Breaking Dense Structures: Proving Stability of Densely Structured Hybrid Systems

TL;DR

This paper improves a graph-based decomposition technique for hybrid systems by relaxing the graph structure to handle super-dense switching, enhancing efficiency and applicability for complex cyber-physical systems.

Contribution

It introduces a relaxation method that reduces graph connectivity, enabling more efficient and broader decomposition of densely structured hybrid systems.

Findings

Enhanced decomposition efficiency for complex hybrid systems

Broader applicability to densely connected graph structures

Improved stability analysis through local Lyapunov functions

Abstract

Abstraction and refinement is widely used in software development. Such techniques are valuable since they allow to handle even more complex systems. One key point is the ability to decompose a large system into subsystems, analyze those subsystems and deduce properties of the larger system. As cyber-physical systems tend to become more and more complex, such techniques become more appealing. In 2009, Oehlerking and Theel presented a (de-)composition technique for hybrid systems. This technique is graph-based and constructs a Lyapunov function for hybrid systems having a complex discrete state space. The technique consists of (1) decomposing the underlying graph of the hybrid system into subgraphs, (2) computing multiple local Lyapunov functions for the subgraphs, and finally (3) composing the local Lyapunov functions into a piecewise Lyapunov function. A Lyapunov function can serve multiple purposes, e.g., it certifies stability or termination of a system or allows to construct invariant sets, which in turn may be used to certify safety and security. In this paper, we propose an improvement to the decomposing technique, which relaxes the graph structure before applying the decomposition technique. Our relaxation significantly reduces the connectivity of the graph by exploiting super-dense switching. The relaxation makes the decomposition technique more efficient on one hand and on the other allows to decompose a wider range of graph structures.