Validation of Collision Frequency Estimation Using Extreme Value Theory
Paper in proceeding, 2018

There is a lot of focus right now on how to build an autonomous vehicle, which can handle all the situations that a human driver is experiencing. Less is done on how to ensure that these vehicles are safe enough to be released to the public. Using traditional statistical methods would require one to drive extensive distances without incidents to prove the safety to a sufficient degree. Recent research has shown the possibility of using near-collisions in order to estimate the frequency of actual collisions using Extreme Value Theory. In order to trust these estimations, the precision of these estimates needs to be validated. The results from a 250 000 km field test shows that the Extreme Value estimations are reasonable in relation to a crash statistics estimate for rear-end collisions. This further suggests that extreme value is a method that can be used to predict collision frequencies from data containing no collisions.

Safety

Automotive

Verification & Validation

Statistical inference

Autonomous vehicles

Author

Daniel Åsljung

Chalmers, Electrical Engineering, Systems and control

Zenuity AB

Jonas Nilsson

Zenuity AB

Jonas Fredriksson

Chalmers, Electrical Engineering, Systems and control

Proceedings of the IEEE Intelligent Transportation Systems Conference, 2017

1857-1862

Areas of Advance

Transport

Subject Categories

Electrical Engineering, Electronic Engineering, Information Engineering

Probability Theory and Statistics

DOI

10.1109/ITSC.2017.8317596

More information

Latest update

3/5/2020 1