Hybrid Reachability Analysis for Kuramoto-Lanchester Model

Abstract

Cyber-physical systems are ubiquitous nowadays and play a significant role in people's daily life. These systems include, e.g., autonomous vehicles and aerospace systems. Since human lives rely on the performance of these systems, it is of utmost importance to ensure their reliability. However, their complexity makes analysis particularly challenging and computationally expensive. Thus, it is crucial to develop tools to efficiently analyze cyber-physical systems and their safety properties. Cyber-physical systems are often modeled by hybrid automata, i.e. finite-state machines augmented with ordinary differential equations. In the thesis, we investigate reachability analysis methods for hybrid automata. In particular, we extend JuliaReach, a framework for fast prototyping set-based reachability analysis algorithms, to support verification of hybrid automata. For this purpose, we add to JuliaReach concrete and lazy discrete post operators. Lazy operations are particularly efficient in flowpipe based reachability analysis with long sequences of computations. The implemented algorithms are interchangeable and support all three reachability scenarios available in JuliaReach for the purely continuous setting: techniques to analyze linear systems using support functions and zonotopes as well as Taylor model based analysis for nonlinear systems. In order to evaluate our methods, we apply them to the Kuramoto-Lanchester model. This model exhibits highly nonlinear dynamics and can be easily scaled, and thus is well-suited to assess performance of reachability analysis methods for hybrid automata. Show more