Toward Crash-Free Autonomous Driving: Anomaly Detection and Control for Resilience to Stealthy Sensor Attacks

Cooperative adaptive cruise control (CACC) enables connected and automated vehicles (CAVs) to drive autonomously on the highway in closely coupled platoons. The use of CACC technologies increases safety and the traffic throughput, and decreases fuel consumption and CO2 emissions. However, CAVs heavily rely on embedded software, hardware, and communication networks that make them vulnerable to a range of cyberattacks. Cyberattacks to a particular CAV compromise the entire platoon as CACC schemes propagate corrupted data to neighboring vehicles potentially leading to traffic delays and collisions. Physics-based monitors can be used to detect the presence of false data injection (FDI) attacks to CAV sensors; however, given enough system knowledge, adversaries are still able to launch a range of attacks that can surpass the detection scheme by hiding within the system disturbances and uncertainty—we refer to this class of attacks as stealthy FDI attacks. Stealthy attacks are hard to deal with as they affect the platoon dynamics without being noticed. In this manuscript, we propose a design methodology (built around a series convex programs) to synthesize distributed attack monitors and $H_{\infty }$ CACC controllers that minimize the joint effect of stealthy FDI attacks and system disturbances on the platoon dynamics while guaranteeing a prescribed platooning performance. Computer simulations are provided to illustrate the performance of our tools.