Cyberphysical Fault Propagation and Detection in Connected Electric Vehicles

Connected electric vehicles are an integral part of smart mobility infrastructure that offer improved traffic flow through intelligent decision-making. However, these vehicular networks are vulnerable to cyberphysical faults. A platoon vehicle experiencing a fault or cyberattack may also propagate disturbances to other vehicles via the network, amplifying adverse effects, while obfuscating attempts at diagnosis. This work investigates the propagation of disturbances caused by cyberphysical faults as well as fault detection. We use machine learning-based fault classifiers, including a multi-head attention architecture, to classify simultaneous cyberphysical faults in the EV platoon. We demonstrate that the attention-based method performs well for simple communication topologies, but poorly for increased connectivity. We further demonstrate that the method generalizes to large platoons and succeeds in binary classification for unseen scenarios.