Uber, NHTSA, NTSB and the Industry are responsible for Elaine Herzberg’s Death not Rafaela Vasquez
Elaine Herzberg, the State of Arizona, the public and yes, Rafaela Vasquez, are a victims in Uber tragedy
Rafaela Vasquez and Elaine Herzberg nor anyone else involved in the Uber or Tesla tragedies to date should have been part of these events because they should never have occurred. Uber, and most autonomous vehicle makers are using a development and testing process to create these systems, which relies on public “safety drivers” to train the machine learning. The problem is this process will literally never result in anything close to a true autonomous vehicle. This means that the Uber and the industry is doing the exact opposite of that which it says it is doing. It will never save most of the lives it eventually wants to save and take thousands of lives needlessly in a fatally flawed effort trying. Uber, NHTSA and the NTSB should all know or admit this. The reasons for that are:
· It is impossible to drive and redrive, stumble and restumble on all the miles and scenarios needed to train and/or test the system. RAND stated it was impossible with their calculations extending to 500B miles at 10X better than a human. Toyota stated it would be one trillion miles.
· It is impossible to spend the money to drive those miles. It would be a minimum of $300B.
· The “Safety Driving” process relies on actually having accidents over and over in many cases. If the “safety driver” disengages the testing process will not gather the data it needs because the threads are never experienced by the system. This means thousands of people like Elaine Herzberg, Joshua Brown, Jeremy Banner, Yoshihiro Umeda, Hao Yaning and Walter Huang must eventually occur. (The latter are all in Tesla’s.) The public, press, government, attorneys etc will never allow that process to continue when it is realized that thousands of accident scenarios have to be run thousands of times each, which would result in thousands of casualties, the process will be stopped. (I believe when the first child or family is killed it will be stopped).
· The “Safety Driving” process also relies on something called handover. That process involves the vehicle passing control back to the human to take control when not cannot handle a scenario. Which is often given these systems are in development. Often than handover is either unexpected by the human or the time needed to reacquire proper situational awareness cannot be provided by the AV no matter what monitoring and control system it uses. That time is between 6 and 45 seconds depending on the scenario. NASA, Missy Cummings, a plethora of studies and more recently Waymo, Ford, Chris Urmson and quite hypocritically Volvo, have said the process is dangerous and needs to be skipped. Ford reported their professional drivers fell asleep. NASA has a plethora of data on how ineffective this process is in those situations and has recommended to the AV industry they not use it. At an event in DC 2 years ago I pressed the NTSB Director of Highway Safety Robert Molloy, who leads crash investigations on this. He publicly stated there is a window of time proper situational awareness cannot be provided and that most of the activity should be done in simulation. All of this demonstrates it is unreasonable and often impossible for public “safety drivers” to react in time and react properly. Ms. Vasquez and the world around her are being misled by the industry into believing the process being used is safe and effective and the best or only one that can be used and that the people who die as a result are for the greater good. That is an absolute myth. It is unreasonable and grossly negligent to expect anyone to shadow drive in most complex, critical and dangerous situations. As such the process should not exist. Given that the tragedies involving Herzberg, Brown, Huang, Yaning and Umeda tragedies should never have happened. While Rafaela clearly did not obey company policy by watching video, and should receive some form of punishment, she should not receive a murder charge.
Finally, this is ALL resolved by using proper simulation, informed and validated by the real-world. Where much less shadow driving and virtually no “safety driving” is required. “proper” simulation being what aerospace and DoD use. Not the gaming-based systems the industry uses now.
Note — Several massive and negligent issues on the Uber side that should have stopped the car or provided more assistance to Rafaela.
· Defeated Volvo’s emergency braking
· Was confused over identifying a person or a bike. Why not stop for both or because it was confused by two major objects? Or alert driver? Radar would have not known what the object was. Why not defer to it?
· Where was the driver head or eye tracking and an alarm?
· Uber pulled the second “safety driver”
· Did Rafaela know the risks of handover? Or the actual condition of the car?
· What message did Uber send when they removed the second “safety driver”? That the system was much safer?
More details in my other articles here
The Autonomous Vehicle Industry can be Saved by doing the Opposite of what is being done now
Autonomous Vehicle Industry’s Self-Inflicted and Avoidable Collapse — Ongoing Update
Autonomous Vehicles Need to Have Accidents to Develop this Technology
SAE Autonomous Vehicle Engineering Magazine — Simulation’s Next Generation
Using the Real World is better than Proper Simulation for AV Development — NONSENSE
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.
Key Industry Participation
- Founder SAE On-Road Autonomous Driving Simulation Task Force
- Member SAE ORAD Verification and Validation Task Force
- Stakeholder for UL4600 — Creating AV Safety Guidelines
- Member of the IEEE Artificial Intelligence & Autonomous Systems Policy Committee (AI&ASPC)
- Presented the IEEE Barus Ethics Award for Post 9/11 Efforts
My company is Dactle
We are building an aerospace/DoD/FAA level D, full L4/5 simulation-based testing and AI system with an end-state scenario matrix to address several of the critical issues in the AV/OEM industry I mentioned in my articles below. This includes replacing 99.9% of public shadow and safety driving. As well as dealing with significant real-time, model fidelity and loading/scaling issues caused by using gaming engines and other architectures. (Issues Unity will confirm. We are now working together. We are also working with UAV companies). If not remedied these issues will lead to false confidence and performance differences between what the Plan believes will happen and what actually happens. If someone would like to see a demo or discuss this further please let me know.