基于Q-learning 的定制公交跨区域路径规划研究

考虑城市大客流通勤者跨区域出行需求，结合城市公交线网中乘客出行密集、客流走向规律等特点，提出一种跨区域定制公交的搭乘方案. 通过改进的Q-learning 模型对公交线路进行优化，为城市通勤者提供更加便捷和高效的出行服务. 通过综合路段拥堵状态、乘客需求及居民小区位置，设定了Q-learning 强化学习的奖惩函数，提升定制公交区域路径的直线系数、满载率、通行时间. 结果表明，所提出的改进方法能够降低通勤者跨区域通行的旅行时间，有效提高髙峰时段定制公交线网的通行效率.

Cross-regional Customized Bus Path Planning Based on Q-learning

This paper investigates a customized transit scheduling strategy for urban residents commuting across multiple regions with comprehensively considering the demand of massive urban commuters, as well as the characteristics of transit passenger density and flow in urban network. The Q- learning reinforcement learning improved method is applied to optimize the bus route. Through the integrated road congestion status, passenger demand and residential area location, the reward and punishment function of Q-learning reinforcement learning is set, and the linear coefficient, full load rate and transit time of the customized bus area path are improved. The results show that the proposed improved method can reduce the travel time of commuters across regions, and effectively improve the efficiency of customized bus lines during peak hours.