I’ve been involved in both automotive and aerospace. I did the device drivers & “BSP” for the Boeing 747 flight management computer. I’m currently working on next gen passenger vehicle architectures.
There are vast differences between the two. As pointed out, the use-cases are very different, as well as the economics. So are the regulatory environments.
In systems requiring functional safety at the highest levels, software has high standards - you must be able to show that every execution path in the code has been tested. Avionics systems are really far less complex, hence the ability to generally fly themselves. For passenger vehicles, to successfully navigate the complex environments and drive, at least, as well as a human, all kinds of exotic software must be used. Neural networks, massive parallel processing, complex operating systems, all mount up to be 10’s of millions of lines of code. You can show that it works but you must also be accountable for HOW it is developed and tested. It’s too complex. The standard practices break (economically).
I’ve seen many autonomous vehicle start-ups (along with OEM programs). Most of the people involved are R&D types. They’ve no idea what “functional safety” is respective to software. Many executives have been shocked, you have a working prototype only to be told that the development must be done again from the ground up - to be done *properly*...all these exotic architectures need certification, which isn’t happening (cost).
I appreciate the comments, I understand each and don’t necessarily disagree. Autonomous vehicles are inevitable, at some point - but it’s going to have to take another path. I hope by getting things to the point where cars are just really hard to crash.
For those that don’t want software in their car doing anything, I can only say good luck - but the industry has become better. The Toyota Prius runaway throttle cost them billions. There’s some really good YouTube videos on the matter for those interested, all kinds of lessons learned.
I have a take on this that might seem completely out of left field, but it's based on years of experience working on the "highway" side of motor vehicle operations. I truly believe that one of the biggest obstacles in place for the adoption of automated vehicles is that the whole concept is based on the wrong fundamental principle of how the system works.
I started my career in civil engineering back in the 1990s when "Intelligent Vehicle-Highway Systems" (IVHS) first became part of the jargon of people involved in planning and designing transportation infrastructure. If anyone remambers those days, there was a fundamental issue that had to be addressed before any serious progress could be made in making a great leap in vehicle technologies that would eventually culminate in the development of self-driving cars.
The two competing paradigms at the time were: (1) Vehicle-to-Vehicle (V2V) technology, and (2) Vehicle-to-Infrastructure (V2I) technology.
V2V technology is what the automotive industry is developing today, where cars and trucks instrumented with all kinds of sensors and transmitters "talk" to each other and help guide themselves safely on a highway network. The vehicles drive themselves.
V2I technology involved instrumented vehicles operating on a roadway system built with roadside sensors and transmitters that guided the vehicles. The road and associated hardware drive the vehicles.
Very early on in this process, V2I was largely abandoned because it was determined that V2I infrastructure was far too expensive for state highway departments to adopt across entire road networks. V2V won the day almost out of necessity because the automotive industry was far more adept at testing and implementing technology than government agencies were. And it helped that people who buy cars would still retain a lot of control over their operation and aesthetics.
I believe we should be going back and thinking about this all over again ... because V2I will ultimately work better, and will eliminate many of the legal and technological obstacles that are holding V2V development and implementation back.