solution that they are planning to release in the future. It uses a Python interpreter running on board the aircraft that can manipulate the waypoints and probably also the flight vector. But we will see how that works when that feature is documented and released to the public. There are many more things that we already can do, but this is how far we've got in the past 10 years. Now let's try to predict a little bit into the future. Or rather, let me tell you what I hope you will be working on next. Let us start with a great example of what it's working very well so far. Paparazzi Sumo is, for example, that I have here on the stage with me. It's the aircraft here. It is it has been used for quite a while now by metrological research groups as a reusable weather balloon with basic training, the researchers that never flew in RC airplane before before it can take the soon launch it and gather humidity and temperature and wind information in our columns, thanks to the efforts of the people like Martin Miller. It is. It has become a standard tool that is as important as about balloons themselves. Based on that, I believe we will see more drones used as sensor platforms. For that. I hope we will see more people like Martin Miller, who will make sure the solutions are standardized a
nd deliver very repetitive results. On the other side of the spectrum. I was shocked to hear that a simple seeming competition, the outback challenge is still unclaimed. The mission consists of taking off from a runway flying to a field that is 9.3 kilometers away. Searching as 7.5 square kilometer area for a bright yellow dummy and then dropping a half liter bottle of water. It's a bit bigger than this to the designated position. This sounds like a very simple job, but since the first competition in 2009, the main prize is still unclaimed. Teams were getting very, very close, but not close enough to claim the main prize. I have been told that the main reasons why teams fail is because the rules and requirements of the mission are overly demanding. So let's hope that the big prize will be claimed this year. But this has been made clear that the civil use drones are still very much in their infancy. Drone technology, as well as the regulations regarding them, are in need of advancement in order to mature. There is still a lot of pioneering work to be done. One striking development is miniaturization. There are plenty of very cheap and whatever quality toys in this area, but I believe this will be a starting point for an interesting burst of innovation, just like UFO did during its time. There are already a few of the first really cool open source projects emerging, for example, Crazy Fly by Bit Grace. It comes with all the basic 10 degrees of measurement sensors and their own custom radio that you can connect to your laptop. The API is very well documented and the source code is released under open source license. This makes it the device very compelling to anyone interested in experimenting with drones from the comfort of their chair. People, people are already doing very interesting things with it, like controlling it, using Microsoft Kinect or adding sensors, so it bounces off the walls without touching them or adding their own 3D print and custom parts to the air
frame. I think we will be seeing very cool stuff emerging based on this project. Other project that I am closely involved in is the development of Liza s in collaboration with the Technical University in Delft, the Liza, as is just the brain electronics. So just the pure autopilot of a UAV, I designed it to be as small as possible, making the size of a medium postage stamp about two point two by two centimeters. It contains the 10 degrees of measurement sensors, a powerful arm microcontroller with half a megabyte of flash, as well as a GPS receiver module while weighing only two point five grams. I have it here on the stage with me, so maybe it will be even smaller. Yeah. So. Just like Crazy Fly. I also developed a custom radio. The main reason was to save some more space and wait. The difference between super bitter and crazy fly solution is the compatibility with the Spectrum RC control system. This solutions allows us to have telemetry with the laptop, computer or tablet and standard model aircraft controller at the same time using only one module on the aircraft. Normally, these two channels are implemented using two separate radio systems. The initial motivation was an auto pilot for a low cost micro helicopter airframe as a retrofit solution, providing additional lead to this standard tandem. Also a GPS receiver module, making it a perfect fully autonomous for fully autonomous flights. And because it is just a pure brain part of the system, it makes it easy to mount on other airframe types do like Quadro copters or airplanes like this tiny things from China, for example. So you can just retrofitted with the with the board, and you should be able to fly autonomously with it. As I mentioned, we mounted aGPS on the laser as making it useful for outdoor applications. It is quite still not quite accurate as we would like, but thanks to companies like Swift Navigation, we will be getting a sub 1000 AHTC GPS systems that will provide a centimeter accuracy at 50 hertz
. This is quite an improvement from current GPS solutions that provide best case accuracy of three meters at 10 hertz. This is a significant step forward for outdoor applications, for example, from small separation swarms where you fly the aircraft really close to each other and you can draw things in in air with them, for example, or fully autonomous aerobatic flight, because this module also supports really high accelerations. What you have big problems with commercial modules or this is also a commercial module, but the modules that you can currently buy on the market or precision farming mapping and so on. But what about indoor applications? Sadly, indoor applications autonomous flight is still very, very challenging. And there's there are some solutions available, mostly for research groups that with really, really deep pockets. For example, infrared camera tracking systems. These are the same systems used to capture the motion of actors for 3-D movies like Gollum four from Hobbit, for example. They're great. Very accurate and fast thanks to the solutions like this. Groups like Grass at Pennsylvania and University and idiocy at E.T.A. Sir, it can create incredibly amazing demos. I bet you have seen their videos of swarm flights snapped, optimize flight planning or ball juggling Commodore copters. The problem is that an infrared tracking system like this is a very expensive and because of this out of reach or even the very rich hobbyists or less wealthy universities. Beside the cost, one has to dedicate a room for the micro UAV in order to mount the cameras in the room at very accurate positions. You can't just go with your drone anywhere and perform autonomous flight. So I'm throwing it out there. We need to work on some solutions that we will have the following feature set high accuracy. Relatively cheap. No environment modification needed. Light and very small. Low energy and computational requirements, there are many approaches to solving this problem. Maybe
laser based radar, maybe computer vision, who knows? There are some very promising technologies on the horizon. One of them comes from a company called Technical Illusions. They are developing a new kind of virtual reality 3D glasses. The thing in the middle of the glasses is a tiny cell phone camera connected to an FPGA. They also will make a metal gate tariff, a version of it to bring down the cost in PCs, as well as a need of power. It can track a blinking infrared led pattern at 100 frames per second, providing a submillimetre, accurate position and rotation of the user's head. I am looking forward of mounting one of those on a nano drone. Another solution comes from two Delft. It was released just a few days ago. It is a tiny stereoscopic vision module from four. They are micro flapping wing platform called Delphi. The system is based on the parallax effect of binocular vision, just like we humans and every animal with front facing eye does using two cameras, this module can also see depth not only for a flat image. It is not a totally new concept, and what it is is the size. As you can see, it is pretty small and thus light. It can detect obstacles. And according to Mavela, it is also quite affordable. It enables to Dell fly to and navigate autonomously. Definitely. The videos look very promising, so here is like a time lapse recording of the position of the aircraft, how it is flying in a room completely with no human control input. I think we are almost out of time, so. So let me let me leave you with a thought in the world of PC software. We know about continuous integration, coding standards, static code tests and so on. It is really hard to embed for embedded systems and thus for all drone types. I know many projects out there do software compilation tests and static code analysis on their software. But something I know we should work on is better reliability testing and flights for flight software we put on our toys. I know this will not happen overnigh
t, but it is something we should make a priority of. I know this is not as much fun as flying them or developing new algorithms, but I would be. It would be a huge benefit not only in this field, but other embedded system areas to be able to do better regression and reliability testing. I know of the fact that there are some approaches provided as part of commercial development packages. What we need, though, are open source solutions for automatic regression and unit testing tools for embedded systems. If anyone here would like to volunteer any solutions, that's why I'm telling you that I would really love to hear from you. Thank you for your time and I think we will be going to questions now. I guess. As always, we have four mikes, two there, two there, just line up. Also IAC and Twitter. We already have a question there. Go ahead, please. OK, so all the numbers survey pit floods pretty useful for producing aerial photos, for example, for open street maps where you can draw a map based on these photos. Are there right now, economically affordable solutions for that? And can you say something about the legal complications there? So affordable solutions? The most affordable solution is definitely to build it yourself from some artsy parts a that you just buy off the shelf and then put an autopilot on it from any of those projects, like paparazzi. In this case, you can buy the electronics and put them in. I'm pretty sure this is more affordable than buying an off the shelf, fully integrated and solution the the the so basically that it's there. So this sumo project, for example, is one of those that are built by by a by a group of people that are using an Open-Source platform for this. And there are other solutions. I think our drone, the company, the parrots, they bought something there. I don't know how affordable they are because I'm completely uninterested in buying anything from the shelf anyways. So I don't know. I rather build it myself anyways and to exactly
suited the application. And it definitely is fully doable with the paparazzi platform, and I'm pretty sure that others have some solutions for that too. The legal implications this is this is a this is difficult to say. It's like I'm not a lawyer, and that's why I'm not talking about the legal stuff at all because I know what I may do, where I may fly, and I am happy that I grasped that and getting into like, what are the regulations? You have to really look locally where you are, what right, what rights you have, who you have talked to talk to. It's like still everything in the infancy. As I said in the talk, it's we are still working on the regulation side of things, as you everyone here might know. But but it's like we as a community from the open source side, we have to really, really push hard so that the regulations make sense because what is seems to be coming is really horrible. And but I'm speaking to the choir here, right? OK, thank you. Number three, please. I was also asking, I'm going to ask about the legal implications in regards to arrest and may go to or altitudes you can fly. Maybe there's a certain space where you can fly under a certain altitude. Do you know at least where we can research that sort of stuff? What resources do you use to find out about that? Yeah. Well, it's like Google for it. But yeah, no. It's like in there is stick here in Germany, there is a club for RC airplane fliers. This is basically the rule so that you will apply to most, most of people that are trying to do this. The problem is exactly the regulation part of things is the problem that as soon as it is not a hobby anymore, there's not really any regulations available. And if they are available and people think that they know what the regulations should be, it is. No, you may not and please don't compete with the big guys because we have the money. So basically, that's what it ends up being for the altitude part of things. It's I think in Germany, it's like 500 meters alt
itude maximum grant or an off the ground. So AGL, not not for altitude. So if you are in the mountains above 500 meters, you still may fly 500 meters above the mountain. But about this is only far away from an airport. As far as I know, if you're near and near an airport, you, I think, are very restricted to 100 meters or something, but it's similar rules apply. Also in the U.S., it's like 400 feet maximum altitude. And so if you are below five, um, am I right, five kilos, then you may fly also outside of a designated field. Yeah. So thanks. So we offer Amazon lawyers to sort it sort of things out for us and we may fly to. Yeah, that wasn't a nice gag. What they did there. Yeah, it's like No. Four, please. Hi. If I understood you correctly, you said you're only working on free software platforms and. I was wondering if you think about introducing nonmilitary classes to those licenses. Well, if you add them, you are not Open-Source anymore, right, because you are excluding a group, so you may not discriminate anyone if you want to be really open source. So adding a nonmilitary rule to it is against that rule. Thus, you are not really open source. So this is my take on this. And yes, I would love to block the military. But on the other hand, they have much better choice anyway, so it doesn't matter. One more from the internet, please. OK, how secure are the solutions you presented for controlling a drone against the though? So how easy is it to take control over a drone that my, yeah, my pal is controlling right now? So this is what I mentioned. Also in the talk, it is a problem currently. It's like it's completely non encrypted. It's like, basically we are at this stage, like the internet started out a practically is still it's like all unencrypted stuff running everywhere around. So like you run, there was a great example from the hack five guys. Just again few days ago, it was a great video where Darren Kitchen took like an IRA drone. And I think a DJI other quadco
pter put a pineapple on it, flew to the drone and overtook its Wi-Fi and controlled it from there. So it's it's obviously bad we have to work on this now. But yeah, that's it's really necessary to do. It's like, we need to secure this. OK, thank you. Number two, please. Yeah. Commercial aircraft use a radio beacon system even even today to navigate between waypoints over pretty much the entire globe. And this is without this was done before GPS. Yeah, have it since many institutions, buildings, etc. have a Wi-Fi network or some other industrial radio network. Has there been any research done in using these, like individual APIs or individual radios for for navigation? I'm thinking of like, you know, following a Mac address to a Mac address. Yeah. So there were some efforts. I saw some, some some universities mentioning that they were working on things like that for if for a fact, I could take the words. Also recently, some news that Apple was deploying like position tracking of iPhones to track their customers in their stores with quite high accuracy. So theoretically, that technology would be there to use Wi-Fi hotspots to as basically radio beacons. I don't know of any go to solution. That's the problem. It's like there are some DIY things here and there with varying degree of accuracy and reliability and documentation. But it's like, I don't know if I think this is also one of the possible approaches, but I'm just looking forward to see if someone comes up with some really nice solution where it's like, Oh, this works. Yeah, because this seems like you could be a highly accurate because they're they're these tend to be fixed position, you know, your IP networking staff, how are you going to know exactly where they are? So you should be able to construct a route? Yeah, but they're being already used, for example, for augmenting your cell phone GPS position accuracy today. So. And like Google and Apple, I think they're mapping also, though. So yeah, it's a I'm look
ing forward to see some something that is not a huge brick that I could put on on one of my quadcopter send testers. I'm curious to see that. Thanks, internet. Just a short, short one. Are you? Do you know about the challenge about that? Excellent. Are they allowed to refuel during the mission? As far as I know, I think, yeah, there is something you may come back after you scan the field back home so that they can drop off pictures, find it so that you don't really need to solve everything, including the finding out anonymously. So I I bet you can refuel in that stage when you come back, but you still can reach you. I don't think you can go to the field several times and refuel between passes, so you have to do go there and scan the whole field in one go. I think this is what it is, but don't hold that to me. So. OK, thanks. Mike, one, please. What are the one or two main areas where commercial systems are more advanced than open source solutions? And why is it is the time, the money, the smarts? That's a difficult one. So, yeah, definitely there are commercial systems that do some things better or different than the open source solutions. One thing is there is the obvious one that you have with a lot of open source projects in general. They are much more tested, I think, or they are like, basically, they are not adding so much stuff so quickly so that this system is is getting a little bit unstable at times. So they are like, OK, very conservatively, we are the future. We test it truthfully and then we consider maybe we should release it. At least some companies do, like not all of them. Like, I know that 3D robotics is releasing hardware that is like half tested and then they have to do revision after revision. And not even, you don't know as a customer when you buy it that it's something changed, that the previous one was broken. But anyways, maybe they are doing some communication telling the customers, but this is what I heard. But then there are commercial com
panies that are like, fully commercial and closed that really do. Also, it's more robust. I would say the solution this is from the feature set from the feature set. There are more like the mechanical solutions, for example. I think the there are some some solutions to have, like a hook on the wing so you can land on the boat very easily. So you have a string hanging from a crane on the boat and you can hook it in. But I am pretty sure we could just build it. Also, go on a C and see cut apart and do that to ourselves, Stu. But no one I don't know of anyone trying that themselves. Otherwise, from the features again, from the feature set, I don't know it probably enough about the commercial systems to say what they really have. Maybe I should really actually do that and see what we need more to concentrate more in open source. But what we have to concentrate more is definitely reliability and rip it repetitively anyway. So basically thinks like this sumo. That's why I showed Sumo, because this is Martin Miller with his guys really put the energy to make this thing fly every single time and work every time. This is like a rare thing in their whole open source world where it's like, This is the solutions. This is how you build it and you have a tool like a hammer. Internet. How critical do you see the fact that the UAV challenge you mentioned is platinum sponsored by Northrop Grumman, which is a military cooperation? Yeah. So from the moral standpoint, I think it's it's crap, but from the other standpoints, they have too much money anyway. So we should take it. OK. So your your point is still that the military already has better stuff and. Well, they they have better stuff that costs way too much money. This is why they're interested in our stuff because it's much cheaper and we choose solutions that they wouldn't even come up with because it doesn't have 50 certificates on it, basically. OK. Mike, one. OK, I've heard as a branch of science that investigates real animal
s like birds and flies and to take inspiration from that to build machines. Does it play a role for your research? Yeah, yeah, definitely. It's like. So the first quadcopter ever was built by imager. It was before even the helicopter that we know today was even conceived. It was a quadcopter, the first vertical takeoff vehicle at all, and he was inspired by by dragonflies and their flight. So the propellers that he used on this were already back then inspired by by birds. And definitely the further you go like, for example, they fly the flapping wing aircraft, it's of course it's biomimicry, right? So you are trying to copy copy the animals. OK, internet, go ahead. What kind of scenarios can we expect in the future, apart from the current positioning, acceleration and orientation sensors? Do you have any ideas on which sensors additional sensors? So yeah, that's that's basically the the problem here. It's like for the EC, we have everything to do autonomous flight very well outdoors. It's like on the airplanes you have one additional sensor is that Peter, too. So to get airspeed, so you can do control on that, but this is also a normal part of the control system on an airplane. But on the road aircraft's sake we are putting, I think we have everything to do with autonomous flight by now. But what will come are other sensors that make it more, maybe more accurate to say you don't like to keep the altitude more correct at ground level. There are ultrasound sensors being used, but it really also depends on the mission and the and the environment that you're running and don't facing cameras just like our airdrome does for optical flow. This is something they are using. I'm into. Yeah, pick the guys picks, OK? They have optical force sensor that they are putting on their stuff. It's a it's it's definitely coming to, but it really the minimum set, in my opinion, to do. Autonomous flight is the tandem plus GPS solution. So and then you have basically everything you will yo
u will get definitely other sensors, depending on the mission and what you want to do. The indoor navigation, though, yes, the indoor navigation, there will be something will come. I really hope so. That's why I hope that's why I'm talking to you guys, because the smarts are we have to put our heads all together because it's so interdisciplinary. This is also the thing here with the drones. It's like it's machining, it's electronics, it's software. It's it's our dynamics. It's everything, it's like it's not and not a area where you can just concentrate only on your part what you know about and and you will you won't be able to operate. You have to talk to so many people in many disciplines. And that's why I think I really hope that someone will come up with really great solutions that no one thought about before because they are just thinking out of the box. Thanks, Mike. One, please. OK. Do you have an overview of the race integration rapid operation system in open source platform? Yeah. So Ross Ross integration is is ongoing. I am, I know that, but there is like there is some adapter for modeling, I think for the autopilot stuff. And paparazzi has also some adapters for us. It has a converter from the Ivy Ivy bus that it is using on the ground station. How good is this working? Uh, as far as I know, it works, so I didn't try it myself, so I can tell you, but I. Hmm. Yeah. But this is not an open source platform, as tech has as a commercial platform has a Ross integration. But yeah, this is something I would love to see. Like someone who is actually really using Ross Tech I use draws in different research in the past myself. It's it's a great platform, so I would love to see it properly integrated with paparazzi, for example. As far as I know, as I said, it has an adapter. People were using it. I don't know how well it works, though, because most of the missions and most of the stuff that is being done with this platform normally doesn't require to run with Ross. A
dditionally, it is for definitely for multi-platform operation where you have to communicate with different with rovers and robots and humanoid robots. You would need that, yes. But again, I don't know of anyone who is involved in this project to be actually involved in this. You're welcome. Thanks, Mike. Three, Please. When you're talking about regression tests, what do we really want to test and where would you like to start? So as I'm mostly a low level embedded person, it's like I would love to find a way I say can treat really thinking hard how to do it. But it's like I didn't come up with a reasonable solution is to be able to even be able to test the drivers on the lower level. It's like with the IO, with the hardware connected, there's this really hard problem because you probably have to build some custom hardware to be able to run those tests with feedback loops and like switches and whatever. People are doing some things along those lines. But this is something I would start with because it's this is what I do. But other things is on the very high level. There is some simulation stuff. This is an interesting thing. I know people are working on this, this will come, but it's basically that you have a simulation running. You basically take the newest revision from the code base and you fly an autonomous flight plan in simulation. And if it deviates from certain parameters, then it will start screaming like something broke. And this is something completely the opposite side and everything in between needs to be tested to. And so basically, that's why I'm saying it's like, I know there are people that are software engineers that are like, Oh, a drone Quadro copter. I don't know. It's like, it's too much, too much soldiering and too stinky. But there is there are real software engineering problems in that in that area that that we need the help of those people that are really have their head deep in software to solve this. So you would like to decouple the sen
sors from the PC B and inject simulated simulators or areas. It's like, yes, there is software in the loop. It already exists, but it is not an automated test system. OK. That's what I mean. It's like, yes, you can do it on your laptop. You can run it in simulation. No problem, but it's like, how do you integrate it in a platform that runs on a server somewhere, a client and just just goes through this every day when there are new comments? OK? Internet, please. Okay. What batteries have the best power density and are available to the public? Who? That's like another home, and we don't need to manufacture another, but just the technology a. As far as I know, it is still lithium polymer, but I might be wrong. That changed by now. I know that leafy pool or is the abbreviation have the advantage that you can charge them faster. But I I'm pretty sure they don't have the density of capacity. So I think lithium polymer is still the best. But tell me if I'm wrong. OK, thanks. I would like to know myself better game theory. Something better than this last question. Number four, please. For indoor navigation. Would it be feasible to build something like a GPS? But I assume frequencies. I bet it would. But it's and my RF knowledge is not good enough. So. So building such a thing is over my capabilities, basically. And that's yeah, it's I definitely I think it is a solution. I saw something there. There is also a group that made a prototype. They built some antennas and on poles, put them in four corners of the room or three corners of the room. And then they had that receiver and basically they measured the time of flight. You can do it much better where you measured the carrier phase, right? It's like, then you have it much more accurate, just like you do it with GPS. So but there is again, there is no there are some very, very, very expensive commercial systems that simulate satellites indoors. So GPS satellites. I think this exists, but it is it is completely beyond any go
od or bad in price. This is like this is not even the Richard universities can afford that. I think so. Yeah, but if this would be, this is definitely a direction that would work nicely if someone would build a platform that like that. OK, and we are out of time, please, once again, thank Estan.
