Symbolic Simulation of Parametrized Hybrid Systems with Affine Arithmetic

Abstract

The purpose of this research is to develop a highly reliable simulator of hybrid systems, i.e., systems involving both discrete change and continuous evolution. In particular, we aim at rigorous simulation of parametrized hybrid systems, which enables not only the analysis of model's possible behavior but also the design of parameters that realize desired properties. Simulators with interval arithmetic can reliably compute a reachable set of states, but preserving the dependency of uncertain quantities in models is still challenging. In this paper, we discuss a simulation method that is based on symbolic computation and cooperates with the interval Newton method and affine arithmetic, which is able to preserve first-order dependency of uncertain quantities. We implemented the algorithm on the symbolic simulator we have been developing and evaluated the performance of the method with example models.