If all the autonomous vehicle makers pooled their vehicles, they would still never get remotely close to L4
RAND stated that it would take 500 billion miles to create an autonomous vehicle 10X better than a human. Toyota has stated one TRILLION miles would be needed. If I use extremely conservative math, vehicle, fuel, driver, no engineering costs and an average of 50mph, it would take 234,000 vehicles driven 24X7 for 10 years to get those miles. At a cost of over $300B. Even if you go with RAND’s numbers that is 117,000 vehicles driven 24X7 for 10 years at a cost of over $150B. It can’t be done. (Not only due to these issues but safety issues that cannot be resolved. See more in my link below.) But let’s say it could be done. Right now, if all of the folks making these systems pooled their vehicles together, they wouldn’t be remotely close. (Yes, simulation is being used to offset some miles. However public shadow driving is still the primary platform and the simulation systems being used in this industry have significant performance problems.) I read that Apple just got to 70 cars. Audi has 12. Waymo ordered 500 vans. Let’s say they use all of them. And let’s say every AV maker had 100 vehicles. At about 50 AV makers that is 5000 vehicles. 4% of what is needed best case. And again, that is if they had that many and if they used them all together, no redundant scenarios. Even if they ALL had 500 vehicles each that is 25,000 total. Only 21% of what is needed best case.
The solution is to replace 99.99% of that public shadow driving with aerospace/DOD/FAA level simulation and systems engineering practices
More here
The Crash of the Autonomous Vehicle Industry
