Flutterby™! : Crash Rates

Next unread comment / Catchup all unread comments User Account Info | Logout | XML/Pilot/etc versions | Long version (with comments) | Weblog archives | Site Map | | Browse Topics

Crash Rates

2023-09-06 23:37:18.632099+02 by Dan Lyke 2 comments

Missy Cummings, George Mason University: Assessing Readiness of Self-Driving Vehicles (preprint)

State agencies are increasingly faced with self-driving permit and licensing requests as self-driving operations expand. However, these expansions have led to congestion and problematic interactions with first responders, as well as increasing public distrust. To respond to these self-driving permit requests with evidence-based recommendations, government agencies need straightforward tools to help them objectively and holistically assess such requests. To this end, using self-driving disengagement data from California, as well as federal non-fatal and CA transportation network companies’ crash reports, this effort demonstrates how the combination of human- and autonomy-initiated disengagements, coupled with non-fatal crash rates, can provide insight into assessing self-driving vehicle readiness for commercial operations. Additional results show that Cruise’s and Waymo’s robo-taxis in San Francisco are 4-8x morelikely to be involved in non-fatal crashes, equivalent to the CA crash rates of Uber and Lyft. One major drawback to this approach is a lack of reporting by the majority of companies conducting self-driving operations on public roads in CA. This lack of reporting and companies’ avoidance of publicly address emerging problems, while simultaneously claiming their technologies are superior to human drivers, suggests there are systemic problematic safety cultures in the self-driving community. If self-driving companies do not adopt more transparent and responsive safety practices, their non-fatal crash rates could continue to exceed that of human drivers. They also risk further eroding public sentiment, which could lead to further public rejection of what otherwise could have been a promising technology.

I'mma just repeat that in bold: "Cruise’s and Waymo’s robo-taxis in San Francisco are 4-8x morelikely to be involved in non-fatal crashes, equivalent to the CA crash rates of Uber and Lyft"

[ related topics: Interactive Drama broadband Bay Area moron California Culture Community Education ]

comments in ascending chronological order (reverse):

#Comment Re: Crash Rates made: 2023-09-07 12:35:46.003635+02 by: brainopener [edit history]

Similar from the paper:

...while Cruise and Waymo self-driving vehicles appear to be on par with non-fatal crashes occurring in similar human driving scenarios, these results beg an important regulatory question as to why such high crash rates are acceptable regardless of whether they are the result of human or robot actions?

  1. I wonder how the crash rates of old-style taxi cabs compare?
  2. It's always been a concern out there on the fringe that a variety of factors would make human driving worse --- effectively let machine driving and human driving skill "meet in the middle" of the intersection collisionally . I wonder if that's in the data?
  3. I don't know. Maybe the only way out is through and we just need to get to an all robot car present quickly so we can move on to something better.

#Comment Re: Crash Rates made: 2023-09-09 16:43:44.026251+02 by: Dan Lyke

  1. Yeah, I wonder if there's any way to extract that information out from the general safety data.
  2. Related is the question of how interactions with automated systems are different: pedestrians are going to adapt to how the automated systems behave differently...
  3. I don't know either, but I do know that geometry makes designing mobility systems around private automobiles kind-of a dead-end thing.