Using Disparity to Enhance Test Generation for Hybrid Systems

This paper deals with the problem of test generation for hybrid systems, which are systems with mixed discrete and continuous dynamics. In our previous work [9], we developed a coverage guided test generation algorithm, inspired by a probabilistic motion planning technique. The algorithm is guided via a process of sampling goal states, which indicate the directions to steer the system towards. In this paper, we pursue this work further and propose a method to enhance coverage quality by introducing a new notion of disparity. This notion is used to predict the situations where the goal states can not be ‘directly’ reached. We then develop an adaptive sampling method which permits improving coverage quality. This method was implemented and successfully applied to a number of case studies in analog and mixed-signal circuits, a domain where hybrid systems can be used as an appropriate high level model.