数据驱动的城市交通子区饱和度分层均衡控制

Urban Traffic Sub-region Based on Data-driven Hierarchical Saturation Balance Control

传统交通子区控制以子区为最小控制单元,难以考虑到不同饱和度情况下的子区内各交叉口的控制方式,为此提出一种数据驱动的城市交通子区饱和度分层均衡的控制方法 .首先,从子区内单交叉口的控制出发,建立交叉口排队均衡的无模型自适应控制方法,实现单交叉口各个相位排队长度的均衡;其次,在单交叉口排队均衡基础上,考虑子区内所有交叉口的综合控制,通过设计交通子区饱和度期望值,构建子区内多交叉口饱和度均衡的无模型自适应控制方法,通过令各交叉口的饱和度趋近期望值,实现子区内饱和度的均衡;结合单交叉口和子区内各交叉口的均衡控制,提出面向城市交通子区的饱和度分层均衡控制方法;最后,采用自主研发的在线交通控制仿真平台对所提方法与传统稳态信号控制方法和传统自组织子区控制方法进行了对比分析.结果表明:与两类传统方法相比,所提方法的在子区1中平均停车次数降低了18.56%,平均延误时间降低了22.43%;在子区2中平均停车次数降低了19.15%,平均延误时间降低了17.22%.

A control method of saturation leveling equalization of urban traffic sub-area is presented based on the data drive.It takes into account that it is difficult to establish an accurate model of traffic flow at intersections in sub-areas and that the control forms of intersections in sub-areas are different under different saturation conditions.Firstly,a model-free adaptive control method for queuing equilibrium at a single intersection is established to balance the queue length of each phase and ensure the effective allocation of phase green time.Secondly,based on queuing equilibrium of a single intersection,the expected saturation value of the traffic subarea is designed,and the model-free adaptive control of multiple intersections in subarea is constructed so that the saturation of each intersection tends to the expected value,and the equilibrium of saturation in subarea is realized.Based on this,a hierarchical equilibrium control strategy of intersection saturation in the subarea is proposed.Finally,a self-developed online traffic control simulation platform was used to compare the proposed method with the traditional steady-state signal control method and the traditional selforganizing subregion control method.The results show that compared with the two traditional methods,the average number of stops and the average delay time of the proposed method are reduced by 18.56%and 22.43%respectively in subregion 1.In subzone 2,the average number of stops decreased by 19.15%and the average delay time decreased by 17.22%.