A novel hierarchical cooperative merging control model of connected and automated vehicles featuring flexible merging positions in system optimization

Merging sections on freeways are typical bottlenecks for traffic efficiency and safety. The connected and automated vehicles (CAV) technology has great capability in improving traffic performance of merging bottlenecks. Current approaches of system optimal cooperative merging control assume single fixed merging point (CMC-SMP) for on-ramp vehicles to reduce the complexity of their models. Non-fixed merging positions have the potential to further improve merging operations but may lead to the computational explosion. To solve this core contradiction, this study proposes a novel hierarchical system optimal cooperative merging control model considering flexible merging positions (CMC-FMP) to realize safe and efficient merging processes. The upper-level - the tactical planning model is formulated as a non-convex mixed integer quadratically constrained programming, aiming at minimal total travel time in the control zone. It optimizes not only the merging sequence but also vehicles' critical states in merging processes, such as the state of on-ramp vehicles when they merge. The lower-level -the motion planning model generates feasible trajectories and the next step actions for every vehicle. A Monte Carlo Tree Search-based decomposition algorithm (MCTS-DA) is further designed to improve the computational efficiency of the tactical planning model. Meanwhile, solutions of MCTS-DA are proved to have good optimality compared to the direct solving method. A batch-based scheme is developed to realize real-time control. The results reveal that the average delays of CMC-FMP are remarkably lower than those of CMC-SMP. Especially in low on-ramp ratio scenarios, the difference in improvement can reach 64%. Furthermore, the sensitivity analysis indicates that large safe merging margins and short merging sections are disadvantageous for traffic efficiency for the merging flow controlled by CMC-FMP.