We are almost constantly referencing autonomous vehicles. Some say we’re almost there. However, that concerns autonomy on the roads — what about autonomy in the skies? No, I don’t mean the autopilot that large airplanes have used for years. I mean full self flying.
Daedalean is an autonomous flight control software development company trying to make this a reality. Drones, helicopters, and eVTOL (electric vertical take off and landing) vehicles — those are the vehicles for which Daedalean is trying to develop full autonomy. CleanTechnica has had the privilege of interviewing them and witnessing their latest demonstration. (See the story linked above.)
Overall, this article aims to be everything you need to know (but probably didn’t yet know) about full self flying.
First of all, we should discuss all of the challenges this field is facing. When you think of airplanes, you are surely aware that they occasionally need to dodge other airplanes, which is done either through good planning by the control tower or via a super expensive device called TCAS (or CCAS) that is essentially a vehicle-to-vehicle communication system that decides which one needs to descend and which one needs to climb, and then announces it to the pilot loudly by saying “CLIMB, CLIMB” or “DESCEND, DESCEND” when the pilots have less than a minute to comply or face catastrophe. In today’s relatively empty skies, that is doable. In the much denser traffic that we would have with Amazon drones delivering shoes and such, a robotaxi delivering your drunk friend home at night, and a flock of geese making a kamikaze run right at you, you’re obviously going to need something much better than the status quo to ensure your safety. So, currently, the answer the experts give us is a whole cocktail of solutions.
Vehicle-to-Vehicle Communication & AI Control Towers
So, before we discuss current developments, let’s quickly talk about the thing we don’t have yet, for which there are as of yet no viable prototypes. What we really need for any of this to become a viable commercial solution are vehicle-to-vehicle communication systems and AI control towers.
For vehicle-to-vehicle communication, right now, airplanes only have TCAS which I already explained above. In the future, this will not be good enough. While to our knowledge no other system has been invented as of yet, there are a few interesting principles that this system will have to include. The most obvious is communications between vehicles to decide who flies where and at what speed in order to avoid crashes. What this will also include is meteorological conditions. Is there a thunderstorm ahead? Is it about to start raining? Imagine the fasten seatbelts sign going on automatically before the turbulence hits.
The next thing is identifying flying objects that do not transmit and signals like a flock of birds. The idea is that the cameras on these vehicles will be able to see the flock, avoid it, and then calculate their likely trajectory and warn everyone on a flight path that might intersect the flock. Every vehicle in the network can keep updating the current position and trajectory when spotted and be at heightened alert if visual contact with the flock has been lost. When the traffic becomes too dense and too complex, we really will need AI to manage the flow of traffic. Right now, there are no AI-based solutions for this problem only some prototypes that can “assist” air traffic controllers.
Vision-Based Safety Solution
The first line of defense is going to be something we have heard about before, a vision-based safety system. This is exactly what Daedalean is working on right now, and it’s impressive to see what the company has achieved. It is kind of like what Tesla is developing but it spots birds, small airplanes, and unregistered drones.
Currently, when TCAS senses a new player joining the game, it warns the pilots and the right procedure is to attempt to make eye contact with the other plane if possible. Now, obviously, this is often like trying to find a proverbial needle in a haystack, since it will be a tiny spec in the distance and pilots of commercial airplanes cannot directly see what is beneath them or above them. Also, TCAS costs between $15,000 and $150,000, so in its current form it is obviously unsuitable for the future of low-altitude air travel. In any case, we were talking about a vision-based system. In many ways it is similar, but in many ways it is very different from the vision-based Autopilot Tesla is working on. In the sky, if something is detected, the computer pretty much just has to ask “Is it a bird? Is it a plane? Or is it Superman?” Okay, throw in a few other possible options — like drone, helicopter, eVTOL — determine which it is, and react in a predefined way. In any case, the types of objects your autonomous flying vehicle can run into are much more limited than what you might encounter on the ground.
Currently, Daedalean is one of the industry-leading companies in autonomous flight. CleanTechnica recently visited the company’s HQ to learn more about how it has gotten this part of vision-based autonomous flight almost nailed. However, as it turns out, the true challenge isn’t flying — it’s landing. More precisely, the challenge is in determining where not to land. So, I guess we can call this full self landing? Autonomous landing? Let’s go with that second one to avoid confusion.
The irony is that this problem is practically the inverse of self-driving and self-flying, because we are not trying to determine how to avoid everything but instead how to not avoid one or a few suitable targets and then touch down on it as gently as a leaf kissing a pond.
As another comparison, while Tesla has focused on highways for years and is now switching to traffic lights and street traffic, the full self-flying industry is now focusing on vision-based autonomous landing instead of just flying. (To find out more about this and where the industry is currently, please make sure to look at our article about industry-leading company Daedalean.)
Suffice to say that the AI is getting pretty darn good at determining what can or can’t be landed on, and is constantly updating that as objects walk into a potential landing zone.
Where Daedalean is Now
This technology is actually very promising and is most likely going to improve upon what pilots in real life can do. It’s a movie classic to hear that “the vegetation is too dense” and then see that the protagonist will have to land some distance away from the target and then go the rest of the way on foot, or something like that. It is very likely that once a full self-flying vehicle becomes better than a human pilot, it can calculate landing zones very precisely and reach areas that a human pilot would not risk. The autopilot suite will also benefit from other sensors, like radar, that will even allow it to land when there is fog. Adding sensors like infrared and radar will open the door to a lot of new possibilities that are simply not possible for a human pilot.
Timeline of the Different Aspects of Full Self Flying
Almost everyone has at least once seen a headline something like,“Flying cars are just 5 years away — here, look at this amazing prototype this company made!” Actually, the 5-years-away timeline could even be seen in the news 10 years ago, yet to this day no such vehicle is here. However, unlike nuclear fusion, which is just 15 years away and always will be, the problem here is not an underestimate of the complexity of the science behind it, but rather a desperate scramble for funding. Here are the issues and how many years we might need to resolve them:
Without sufficient battery density, electric air propulsion will not be viable. Tesla’s vehicles for sale right now have batteries with roughly 250 Wh/kg. For flight to be viable, many experts say you need a minimum of 400 Wh/kg, something that Tesla’s/Maxwell’s theoretical limit of 500 Wh/kg might actually realize. This will likely require 5–10 years, whether you are being optimistic or pessimistic, and might also be dependent on the unicorn of solid-state batteries that many automakers and battery manufacturers are riding their hopes off a cliff for.
Everyone knows that for any kind of AI operation, you need hardware that is both powerful and not too power hungry. What you may not have considered, however, is that for flight, there is a very important third metric — weight. While a truck might be able to carry a whole server farm, an eVTOL vehicle, and especially a drone, can’t. Right now, the hardware is not powerful enough and light enough. (Another issue is that any new hardware solution needs to be certified for operation in avionics.)
With VTOL vehicles, unless you have full self flying autopilot technology, you will need a human pilot, and as many of you know, while a lot of people have driver’s licenses, relatively few people have pilot licenses. Thanks to our recent interview with Daedalean, we now know that vision-based flying and not hitting something in today’s low-altitude traffic (or lack thereof) will be perfected in the next few years. Flying in conditions that aren’t perfect or during the day might require some creative thinking and take some more time. It will require additional sensors, like radar, and infrared might be especially crucial at night. Optimistically, this could be accomplished within the next 5 years. The real hurdle is developing a vehicle-to-vehicle communication system and make it an industry standard that everyone uses. The other big hurdle is an AI control tower and enhanced vision-based software that can handle the high-density low-altitude traffic of the future.
Technically speaking, in Europe (for example) there is only one law that needs to be rewritten for fully autonomous robotaxis to become legal, and that is that right now a human must always be able to take control of the system and override the autopilot. In a robotaxi, a customer shouldn’t be able to take over control of the autopilot, because in most cases, said passenger does not know how to fly. More realistically, additional laws should be made for testing and certifying full self flying vehicles to ensure that they are safe.
Conclusion: Real Timeline of Final Robotaxi Product
Technically speaking, if cost was not a concern at all and we could use some of the technology that is now only available in satellites, and we didn’t have to certify the vehicle, the first autonomous eVTOL vehicle might go for sale in 2025 for an absurd price that only multi-millionaires could afford. Large corporations buying them and offering services using them is not likely to be possible before 2030. Anyone claiming otherwise is likely looking for funds and wants to hook investors.
Related: Daedalean & The AI That Knows Where Not To Land
I don't like paywalls. You don't like paywalls. Who likes paywalls? Here at CleanTechnica, we implemented a limited paywall for a while, but it always felt wrong — and it was always tough to decide what we should put behind there. In theory, your most exclusive and best content goes behind a paywall. But then fewer people read it! We just don't like paywalls, and so we've decided to ditch ours. Unfortunately, the media business is still a tough, cut-throat business with tiny margins. It's a never-ending Olympic challenge to stay above water or even perhaps — gasp — grow. So ...
Sign up for daily news updates from CleanTechnica on email. Or follow us on Google News!
Have a tip for CleanTechnica, want to advertise, or want to suggest a guest for our CleanTech Talk podcast? Contact us here.