Falsifying Motion Plans of Autonomous Vehicles With Abstractly Specified Traffic Scenarios

Abstract

Verifying the safety of autonomous vehicles is one of the major challenges towards their deployment on public roads due to the vast number of possible situations that can occur in traffic. Scenario-based testing has been proposed to reduce the number of required tests using catalogs of abstractly defined scenarios. However, when specifying test scenarios abstractly, there is still an infinite number of possible concrete scenarios that can be derived from a specification. Available computational resources can thus be easily exceeded when using inefficient strategies for deriving concrete scenarios. In this work, we propose a novel method that synthesizes concrete scenarios complying with abstract scenario specifications. We compute the set of compliant trajectories for surrounding traffic participants and only sample trajectories within those sets. Our synthesis integrates a falsification algorithm that searches for specified failures of the vehicle under test. Compared to existing work, we can efficiently find scenario concretizations, especially for complex maneuver specifications. By directly considering failure specifications during the scenario synthesis, we avoid executing irrelevant simulations that cannot possibly result in failures. Our approach is demonstrated in several scenarios using Monte Carlo tree search as a search algorithm.