Friday, June 8, 2018

NTSB report: Tesla autopilot responsible for fatal crash

It accelerated into a concrete wall:
The National Transportation Safety Board has released its preliminary report on the March crash that killed driver Walter Huang in Mountain View. The report provides a second-by-second description of the events that preceded Huang's collision with a concrete lane divider. 
The report confirms that Autopilot was engaged ahead of the crash, and it appears to confirm that a navigation mistake by Autopilot contributed to Huang's death.
Are you willing to trust your life to a huge software program that nobody really knows how it works?

4 comments:

Rick C said...

"Are you willing to trust your life to a huge software program that nobody really knows how it works?"

If you read the comments at Arse Technica, you will find there are not only lots of people willing to, but lots willing to require you and me to as well.

Borepatch said...

Rick, there are a lot of techno-nerds at Ars. Fortunately, there are 100 Million + non-autonomous cars on the road. Congress won't do anything to make people replace them, because they don't want 100 M + people voting against them in the next election.

Personally, I think that insurance rates will be the tell on where this technology goes. As with all tech, the first 20% of the effort gives 80% of the results. Getting to 100% is the challenge, and I suspect that the safety savings of that final 20% won't remotely justify the expense of the final 80% effort.

Differ said...

I would like to know to what Development Assurance Level (RTCA DO 254 & 178) the autopilot computer in the car and the software installed on it are designed and built. Did they follow appropriate development guidance e.g. SAE ARP 4754 for system development? I know these are aviation standards, but an autonomous car exposes occupants and bystanders to the same risks, albeit on a smaller scale. Are there similar requirements and guidelines for the automotive industry?
I would expect they cross-polinate from SpaceX which must comply with stringent requirements; so if they ARE using good standards doesnt that further indicate the extreme complexity of autonomous vehicles compared to aerospace?

Borepatch said...

Differ, I'd be very surprised if any of these standards are followed. I don't think that the automotive industry adheres to standards for this sort of thing.