Scalable Reachset-Conformant Identification of Linear Systems

Abstract

By monitoring the set of reachable outputs, safety can be verified. However, to compute the reachable set of real-world systems, we require models that are able to produce all possible system behaviors. These kinds of models are called reachset-conformant, and their identification is a promising new research direction. While many existing reachset-conformant identification techniques require the computation of the halfspace representation of the zonotopic reachable sets, we propose an approach that leads to the same optimal identification results using the more scalable generator representation. Thus, our approach offers greater efficiency for high-dimensional systems and long time horizons. The scalability and accuracy of both approaches are compared in numerical experiments with linear time-variant systems.