NHTSA brings Missy Cummings on board — That will likely mean very little in the end
This week NHTSA announced Missy Cummings was appointed to a key autonomous vehicle advisor role.
That appointment is complicated. This is because Missy is right about some of the engineering involved and advocates for some of the right positions. However, far more often she is incorrect and advocates for the wrong positions, especially involving safety. She joins the rest of the duplicitous industry in scapegoating Tesla from glass houses. They are Splitting hairs in order to stop Tesla from harming and killing more people at a faster rate than they do so they can harm or kill less people later. They are duplicitous, disingenuous, hypocritical, and dangerous. Not as dangerous as Tesla, but nowhere near high ground. In an effort to make my POV on this issue easier to understand I will communicate my POV is a list format.
On the Positive Side
· Critical of machine learning, especially general and deep learning, specifically regarding these areas not being anywhere near mature enough for these domains. Too much confusion, guessing, mistakes etc. Basically, there is no thinking or inference yet. Only recall and execution replay.
· Critical of Tesla’s driver monitoring, especially alarm duration and ability to be defeated and overall maneuvering performance.
· Critical of Tesla’s use of Tesla customer Guinea pigs
On the Negative Side
· Supporting the use of human Guinea pigs when they are trained employees. And being just fine with the public being used as Guinea pigs in all cases.
· Ignoring issues with every other AV maker. This includes similar AEB issues due to poor fidelity radar that leads to stationary/crossing object issues.
· Ownership of stock and a seat on the board with LiDAR maker Veoneer. (I cover my own conflict of interest below.) Missy should have never gone down this road.
· Not pushing for imaging radar and only the addition of LiDAR to Tesla’s system. Right now, the only way these companies get to L4 on public roads in all human drivable weather conditions is to use cameras, LiDAR and imaging radar together. (This assumes the LIDARs are used to perform object classification and creation of tracks. Not used only for discerning location. Currently not many LiDAR companies or users are doing this.)
· Incorrectly contributing Tesla’s AEB issues to software vs hardware. I believe the stationary/crossing issues here are due to poor sensor input to the perception system. Well before the code Missy is referring to is used. I believe this is supported by Tesla’s eliminating the radar they had. While it had poor fidelity discerning the lateral position of objects, which led to the tracks being ignored to avoid false breaking, there are many conditions even a poor radar performs better than a camera. Tesla should have been able to find those sweet spots using Kalman filtering. I believe they couldn’t because their hardware has issues with accepting any other sensor inputs, especially if tracks are the primary input.
· Stating on Twitter she would punch Elon Musk in the face. This is just unnecessary.
· Her refusal to educate herself on aerospace/DoD simulation technology and support its broader use. I tried to make Missy aware of this technology. She refused to have the discussion. This is willful reckless and willful ignorance, especially given she most likely trained and qualified in simulators with the technology. When you loudly trumpet you are an expert in safety you have an obligation to do your due diligence. (I should also note she and the industry, including DOT and NHTSA are dead silent on US DOT VOICES. A group that agrees with the problem and solution statements I put forward.)
Regarding where I believe NHTSA will go, I believe NHTSA could go several ways:
1. They do nothing
2. They force Tesla to change only the “Autopilot” and “Full Self Driving” names and handling flashing emergency lights. This will leave make the systems safer but nowhere near as safe as they should or could be. Emergency vehicles will still be hit when the flashing lights are off or hard to detect since every stationary object is still at risk. It will also leave the use of Human Guinea pigs in place.
3. They force Tesla to fix all the names, flashing emergency light detection, the stationary/crossing object/AEB flaws and driver monitoring fidelity and alarm times over 20 seconds. This will result in a larger safety improvement regarding Tesla but leave the fundamental use of human Guinea pigs in place for the industry. While Tesla is the most egregious there will still me many people needlessly harmed or killed through the industry. This option would put the rest of the industry at risk with regard to stationary/crossing objects as many have the same problem. Albeit not as bad as Tesla because they use radar. Though like Tesla it is a poor one and struggles with lateral object location issues. Which lead to the objects being ignored to avoid false breaking.
4. They force Tesla to fix its particular issues and the entire industry to fix its stationary and crossing object issues as well as to move to proper simulation for most of the development and testing. Thus, nearly eliminating the need for human Guinea pigs. This would also save the industry from going bankrupt and give it the ability to actually get to L4. And to do so in our lifetime. (Two key parts of the solution are the use of simulation, specifically from aerospace and DoD, and its part in a due diligence engineering progression. Companies should have to prove why use of the real-world is needed, especially for “safety driving”. Amazingly USDOT’s own VOICES group supports this approach. Secretary Buttigieg, are you aware of that?)
Which one do I believe NHTSA will choose? I believe they, like Missy and most of the industry, will go with #2 to avoid touching Tesla’s competitors, ruffling the echo chamber’s feathers, admitting they were wrong the entire time and possibly derailing the ego boosting gravy train.
(My conflict of interest — I created a company called Dactle to try to provide the simulation technology required to get to L4. As I understood going this route would cause a conflict, I tried to avoid it by offering to help the existing simulation companies enhance or fix their tech. Particularly with regard to real-time and model physics fidelity. They refused. With some saying they would do so when their customers figured out there were issues and paid for the changes. That was unacceptable since that would likely only come after several crashes and deaths. (More on this in the article on the failed Iranian hostage rescue mission in 1980 and how it led to the simulation technology needed here below.)
More detail here. Including how to do this right.
Tesla “autopilot” development effort needs to be stopped and people held accountable
NTSB Chair falsely states “We have done all we can do” with regard to Tesla’s “Autopilot” debacle
NHTSA should impose an immediate “Autopilot” moratorium and report initial investigation findings in 30 days
The Deadly Industry, NHTSA and NTSB “Autopilot” Name Change Diversion
The Autonomous Vehicle Industry can be Saved by doing the Opposite of what is being done now to create this technology
SAE Autonomous Vehicle Engineering Magazine — Simulation’s Next Generation (featuring Dactle)
How the failed Iranian hostage rescue in 1980 can save the Autonomous Vehicle industry
USDOT introduces VOICES Proof of Concept for Autonomous Vehicle Industry-A Paradigm Shift?
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.
Industry Participation — Air and Ground
- Founder SAE On-Road Autonomous Driving Simulation Task Force
- Member SAE ORAD Verification and Validation Task Force
- Member UNECE WP.29 SG2 Virtual Testing
- Stakeholder USDOT VOICES (Virtual Open Innovation Collaborative Environment for Safety)
- Member SAE G-34 / EUROCAE WG-114 Artificial Intelligence in Aviation
- Member Teleoperation Consortium
- Member CIVATAglobal — Civic Air Transport Association
- Stakeholder for UL4600 — Creating AV Safety Guidelines
- Member of the IEEE Artificial Intelligence & Autonomous Systems Policy Committee
- Presented the IEEE Barus Ethics Award for Post 9/11 DoD/DHS Whistleblowing Efforts