-
Want to remove ads? Register an account and login to see fewer ads, and become a Supporting Member to remove almost all ads.
EVNow
Well-Known Member
Since no one has implemented Level 5 yet, I think at this time its all smoke and mirrors. Charitably, it is the best guess of Tesla & NVidia.
Reminds me of when my team bought expensive SAN with a lot of extra capacity - only to retire it in a few years and getting cheaper & better cloud storage.
Reminds me of when my team bought expensive SAN with a lot of extra capacity - only to retire it in a few years and getting cheaper & better cloud storage.
To those trying to draw analogy to the problem of automating aircraft, I'd like to make a few comments from the perspective of a general aviation pilot and engineer in the aerospace industry. First of all, if you exclude takeoff, commercial airliners can already automatically fly you to where you are going and land the plane, and it happens on a regular basis.
The problem from a control and procedures standpoint is extremely easy compared to autonomous cars. All routes and approach paths are planned in advance and guaranteed by your national aviation authority to be free of terrain and obstacles. Air traffic control guarantees that you are separated from other vehicles in the air and on the ground. Having this kind of consistent environment means that the only technology required is precise navigation (e.g. augmented GPS or high-integrity ground-based navigation beacons) and aircraft control algorithms, all of which are well-understood and proven technology. (See also: Instrument landing system - Wikipedia and PID controller - Wikipedia) No computer vision necessary. The situation is the exact opposite for autonomous cars.
The real problem for aviation is guaranteeing integrity of the system. Aviation is extremely risk-averse, to the point where infinitesimal risks of failure are sometimes not tolerated. To give you an idea of how little risk is tolerated, consider the FAA's WAAS, its GPS augmentation system. It enables instrument approaches (i.e. no vision) down to two hundred feet above a runway before the pilot must maneuver and land visually. The risk tolerance for the system outputting hazardously misleading information (HMI) on an approach is 2x10^-7 per approach, which is two bad-information approaches out of ten million, and even getting bad information on an approach will not necessarily lead to a crash. (See also: WAAS Performances - Navipedia) Keep in mind that this risk tolerance is for planes that have a trained human pilot in the loop who would be hand-flying the approach or monitoring the situation with autopilot. What would the tolerance be if no pilot were present? How much redundancy and system safety self-checks would be required to maintain integrity? How many flight hours would be needed to demonstrate that the system is actually safe?
The takeaway is that you can divide autonomous system development into multiple distinct problem areas:
For autonomous vehicles, I'm not too privy to the details, but my impression is that #2 is mostly if not completely solved in many R&D vehicles (over all companies, not necessarily restricted to or including Tesla). Looking at some of the Mobileye academic talks, it sounds like there is a lot of work done on #1. There is quite a lot of mathematical framework developed for autonomous tactical planning and driving policy. #3 feels very sticky to me. A lot of tactical planning relies on computer vision based on neural networks. To my non-expert knowledge, they are not very well-understood and there exist many simple tricks that can cause them to incorrectly identify objects (e.g. adding noise to an image). What would it take to prove the integrity of such a system? Given that the operational environment of driving is much more dynamic than that of aviation, and that the integrity problem is hard enough for aviation already, this seems nearly insurmountable.
Just wanted to share some scattered thoughts. Thanks for reading.
The problem from a control and procedures standpoint is extremely easy compared to autonomous cars. All routes and approach paths are planned in advance and guaranteed by your national aviation authority to be free of terrain and obstacles. Air traffic control guarantees that you are separated from other vehicles in the air and on the ground. Having this kind of consistent environment means that the only technology required is precise navigation (e.g. augmented GPS or high-integrity ground-based navigation beacons) and aircraft control algorithms, all of which are well-understood and proven technology. (See also: Instrument landing system - Wikipedia and PID controller - Wikipedia) No computer vision necessary. The situation is the exact opposite for autonomous cars.
The real problem for aviation is guaranteeing integrity of the system. Aviation is extremely risk-averse, to the point where infinitesimal risks of failure are sometimes not tolerated. To give you an idea of how little risk is tolerated, consider the FAA's WAAS, its GPS augmentation system. It enables instrument approaches (i.e. no vision) down to two hundred feet above a runway before the pilot must maneuver and land visually. The risk tolerance for the system outputting hazardously misleading information (HMI) on an approach is 2x10^-7 per approach, which is two bad-information approaches out of ten million, and even getting bad information on an approach will not necessarily lead to a crash. (See also: WAAS Performances - Navipedia) Keep in mind that this risk tolerance is for planes that have a trained human pilot in the loop who would be hand-flying the approach or monitoring the situation with autopilot. What would the tolerance be if no pilot were present? How much redundancy and system safety self-checks would be required to maintain integrity? How many flight hours would be needed to demonstrate that the system is actually safe?
The takeaway is that you can divide autonomous system development into multiple distinct problem areas:
- Route planning / procedures
- Control
- Integrity
For autonomous vehicles, I'm not too privy to the details, but my impression is that #2 is mostly if not completely solved in many R&D vehicles (over all companies, not necessarily restricted to or including Tesla). Looking at some of the Mobileye academic talks, it sounds like there is a lot of work done on #1. There is quite a lot of mathematical framework developed for autonomous tactical planning and driving policy. #3 feels very sticky to me. A lot of tactical planning relies on computer vision based on neural networks. To my non-expert knowledge, they are not very well-understood and there exist many simple tricks that can cause them to incorrectly identify objects (e.g. adding noise to an image). What would it take to prove the integrity of such a system? Given that the operational environment of driving is much more dynamic than that of aviation, and that the integrity problem is hard enough for aviation already, this seems nearly insurmountable.
Just wanted to share some scattered thoughts. Thanks for reading.
ElecFan
Member
Thx BinaryField. Having been a commercial pilot myself in the distant past, I've wanted to write such a contribution myself. You've done a great job.
As for the content: What about Cat IIIc operations which allow an aircraft to operate in zero visibility? Isn't this in fact autonomous operation? (be it that at present traffic control has the last saying in the actual operation (when, where)).
I think that, if FSD is going to be safe enough in the near future, we need to create separate infrastructure for these operations. Special lanes for FSD-cars, FSD-tunnels and/or 'skylanes' etc.. If these car can communicate with each other, they can travel at high speeds close behind each other which would greatly reduce both travel times and energy consumption. All non-fsd cars (and failing fsd cars) can use the 'normal' infrastructure. As time progresses there could be less and less non-fsd cars and more and more infrastructure could be made for fsd-cars. Integrity of the system can be improved in this way, as I see it.
As for the content: What about Cat IIIc operations which allow an aircraft to operate in zero visibility? Isn't this in fact autonomous operation? (be it that at present traffic control has the last saying in the actual operation (when, where)).
I think that, if FSD is going to be safe enough in the near future, we need to create separate infrastructure for these operations. Special lanes for FSD-cars, FSD-tunnels and/or 'skylanes' etc.. If these car can communicate with each other, they can travel at high speeds close behind each other which would greatly reduce both travel times and energy consumption. All non-fsd cars (and failing fsd cars) can use the 'normal' infrastructure. As time progresses there could be less and less non-fsd cars and more and more infrastructure could be made for fsd-cars. Integrity of the system can be improved in this way, as I see it.
Much appreciated!
Category III-style approaches were what I had in mind with my top sentence in bold. You know more about this than I do, but I think I recall that pilots need to have special training to initiate such approaches, which I take to mean that in the engineering risk analysis, there is integrity credit taken for having a pilot monitoring the state of the aircraft. (Or at least it should be, to avoid unnecessarily smothering system performance.) My only point was that if we put high safety assurance aside, we could basically have autonomous aircraft today, as you pointed out. This shows that the hardest part of the autonomy problem is not necessarily vehicle control or object recognition, contrary to what many people have expressed.
Your idea makes a lot of sense to me, since it emulates the controlled and coordinated environment of aviation. I guess the only downside is that it requires new infrastructure to be built, as opposed to attempting to train cars to operate in the current environment. I wonder which way is harder in the long run. My money is that the latter is harder, even if it is more appealing.
ElecFan
Member
Thanks, BinaryField.
I guess it will not be infrastructure OR cars, rather the one AND the other. Infrastructure will not have to be built, as it can be modified. F.e. in Holland we have sometimes separate express lanes on highways, which can easily be transformed to FSD-lanes. I guess rail-infrastructure can be modified to allow coupled fsd-cars to operate there and we can use the metronetworks in cities to allow fsd-cars to be transported in and out of citycentres. There is so much infrastructure already, it can be transformed to allow for fsd-cars to operate there.
At the same time cars will become increasingly 'smart' in self driving and (like aircrafts) will be able to operate on their own. Like aircraft pilots today, drivers will need to monitor the systems and intervene whenever the system fails. Emergency buttons are always present in cockpits of modern airliners and this should also be the case in cars, however smart these systems may be.
In that way we use the best of both worlds. Computers for 'normal' operations, humans for extraordinary circumstances that computers cannot deal with.
I guess it will not be infrastructure OR cars, rather the one AND the other. Infrastructure will not have to be built, as it can be modified. F.e. in Holland we have sometimes separate express lanes on highways, which can easily be transformed to FSD-lanes. I guess rail-infrastructure can be modified to allow coupled fsd-cars to operate there and we can use the metronetworks in cities to allow fsd-cars to be transported in and out of citycentres. There is so much infrastructure already, it can be transformed to allow for fsd-cars to operate there.
At the same time cars will become increasingly 'smart' in self driving and (like aircrafts) will be able to operate on their own. Like aircraft pilots today, drivers will need to monitor the systems and intervene whenever the system fails. Emergency buttons are always present in cockpits of modern airliners and this should also be the case in cars, however smart these systems may be.
In that way we use the best of both worlds. Computers for 'normal' operations, humans for extraordinary circumstances that computers cannot deal with.
Similar threads
