Abstract: : The traffic signal control system plays a key role in the road network, and its control performance directly affects the traffic safety and delay time in the intersection. Traditional control in five-road intersection does not have the ability to adjust itself, which wastes green time. This paper adopts a method that use artificial fish swarm algorithm (AFSA) to optimize dynamic-fuzzy neural network (D-FNN) to achieve multi-phase and variable phase sequence intelligent control in five-road intersection. Taking the reciprocal of average vehicle delay as the food concentration of AFSA, and the weights and thresholds of the dynamic-fuzzy neural network which need to be modified are used as the individual state of artificial fish. A set of optimal dynamic-fuzzy neural network parameters are obtained through iterating and updating. After doing simulation analyses in the case of different rates of vehicles arrival, the result shows that this method is better than the traditional control in automatically adjusting the signal cycle, and it reduces the average delay of vehicles for about 11%.

Key words: raffic signal control, five-road intersection; artificial fish swarm algorithm, dynamic-fuzzy neural network, average delay of vehicles

摘要： 交通信号控制系统在路网中起关键性的作用，其控制性能直接影响车辆通行安全和路口延误时间。针对五岔路口的传统控制不具备自调整能力，造成绿灯时间的浪费，提出了人工鱼群算法(AFSA)优化动态模糊神经网络(Dynamic-Fuzzy Neural Network ,D-FNN)的方法，实现五岔路口多相位变相序智能控制。以车辆平均延误的倒数作为AFSA的食物浓度，将需要修正的动态模糊神经网络的权值和阈值作为人工鱼的个体状态，通过迭代更新得到一组最优的动态模糊神经网络参数。在不同车辆到达率情况下进行仿真分析，结果表明：该方法比传统的控制方法在自动调节信号周期方面效果更好，车辆平均延误大约减少11 %。

关键词: 交通信号控制, 五岔路口, 人工鱼群算法, 动态模糊神经网络, 车辆平均延误