Skip to main content

An analysis of real-world crashes involving self-driving vehicles

A study by the University of Michigan performed a preliminary analysis of the cumulative on-road safety record of self-driving vehicles for three of the ten companies that are currently approved for such vehicle testing in California (Google, Delphi, and Audi). The analysis compared the safety record of these vehicles with the safety record of all conventional vehicles in the US for 2013 (adjusted for underreporting of crashes that do not involve a fatality).
October 30, 2015 Read time: 2 mins
A study by the University of Michigan performed a preliminary analysis of the cumulative on-road safety record of self-driving vehicles for three of the ten companies that are currently approved for such vehicle testing in California (1691 Google, 7207 Delphi, and 2125 Audi).

The analysis compared the safety record of these vehicles with the safety record of all conventional vehicles in the US for 2013 (adjusted for underreporting of crashes that do not involve a fatality).

Taking into account the fact that the distance accumulated by self-driving vehicles is still relatively low compared with conventional vehicles and that the vehicles were driven only in limited conditions, the study came up with some interesting results.

These, including the facts that self-driving vehicles may have a higher crash rate per million miles travelled than conventional vehicles, and self-driving vehicles were not at fault in any crashes they were involved in, are available in the report abstract on the university’s website. (link %$Linker: 2 External <?xml version="1.0" encoding="utf-16"?><dictionary /> 0 0 0 oLinkExternal http://www.umich.edu/~umtriswt/PDF/UMTRI-2015-34_Abstract_English.pdf Visit Umich false http://www.umich.edu/~umtriswt/PDF/UMTRI-2015-34_Abstract_English.pdf false false%>).

For more information on companies in this article

Related Content

  • Future Railway launches pantograph design competition
    March 12, 2014
    As part of the UK’s rail electrification programme, FutureRailway is launching a competition to design a Pantograph Dynamic Behaviour Measurement Device for use in Rolling Stock Maintenance Depots. Currently pantographs cannot be run too close together and are limited in the speed they can achieve. Electric trains which can run at faster speeds whilst coupled together in multiple could improve both train performance and network capacity. Improvements in pantograph capability are thought to be needed to r
  • StreetLight Data reveals longest commutes to work across US
    March 12, 2018
    Commuters in the US state of Maine travel 9.8 miles to work on a daily basis, a 72% longer journey than the 5.7-mile median commute in Wyoming, according to the latest report by StreetLight Data. The transportation analytics company conducted its study on nearly 30,000 ZIP codes in areas with over 1,000 residents as part of a strategy to help encourage businesses and governments to start fixing the accessibility gap. The inquiry also highlighted the longest and shortest commutes in each state by ZIP cod
  • Asfinag displays Unterwegs service for traffic and weather conditions
    October 22, 2012
    Asfinag, the Austrian motorway operator, is showing its full service portfolio in the areas of traffic information as well as traffic management at the ITS World Congress. The company’s Unterwegs service provides current traffic and weather conditions on Austrian motorways and expressways for users whether at home, at work, or on the move. Before leaving the office for the journey home, mini applications and gadgets provide commuters and users of modern operating systems like Windows7, Mac OSX with Asfinag
  • Vision Components’ ANPR library identifies damaged plates
    October 29, 2014
    Vision Components’ Carrida software engine is a powerful OEM Library, which is easy to integrate into existing security and surveillance applications, including access control, toll control and traffic analysis. The high-performance, hardware-independent software tool assures reliable high speed recognition of vehicle number plates. With a typical processing time of 30ms and a recognition accuracy of more than 96%, the ALPR/ ANPR library precisely and reliably identifies number plates, even if dirty, dam