Abstract:

A signal-path-speed three-level collaborative optimization control-method was proposed to develop an intelligent transportation system for multi-intersection mixed traffic flow environments. At the upper level, the model predictive control (MPC) was adopted to solve the signal timing optimization problem in a rolling horizon manner, and the trust region bayesian optimization (TuRBO) was introduced to adaptively shrink the search region for efficiently approaching the global optimum. At the middle level, the K-Shortest Path (KSP) algorithm was employed to generate candidate path sets; and the method of successive averages was combined to iteratively update path flow assignment for dynamic route guidance. At the lower level, a speed planning model was constructed based on the MPC; and the TuRBO was utilized to optimize the speed trajectories of connected vehicles online, achieving microscopic speed collaborative control. The SUMO (simulation of urban mobility) was used to simulate. The results show that compared to the baseline scenario without collaborative optimization, the proposed method significantly reduces the total vehicle travel time under various connected vehicle penetration rates, with the best optimization rate reaching 67.6%, outperforming single-level or two-level control models. Therefore, this method provides a solution for collaborative control in intelligent transportation systems.

Key words: intelligent transportation systems, multi-level optimization, collaborative control, intelligent connected vehicles, optimal control strategies