Physics-Informed Data-EnablEd Predictive Control for Regulating Mixed Traffic Flows

The mixed traffic flow consists of both Connected and Automated Vehicles (CAVs) and Human-Driven Vehicles (HDVs), with the overall dynamics heavily influenced by the unpredictable driving behavior of HDVs. In this context, we propose a Physics-Informed Data-Enabled Predictive Control (PI-DeePC) method tailored for the optimal control of CAVs. This method seamlessly integrates partial system physics to achieve a balance between safety and traffic efficiency in regulating mixed traffic flow through distributed CAVs. By incorporating system physics described in an input/output mapping function, e.g., a state-space equation, the explorations in the data and physical space can be harmonically achieved in a unified framework to synthesize optimal control decisions. This study encapsulates the physics of mixed traffic flow, incorporating variables such as velocity and acceleration, into a comprehensive state-space equation. Subsequently, we implement the proposed PI-DeePC framework to strategically regulate mixed traffic flows, taking into account the effects of measurement noise and the uncertain, diverse behaviors of HDVs. Simulation results demonstrate the efficacy, reliability, and robustness of the PI-DeePC in controlling CAVs within an unknown environment involving significant noise, as visualized in the enhanced control performance.