Safety Assessment of Autonomous Cars using Verification Techniques

A common requirement for autonomous cars is a safe locomotion which is evaluated by the method of hybrid verification applied online. The approach checks avoidance of static obstacles and dynamic traffic participants, which are described by imprecise data on their positions and velocities. The nonlinear dynamics of the autonomous car and other traffic participants is conservatively abstracted to Markov chains, which allows the efficient computation of their future positions probabilistically. The result is the probability of a crash for a given time horizon, showing if a given control strategy may lead to unsafe situations.