基于改进的隐马尔可夫模型交通拥堵识别研究

Research on Traffic Congestion Recognition Based on Improved Hidden Markov Model

为准确高效地追踪识别城市区域交通路况信息,提供合理的交通出行策略,针对原始的隐马尔可夫模型(hidden markov model,HMM)初始状态参数难以选择且训练过程极易陷入局部最优解的问题,提出了一种改进的隐马尔可夫模型的交通拥堵态势识别机制,有效地拟合了城市道路相邻交叉口交通拥堵状况.将粒子群优化(particle swarm optimization,PSO)算法引入到隐马尔可夫模型的训练中,结合Baum-Welch算法分别对该模型的状态数等参数进行优化,最后根据Viterbi算法聚类出城市道路交叉口最佳拥堵状态序列.根据采集的真实交通流和GPS数据、车辆延误时间特征数据进行实验,其结果表明,改进的隐马尔可夫模型在道路交通拥堵识别的准确率和稳定性上有明显提升.

In order to accurately and efficiently track and identify the information of urban traffic conditions,and provide a proper traffic strategy,an improved traffic congestion identification mechanism based on Hidden Markov Model(HMM)was proposed to effectively fit traffic congestion at adjacent intersections of urban roads.The proposal of the mechanism was aimed at the problems that it is difficult to select the initial state parameters in the original Hidden Markov Model and easy for the training process to fall into a local optimal solution.The Particle Swarm Optimization(PSO)algorithm was introduced into the training of HMM,combined with Baum Welch algorithm,to optimize the state number and other parameters of the model.Finally,the optimal sequence of congestion states at urban road intersections was clustered based on the Viterbi algorithm.Experiments based on the collected real traffic flow,GPS data and characteristic data of vehicle delay time were carried out.The results show that the improved HMM can significantly improve the accuracy and stability of road traffic congestion identification.