Eslami and Yu's platoon control blocks false data injection attacks

In a new paper on arXiv, Ali Eslami and Jiangbo Yu tackle a critical security challenge for autonomous vehicle fleets: False Data Injection (FDI) attacks on vehicle-to-vehicle (V2V) communication channels. The team models each vehicle's longitudinal motion as a switched third-order system, capturing mode-dependent powertrain dynamics as the convoy shifts between operating states. Instead of assuming attacks have limited size, they consider attackers with bounded rates but potentially unbounded amplitudes—a more realistic and dangerous threat model. To counter this, each vehicle runs an auxiliary observer on every communication link, producing real-time estimates of injected attack signals. These estimates feed a resilient controller that compensates for the malicious data, ensuring the entire platoon remains uniformly ultimately bounded even under extreme attack scenarios.

The analysis goes beyond raw stability. For a predecessor-following topology, the authors establish string stability of the nominal switched platoon—meaning disturbances don't amplify as they propagate down the convoy. When attack-estimation errors are present, they prove the effect on acceleration propagation is bounded, offering a graceful degradation guarantee. This matters because real-world platoons rely on dense, low-latency V2V data for safe following distances; a successful FDI attack could cause cascading collisions. By providing per-link observability and a certified stability margin, this work gives control engineers a concrete framework for building cyber-resilient platooning systems. Numerical case studies validate the theoretical guarantees, moving the approach closer to deployment in autonomous trucking and connected vehicle fleets.