Resilient Supervisory Control of Autonomous Intersections in the Presence of Sensor Attacks
This addresses safety-critical issues in cyber-physical systems like autonomous vehicles, but it is incremental as it builds on existing supervisory control methods with added resilience features.
The paper tackles the problem of supervisory control for autonomous intersections vulnerable to sensor attacks, which can cause collisions or deadlocks, and presents a resilient control system that ensures safety, non-deadlocking, and maximal permissiveness despite such attacks.
Cyber-physical systems (CPS), such as autonomous vehicles crossing an intersection, are vulnerable to cyber-attacks and their safety-critical nature makes them a target for malicious adversaries. This paper studies the problem of supervisory control of autonomous intersections in the presence of sensor attacks. Sensor attacks are performed when an adversary gains access to the transmission channel and corrupts the measurements before they are received by the decision-making unit. We show that the supervisory control system is vulnerable to sensor attacks that can cause collision or deadlock among vehicles. To improve the system resilience, we introduce a detector in the control architecture and focus on stealthy attacks that cannot be detected but are capable of compromising safety. We then present a resilient supervisory control system that is safe, non-deadlocking, and maximally permissive, despite the presence of disturbances, uncontrolled vehicles, and sensor attacks. Finally, we demonstrate how the resilient supervisor works by considering illustrative examples.