基于ε-支持向量回归理论的区域交通信号智能控制

城市交通信号控制是当前智能交通领域的研究热点之一.针对区域交通信号协同控制的实时性和准确性,提出一种基于ε-支持向量回归(SVR)非线性回归理论的智能控制方法(ICSRTS).该方法在无线传感网络结构的基础上结合已有的数据汇聚算法,并采用分簇策略将区域交通控制系统建模成一类集成信息调度与控制的离散切换系统.在离散切换系统中,不仅考虑了数据包传输的网络时延和丢包率,而且观测器利用改进的ε-SVR训练方法实现对多数据源融合的交通信号状态的在线预测并通过控制器进行总体协调控制.运用Lyapunov 函数方法验证了该系统的渐近稳定性及其可调度性. 仿真结果表明,ICSRTS方法相比普通模糊神经网络控制和普通ε-SVR预测算法在交叉口平均延误时间方面具有较好的性能.因此,该方法能实时、有效地对区域交通信号进行协调控制,从而减少了区域内的交通拥堵和能源消耗.

Intelligent control based on ε-support vector regression theory for regional traffic signal system

Intelligent control of urban traffic signal is an important element of intelligent transportation system. In order to meet the real-time and accuracy for the regional traffic signal coordinated control, this paper presented an Intelligent Control Strategy for Regional Traffic Signal (ICSRTS) based on ε-SVR (Support Vector Regression) nonlinear regression theory. Combining with the existing data aggregation algorithm, ICSRTS was based on the wireless sensor network structure, and it adopted the clustering strategy to create a model of discrete switching system, which integrated the information scheduling and control for the regional traffic system. In the discrete switching system, the network delay and packet loss rate for data transmission were considered, furthermore, the observer used the modified ε-SVR theory to realize the online prediction of the multi-source data based traffic state, then the controller carried out coordination control of the overall traffic signal. The asymptotic stability of discrete switching system was analyzed using Lyapunov function. Simulation results show that ICSRTS has better performance in the intersection average delay time compared with ordinary fuzzy neural control and ordinary ε-SVR prediction algorithm. Therefore, this method can realize the regional traffic signal coordinated control in real-time and effectively, and reduce the area of traffic congestion and energy consumption.