IIS-Guided DFS for efficient bounded reachability analysis of linear hybrid automata

• Authors: Lei Bu;Yang Yang;Xuandong Li
• Venue: HVC'11 Proceedings of the 7th international Haifa Verification conference on Hardware and Software: verification and testing
• Year: 2011

In the authors' previous work, we proposed a linear programming (LP) based approach to check the reachability specification along one abstract path in a linear hybrid automaton (LHA) at a time by translating the reachability problem into the satisfiability problem of a linear constraint set. Then a depth-first-search (DFS) is deployed on the graph structure of the LHA to check all the paths with length in the threshold to answer the question of bounded reachability. In this DFS-style bounded model checking (BMC) algorithm, once a path is found to be infeasible by the underlying LP solver, a backtracking on the graph structure will be conducted. Clearly, the efficiency of the algorithm depends on the accuracy of the backtracking. If the DFS can backtrack to the most reasonable location, the state space need to search and verify can be reduced significantly. Fortunately, once a linear constraint set is judged to be unsatisfiable, the irreducible infeasible set (IIS) technique can be deployed on the unsatisfiable constraint set to give a quick analysis and find a small set of constraints which makes the whole program unsatisfiable. In this paper, we adopt this technique into our DFS-style BMC of LHA to locate the nodes and transitions which make the path under verification infeasible to guide the backtracking and answer the bounded reachability of LHA more efficiently.