Temporal Logic Falsification of Cyber-Physical Systems using Input Pulse Generators

Abstract

Falsification is a testing method for cyber-physical systems where numerical optimization is used to find counterexamples of a given specification that the system must fulfill. The falsification process uses quantitative semantics that play the role of objective functions to minimize the distance to falsifying the specification. Falsification has gained attention due to its versatile applicability, and much work exists on various ways of implementing the falsification process, often focusing on which optimization algorithm to use, or more recently, the semantics for the formal requirements. In this work, we look at some practical aspects of input generation, i.e., the mapping from parameters used as optimization variables to signals that form the actual test cases for the system. This choice is critical but often overlooked. It is assumed that problem experts can guide how to parameterize inputs; however, this assumption is often too optimistic in practice. We observe that pulse generation is a surprisingly good first option that can falsify many common benchmarks after only a few simulations while requiring only a few parameters per signal.