MOBATSim: MOdel-Based Autonomous Traffic Simulation Framework for Fault-Error-Failure Chain Analysis

摘要

Safety analysis is an important topic affecting many different domains in engineering. As one of the most safety-critical domains, the automotive sector needs tools and techniques to assess the safety of individual vehicles as well as the traffic as a whole. A new simulation framework, MOBATSim, is presented where simulation-based fault injection is used to assess the safety of autonomous driving systems on component, vehicle, and traffic levels. The simulation framework is designed in MATLAB Simulink and provides the building blocks for modeling various driving scenarios including the urban environment map and waypoints for vehicles. It is based on autonomous vehicle models which are extended with fault injection models. Several types of faults can be injected during the simulation runtime, and the outputs can be examined to verify the safety criteria, specified by the user. MOBATSim supports full fault-error-failure chain analysis and is capable of revealing the relationships between component faults and failures on the vehicle and traffic levels. An illustrative case study shows the analysis of safety criteria violations caused by particular low-level sensor faults in a user-defined driving scenario.