基于增强学习的网格化出租车调度方法

高度信息化的网格化城市管理可以为出租车运营优化提供新的实时动态乘客需求信息和车辆位置信息。以此为契机，针对城市出租车空驶率高和司乘匹配率低的问题，提出了一种网格化的出租车实时动态调度的增强学习控制方法。通过为出租车提供空驶巡游的动态最佳路线，新的控制方法旨在提高出租车的服务效率，并降低乘客的等待时间。首先，以城市单元网格为基础，明确出租车调度的关键问题；其次，以空驶路线的动态调整为控制手段，建立调度的增强学习模型；最后，给出求解模型的Q学习算法，并通过算例验证新调度方法的有效性。研究表明新方法可以有效提高司乘匹配率、增加总的出租车运营收入、减少乘客平均等车时间和减少总的出租车空驶时间。

Grid-based taxi dispatching method based on reinforcement learning

Highly-informed grid-based city management can supply the real time passenger information and the position information of taxis for taxi operation optimization. On this account, we proposed a grid-based taxi dispatching dynamic control method based on reinforcement learning to solve the problem of the high vacant taxis rate and the low matching rate between taxis and passengers. By providing the optimal cruising routes for the vacant taxis, the new control method aims to improve the service level of taxis and to lower the waiting time of passengers. Firstly, based on the grids of city, we clarified the key problem of taxi dispatching. Secondly, by using the adjustment of vacant taxi route as the control action, we formulated the reinforcement learning model of taxi dispatching. At last, we proposed the corresponding Q learning algorithm to solve the new model. Numerical example demonstrated the effectiveness of the new dispatching method. The results show that the new method can not only increase the match rate between taxis and passengers and the total income of operation of taxi service, but also reduce the average waiting time of passengers and the total travel time of vacant taxis.