Risk minimisation for autonomous shuttles in suburban environments based on virtual validation

In recent years autonomous shuttles have been deployed worldwide. Currently, autonomous shuttle trials rely on safety operators monitoring the environment and intervening in critical situations. One of the next steps is to operate shuttles with remote human supervision only, which will require an extensive safety procedure before deployment. Using virtual validation for risk minimisation is an effort to address the current lack of resilient arguments regarding the safety case needed for autonomous shuttle deployment. In this paper, a virtual validation architecture specifically designed to assess autonomous shuttle deployments is proposed. A knowledge-based approach towards scenario generation was used, integrating safety-related key performance indicators for evaluation purposes. The executed test cases for a specific left-turn manoeuvre of the shuttle show the importance of correct vehicle orientation in critical situations. This research contributes towards a verifiable risk minimisation by virtually validating shuttle trajectories, taking location-specific static infrastructure into account.