Testing a battery management system via criticality-based rare event simulation

Abstract

For the validation of safety-critical systems regarding safety and comfort, e.g., in the context of automated driving, engineers often have to cope with large (parametric) test spaces for which it is infeasible to test through all possible parameter configurations. At the same time, critical behavior of a well-engineered system with respect to prescribed safety and comfort requirements tends to be extremely rare, often with probabilities of order 10-6 or less, but clearly has to be examined carefully for valid argumentation. Hence, common approaches like boundary value analysis are insufficient, while methods based on random sampling from the parameter space (simple Monte Carlo) lack the ability to detect these rare critical events efficiently, i.e. with appropriate simulation budget. For this reason, a more sophisticated simulation-based approach is proposed which employs optimistic optimization of an objective function called criticality in order to identify effectively the set of critical parameter configurations. This article documents a case study on applying criticality-based rare event simulation to a charging process (model) controlled by an automotive battery management system, and discusses lessons learned.