Autonomous Safety...Is It Worth It?
In 2018, we predicted 2019 to be the most challenging year for autonomous vehicle safety. Highly anticipated autonomous services were delayed, and a flurry of autonomous safety efforts were launched to address the challenge of AV safety, most notably:
In January of 2019, the publicly available SOTIF draft was released.
In April, SAE announced the formation of the Autonomous Vehicle Safety Consortium (AVSC).
In May, the 3.5 year Pegasus Project concluded with its 2nd Pegasus Symposium.
In July, the Safety First for Automated Driving or SaFAD white paper was publicly released.
In September, TTTech established “The Autonomous” as an initiative for establishing global safety in autonomous mobility.
Finally, in October and December of 2019, the first drafts of the UL 4600 Standard for Safety for the Evaluation of Autonomous Products were released.
Not only were these immense technical challenges to AV safety realized in 2019, the economic challenges became apparent, as well. An MIT study reported by Business Insider estimated the largest annual cost for operating an AV is ‘Safety Oversight,’ with a cost of over $200,000/yr per AV. In January of 2020, it was announced by Waymo’s head of operations that, “for the foreseeable future,” they will be relying on professional driving services, with costs estimated in “tens of millions of dollars, and could reach nine figures,” according to an anonymous source with direct knowledge.
These costs are prohibitively expensive for commercial autonomy. It is cheaper and safer, today, to employ professional drivers to drive regular cars than it is to employ a team of safety drivers and remote observers for AVs. This is why we are asking the question about AV safety assurance, “Is it really worth it?”
The solution, if one exists, would be an affordable and scalable risk management system for autonomous vehicle development and operations. This could leverage human oversight, which is still critical, over as many AVs as possible, in order to lower costs. And the safety assurance it provides would have to be at least equal to, but more likely to be greater than the assurance of humans, alone. Such a risk management system may not be that far off. In fact, a drafted German law for AV regulation was recently reported to require a "’Technical Supervisor’ … to monitor the safety of the journey,” and monitor, “all relevant parameters… for forensic purposes…”
That is why we wanted to let you know that one of the first-ever workshops on Dynamic Risk Management for Autonomous Systems (DREAMS) is taking place at this year’s European Dependable Computing Conference, and EDCC 2020 registration ends August 14th. The DREAMS Workshop will take place on Monday, September 7th, and for many that is the US holiday, “Labor Day,” but it will be held virtually and the agenda is online. There are many great speakers and presenters lined up. We are very excited that we will be presenting a mathematical risk framework, which is the basis of our autonomous risk management system for ensuring autonomous vehicle safety.
We hope you are able to find the time to check that out. It’s possible that the work presented in this virtual workshop becomes the key to enabling truly safe and viable autonomous mobility services, giving us the clear answer, “Yes, autonomous safety assurance is worth it.” We hope you are having a safe year in 2020, and that our optimism continues to grow. Please keep in touch. We look forward to hearing how things are going.
Regards,
Michael Woon
CEO, Founder