Abstract: In order to solve the problems of unbalanced traffic flow, unsatisfied traffic demand and irregular time distribution in two-way two-lane scenario, a variable one-way traffic control algorithm was proposed. Taking the traffic density and queue length information obtained from the roadside as constraints, the optimal switching model was formed by using fuzzy control and rolling traffic switching factor method, and the optimal transition clearance scheme for multiple regions of variable road sections was obtained by Q-learning clearance algorithm. The simulation results show that taking 90 s as a detection step size, the switching scheme was real-time acquisitioned and generated, with an accuracy of 91.5%. The transition clearance module generates the optimal clearance scheme for dividing areas according to the real-time status, which reduces the clearance time. When the directional unevenness coefficient is 1.5 and 5.0, compared with the current road network, the overall delay of the road network is reduced by 29.4 % and 25.2 %, the average running speed is increased by 6.8 % and 9.1 %, and the vehicle detour time is reduced by 6.2 % and 7.9 %, which significantly improves the overall service level.

Key words: traffic and transportation engineering, intelligent transportation, variable one-way traffic control algorithm, fuzzy control, one-way traffic, internet of vehicles, Q-learning

摘要： 为解决双向两车道场景下不均衡交通流通行需求不满足、时间分布不规律的问题，提出了一种可变单向交通控制算法。将路侧获取的交通密度、排队长度信息作为约束，利用模糊控制与滚动交通切换因子方法形成优化切换模型，通过Q学习清空算法获取可变路段多个区域的最优过渡清空方案。仿真验证表明：以90 s为一个检测步长实时采集生成切换方案，准确率达91.5 %，过渡清空模块根据实时状态生成划分区域最优清空方案，减少了空放时间；在方向不均系数为1.5和5.0时，相比现状路网整体延误降低29.4 %、25.2 %，平均运行速度提高6.8 %、9.1 %，车辆绕行时间减少6.2 %、7.9 %，显著提升了整体服务水平。

关键词: 交通运输工程, 智能交通, 可变单向交通控制算法, 模糊控制, 单向交通, 车联网环境, Q学习