All the Autonomous Vehicle makers combined would not get remotely close to L4
Today I see VW switched from Aurora to Argo. That switch will not matter in the least. As a matter of fact, you could literally combine every Autonomous Vehicle maker together and you will still never get remotely close to a legitimate L4. The use of public shadow and safety driving as well as gaming engines for simulation are that debilitating.
Public Shadow and Safety Driving — There is no way to mitigate the amount of time/work/miles (500B per RAND — One Trillion per Toyota) by driving and redriving scenarios you would have to stumble and restumble on. Spend the associated amount of money (+$300B). Or survive the casualties that will be created when handover is involved in time critical scenarios or when thousands of accident scenarios are run thousands of times each to train the ML.
Gaming Engine based Simulation — With regard to gaming engines. The real-time and model fidelity issues those systems cause especially in complex scenarios will cause differences between what the Planning system believes will occur and what will actually occur in analogous real-world scenarios. That will provide significant false confidence and lead to tragedies.
The Solution — The remedy is to use aerospace/DoD/FAA simulation technology and systems/safety engineering. Including augmenting that bottoms up Agile approach with a parallel top down approach. We also have a built in RL Engine automatic training capability built into the simulation. I will be glad to provide a demonstration this is the right approach. (Note- When I mention this approach folks usually think I am referring to air travel which is not nearly as complex as what is needed here. What is as complex and even more so is simulated DoD urban war games. Same locations, same complexity and the public experiences plus the ground vehicles can drive off the roads and shoot at each other.)
Update 6–12–2019 — Far worse than I stated
Over 1,400 self-driving vehicles are now in testing by 80+ companies across the US
Let’s look at the math on this compared to RAND’s estimate that it would take 500B to get to 10X a human and Toyota’s 1 trillion miles. At 40MPH driven 24X7 how long would it take to get to L4?
RAND 500B miles = 1019 Years
Toyota’s One Trillion miles = 2038 Years
1/10 of RAND’s estimated miles = 102 years
Please find more information in my articles below.
Using the Real World is better than Proper Simulation for Autonomous Vehicle Development — NONSENSE
SAE Autonomous Vehicle Engineering Magazine-End Public Shadow Driving
Common Misconceptions about Aerospace/DoD/FAA Simulation for Autonomous Vehicles
The Hype of Geofencing for Autonomous Vehicles
The Autonomous Vehicle Podcast — Featured Guest https://www.autonomousvehiclespodcast.com/
My name is Michael DeKort — I am a former system engineer, engineering and program manager for Lockheed Martin. I worked in aircraft simulation, the software engineering manager for all of NORAD, the Aegis Weapon System, and on C4ISR for DHS.
I am a member of the SAE On-Road Autonomous Driving Validation & Verification Task Force and was recently asked by SAE to lead an effort to establish a new Modeling and Simulation group.
I am a stakeholder for UL4600 — Creating AV Safety Guidelines.
I have also been presented the IEEE Barus Ethics Award and am on the IEEE Artificial Intelligence & Autonomous Systems Policy Committee (AI&ASPC)
My company is Dactle