[SystemSafety] How Many Miles of Driving Would It Take to Demonstrate Autonomous Vehicle Reliability?

[Mike Ellims]

I've download the report, but so far have only looked at the article
attached and a similar one on IEEE Spectrum.

 

This is possibly a straw man in some regards, e.g. Google has a fleet of
vehicles that have driven 1.3 million miles since 2009, however the company
claims to simulate 3 million miles of driving per day based in the
information gathered (e.g. 200,000 stop signs, and 600,000 traffic lights,
180 million vehicles and one rogue duck) and, I assume, the information they
have collected worldwide via StreetView. I always wondered why they had
StreetView.

 

Incidentally Google do some very interesting stuff on software testing.

 

Likewise Tesla is conducting a large scale experiment (you can choose to
participate) with its Autopilot system on 50,000+ vehicles; 100 million
miles is only 2000 miles per vehicle. Tesla gives figures for 2 Billion
miles driven total  with 1 billion this year and 47 million miles (since 14
October 2015 ) on Autopilot - which I understand records and reports
performance.

 

To the best of my knowledge Google and Tesla have one fender-bender each to
date. Tesla's being more interesting as it was with road works on a
Autobahn. Specifically with the stand out reflectors sometimes used.

 

Assuming that functionality is added incrementally then volume of testing
probably isn't actually an issue. What may be an issue is that testing
captures every possible situation. However most drivers can't cope with
every possible situation anyway so that may be moot.

 

And yes my car has ABS, ESP and traction control; all examples of existing
autonomous driving systems.

 

 

From: systemsafety
[mailto:systemsafety-bounces at lists.techfak.uni-bielefeld.de] On Behalf Of
Robert Schaefer
Sent: 14 April 2016 20:52
To: The System Safety List
Subject: [SystemSafety] How Many Miles of Driving Would It Take to
Demonstrate Autonomous Vehicle Reliability?

 

http://arstechnica.com/cars/2016/04/car-makers-cant-drive-their-way-to-safet
y-with-self-driving-cars/

http://www.rand.org/pubs/research_reports/RR1478.html

 

Question: Is there any sort of correlation between the study of safety in
software in avionics and and the study of safety in software in automobiles?

 

I mean, are we (historically speaking) repeating ourselves? - The first as
tragedy, the second as farce?

 

bob s

rps at haystack.mit.edu



---
This email has been checked for viruses by Avast antivirus software.
https://www.avast.com/antivirus
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.techfak.uni-bielefeld.de/mailman/private/systemsafety/attachments/20160414/a320896d/attachment.html>