What does the handling of the Jeremy Banner’s Tesla tragedy tell us about NHTSA and the NTSB?
It has been about 6 weeks since Jeremy Banner died in a Tesla that appeared to be on Autopilot. In a set of circumstances that seemed to mirror the Joshua Brown tragedy from almost 3 years ago. The road pattern was similar, a tractor trailer was performing a common and legal lane crossing and Mr. Bannon’s Tesla ran right under it killing him. How does it take more than a couple days let alone 6 weeks to confirm if the vehicle was in AP mode? Why isn’t there a moratorium when a death occurs due to an autonomous vehicle, especially when the scenarios are similar, and they are ALL using humans as Guinea pigs to develop and test these systems?
After the Brown crash the NTSB stated that Tesla’s, or any AV for that matter, should be able to handle common occurrences in that area. It appears the Tesla did not do that in the Brown or Bannon scenarios. Beyond that it appears Tesla’s cannot handle avoiding many stationary objects or objects that are several feet above the road. They have hit several barriers and tractor trailers. Killing people in each scenario. If these vehicles are supposed to handle common occurrences shouldn’t they have to prove they can handle them in simulation and on test tracks before they graduate to using us as human Guinea pigs?
Looking at what has occurred over the past 3 years it appears the DoT, NHTSA and the NTSB believe that public shadow/safety driving is the best or only way to develop and test these systems. And the deaths that occur are necessary evils. Except, apparently, if they are children and only when they are in a vehicle covered by the Federal Motor Carrier Safety Administration guidelines. Does this mean they are prohibited from doing more by law or that they have to protect those children due to a law? Do they realize it is impossible to drive the one trillion miles or spend over $300B to stumble and restumble on all the scenarios necessary to get to L4/5? And that the processes are unavoidably unsafe and will harm thousands? That it is impossible, no matter what monitoring and notification system is used, to provide time to regain proper situational awareness in critical scenarios? Or that at some point the AV makers will have to experience thousands of accident scenarios thousands of times over to train their systems on how to handle them? And most importantly. if aerospace/DoD level simulation technology is used, informed and validated by real world data including test tracks, that 99.9% of the public shadow/safety driving is not needed? Or do they know all of this and are afraid to act or have some other agenda or motivation?
Earlier this year NHTSA did act and determined EasyMile’s school shuttle in the Babcock Ranch community in Florida was unsafe. In their report they stated, “Innovation must not come at the risk of public safety,” said deputy NHTSA administrator Heidi King in a statement. “Using a non-compliant test vehicle to transport children is irresponsible, inappropriate and in direct violation of the terms of Transdev’s approved test project.” While that is a step in the right direction what about when the exact same children get in any other autonomous vehicle under development and test? A Waymo, Uber, Lyft, Tesla etc. Virtually every AV maker and OEM is using public shadow/safety driving to develop and test these vehicles. A process that uses us a Guinea pigs both in and outside the vehicles. In June 2018 an ADS Conference was held outside of Washington, DC. Robert Molloy, Director of Highway Safety, the lead over all accident investigations, acknowledged there was a period of time in critical scenarios where no matter what approach is used people cannot regain proper situational awareness. If that is the case, why are we in the same place almost 3 years after Joshua Brown dies witnessing what appears to be an exact duplicate scenario?
A key to the answer regarding the agenda of these organizations may lie in two overlapping areas. A study NHTSA performed in 2015 on the safety of L2/L3 (handover) and where several key former NHTSA officials are and their point of view. Mark Rosekind at Zoox, Nat Beuse at Uber and David Strickland at the Self-Driving Coalition.
In 2015, under Mark Rosekind, NHTSA published a L3 Handover Study that stated this practice could be made safe using monitoring and alert systems. They also stated they tested it to prove it worked well. The problem is their test was extremely flawed and misleading. They determined full control was regained, after being distracted, by simply grabbing the wheel and looking straight ahead. They never attempted to look at the quality of the action taken or if proper situational awareness time could be regained. NASA, Missy Cummings and several AV makers, OEMs and leaders in the space like Waymo, Ford, Volvo and Chris Urmson have said handover or L3 is dangerous and should be skipped. (This despite ALL of them still using the practice. The reasons for that follows.)
Mark Rosekind, Nat Beuce and David Strickland all work for organizations or companies that support, defend and enable the practice of public shadow/safety driving. Their reputations, egos and livelihoods are now heavily invested in this process. This means they have personal and financial motivation to put human’s lives at risk to develop this technology. Mark Rosekind has stated so himself in an article last year that defend this practice saying the lives lost now will avoid more losses later. Nat Beuce helped Uber determine its development and testing after the Elaine Herzberg tragedy. Uber still utilizes a mostly public shadow/safety driving approach. And Mr. Strickland works for an organization representing most of the autonomous vehicles makers and OEMs using the process. The reckless, counter-productive and unethical irony being that these folks and the industry are doing the exact opposite of their stated goals. Since L4 can never be remotely attained using public shadow/safety driving they will not safe the lives that level would save. And they are taking lives needlessly in the untenable and perpetual process to get there. (I have sent most of this data to the current NHTSA administrator Heidi King with no response.)
Finally, there is a much better way to do this. One that takes the impossible to possible and resolves the problems I mentioned. The solution is to use aerospace/DoD/FAA level simulation, safety and engineering practices. (Not the simulation or Agile approach being used by the industry now. The former has too many real-time and model fidelity issues and the latter leaves the hardest scenarios hanging out until the end.) At that very same event last year that Robert Molloy acknowledged the unsolvable issues with handover he stated most of the development and testing to create these systems should be done in simulation or on test tracks prior to using public streets. Again, if that is the case, why are we in the same place almost 3 years after Joshua Brown dies witnessing what appears to be an exact duplicate scenario? (In addition to the primary use of simulation there should be a simple progression. You must prove why you cannot do something in simulation. Then why you cannot do it on a track. And finally, when you are allowed on the road, which would be less than 1% of the it is done in a structured manner. Not a free for all.)
Please see more details in my articles below
SAE Autonomous Vehicle Engineering Magazine — End Public Shadow Driving
The Hype of Geofencing for Autonomous Vehicles
Common Misconceptions about Aerospace/DoD/FAA Simulation for Autonomous Vehicles