针对考虑时变交通需求下, 相邻交叉口混合交通流信号配时的优化问题,以机动车延误、非机动车延误、行人等待时间为优化目标,构建动态多目标优化配时模型,提出了多种群协同进化动态多目标遗传算法(multi-populationcoevolutionary dynamic-multi-objective genetic algorithm,MPCED-MOGA), 把种群分为搜索种群和跟踪种群,在寻找最优值的同时跟踪环境的变化和进行信息交流,使算法快速响应环境变化.该算法首先在FDA系列动态多目标优化问题中的3个典型测试函数下进行测试,并与3种现有的动态多目标优化算法进行对比,结果表明该算法具有较好的收敛性和分布性.在仿真环境下测试该算法在上海市某个实际相邻交叉口的信号配时优化效果,结果表明:和3种现有的动态多目标优化算法和动态定时控制TRRL方法相比,该算法能更好地降低机动车延误、非机动车延误和行人等待时间.

To address the traffic signal timing plan optimisation problem of mixed traffic flow at adjacent intersections under dynamic traffic demand, the vehicle delay, nonmotor vehicle delay, and pedestrian waiting time are used as optimisation objectives, a dynamic multi-objective optimisation model is constructed, and a multi-population coevolutionary dynamic-multi-objective genetic algorithm (MPCED-MOGA) is proposed, where the population is divided into search and tracking populations. The optimal solution is searched by the search population, while the changes in the environment are tracked by the tracking population. To obtain an algorithm quickly responding to the environment variations, the information is exchanged constantly between the two populations during the evolution. The proposed algorithm is tested under three classical test functions in FDA series of dynamic multi-objective optimisation function. The result is compared to those obtained by three existing dynamic multi-objective optimisation algorithms. The proposed algorithm is advantageous in terms of convergence and distribution. The effectiveness of the proposed algorithm for the signal optimisation problem is evaluated under a simulation environment based on a real adjacent intersection in Shanghai. The proposed MPCED-MOGA reduces the vehicle delay, nonmotor vehicle delay, and pedestrian waiting time compared to those of the three existing dynamic multi-objective optimisation algorithms and dynamic fixed time control method (TRRL).