Combined Global and Local Search for the Falsification of Hybrid Systems

TL;DR

This paper introduces a hybrid search algorithm combining local derivative-based and global strategies to efficiently find error trajectories in hybrid systems without length restrictions.

Contribution

It presents a novel combined local-global search method for hybrid system falsification that allows arbitrary trajectory lengths and improves efficiency.

Findings

Effective in finding falsifying trajectories

Handles arbitrary trajectory lengths

Combines local derivatives with global search

Abstract

In this paper we solve the problem of finding a trajectory that shows that a given hybrid dynamical system with deterministic evolution leaves a given set of states considered to be safe. The algorithm combines local with global search for achieving both efficiency and global convergence. In local search, it exploits derivatives for efficient computation. Unlike other methods for falsification of hybrid systems with deterministic evolution, we do not restrict our search to trajectories of a certain bounded length but search for error trajectories of arbitrary length.