Reachability Analysis Plus Satisfiability Modulo Theories: An Adversary-Proof Control Method for Connected and Autonomous Vehicles

Connected and autonomous vehicles (CAVs) are expected to operate with safety guarantee in presence of adversaries from the Internet of Vehicles. This article proposes a control method named reachability analysis plus satisfiability modulo theories (RA-SMT) for CAVs against integrity attacks caused by bounded adversary. This method enables vehicles to possess the reach-avoid specification and strict control safety ensurance even in the worst case scenario. The introduction of state-feedback control decomposes the original complex problem into more manageable reachability analysis and adversary-free control strategy optimization. Precisely, zonotope sets are employed for reachability analysis, and the control strategy is optimally solved and verified simultaneously via SMT. This method is applicable to complex traffic scenarios with the help of SMT, which can describe various constraints flexibly and conveniently. Simulation results reveal the effectiveness and safety of the proposed method in the classical car-following scenario against bounded adversary under various conditions. Particularly, RA-SMT exhibits significantly improved control performance (around 16%) and computation efficiency (around 63%) compared with existing methods. Finally, the RA-SMT is implemented on an autonomous driving platform to validate its practicability.