TRUST-ME, Time-efficient RobUSt verification of auTonomuous vehicles - theory and Methods
Self-driving cars provide the potential to make traffic safer, smoother and more fuel efficient. New verification methods are needed in order to ensure that the accident risk of a self-driving car is sufficiently low, A major challenge is that traffic accidents occur with a low frequency, meaning lots of field test data has to be collected to draw a statistically valid conclusion. This project will develop methods that verify whether these safety requirements are met for the final product. These methods include: Estimating the probability of the self-driving car being involved in an accident, from field test data; Enhancing this estimate with directed testing in simulation or test tracks; “Short-loop” verification for changes in the functionality during the development process.
Volvo Car Corporation, Chalmers