Cooperative Adaptive Cruise Control of Connected and Autonomous Vehicles via Hybrid Iteration

This paper proposes a novel data-driven approach, named hybrid iteration (HI), for the implementation of cooperative adaptive cruise control (CACC) in connected and autonomous vehicles (CAVs). The proposed method leverages the integration of adaptive dynamic programming (ADP), internal model principle, and distributed control technique. By collecting input-state data, an approximately optimal control policy is synthesized through the online iterative technique, HI, to optimize the performance of CACC by minimizing a predefined cost function. Note that the HI preserves the benefits of two established learning algorithms: policy iteration (PI) and value iteration (VI). The learning process initiates with the VI algorithm, followed by the application of the PI algorithm to expedite convergence. HI eliminates the necessity for an initial stabilizing control policy, which is a prerequisite in PI, and exhibits a faster convergence rate than VI. The efficiency and safety of the HI method in CACC systems have been demonstrated through platoon vehicle testing and verification in traffic congestion and cut-in scenarios.