基于车辆诱导的交通灯动态配时优化算法研究

发布时间：2018-06-17 04:16

本文选题：交通灯控制 + 车辆诱导　；参考：《沈阳理工大学》2017年硕士论文

【摘要】：随着我国城市现代化进程的不断推进,交通问题成为影响社会发展的一个大问题。其中,交通拥堵是最为常见并影响较大的交通问题,国内外许多学者致力于交通拥堵问题的研究并提出了相应的解决方案。其中,智能交通系统是一种有效解决交通问题的智能系统。在智能交通系统的重要研究中,自适应交通灯控制系统是目前公认的缓解城市交通拥堵的有效途径。由于城市交通系统的复杂性和不确定性,现有的交通灯定时信号控制系统不能很好解决交通拥堵问题。为此,本文以基于最短路径策略的车辆诱导系统为基础,利用善于与环境交互的强化学习算法来建立智能交通控制策略。首先,我们设计基于Q学习的交通灯控制策略对交通信号灯进行动态配时,以减少车辆在交叉口的平均等待时间。其次,从协同优化的角度出发,提出基于模糊Q学习的交通灯控制策略,利用模糊逻辑控制根据车辆诱导信息获取协同交叉口的信息对Q学习的动作选择进行优化,以提高Q学习算法的收敛速度。最后,为了提高智能交通系统的整体性能,提出基于Sarsa学习的车辆诱导和基于Q学习的交通灯控制协同策略,实现两个系统在数据处理、策略实施和信息产生等方面协同,更好的提升交通系统的性能。本文以交通灯控制算法为基础,将自适应交通灯控制系统、强化学习、模糊逻辑控制优化动作选择策略、车辆诱导系统的性能提升融合在交通灯控制算法中,尤其是把强化学习的自学习特性应用到动态交通系统中。实验结果表明,基于Q学习的交通灯控制策略缩减了交通系统中车辆在交叉口的平均等待时间,减少了交通系统的拥堵现象,提升了交通系统的性能。并且,以该控制策略为基础,分别从强化学习算法的收敛速度和系统整体性能的角度进行改进。实验结果表明,改进策略进一步提升了交通系统的性能。
[Abstract]:With the development of urban modernization in China, traffic problem has become a major problem affecting social development. Among them, traffic congestion is the most common and influential traffic problems. Many scholars at home and abroad have devoted themselves to the research of traffic congestion and put forward corresponding solutions. Among them, Intelligent Transportation system (its) is an effective intelligent system to solve traffic problems. In the important research of intelligent transportation system, adaptive traffic light control system is recognized as an effective way to alleviate urban traffic congestion. Due to the complexity and uncertainty of urban traffic system, the existing traffic light timing signal control system can not solve the traffic congestion problem well. Therefore, based on the vehicle guidance system based on the shortest path strategy, the intelligent traffic control strategy is established by using the reinforcement learning algorithm which is good at interacting with the environment. First of all, we design the traffic light control strategy based on Q learning to dynamically match the traffic signal to reduce the average waiting time of the vehicle at the intersection. Secondly, a traffic light control strategy based on fuzzy Q learning is proposed from the view of collaborative optimization. The action selection of Q learning is optimized by using fuzzy logic control to obtain the information of collaborative intersection according to the vehicle guidance information. In order to improve the convergence speed of Q learning algorithm. Finally, in order to improve the overall performance of the intelligent transportation system, a traffic light control coordination strategy based on Sarsa learning and Q-learning is proposed. The two systems can cooperate in data processing, strategy implementation and information generation. Better improve the performance of the transportation system. Based on the traffic light control algorithm, this paper integrates the adaptive traffic light control system, reinforcement learning, fuzzy logic control optimization action selection strategy and vehicle guidance system performance enhancement into the traffic light control algorithm. Especially, the self-learning characteristic of reinforcement learning is applied to dynamic traffic system. The experimental results show that the traffic light control strategy based on Q learning reduces the average waiting time of the vehicle at the intersection in the traffic system, reduces the congestion of the traffic system, and improves the performance of the traffic system. On the basis of the control strategy, the convergence rate of reinforcement learning algorithm and the overall performance of the system are improved respectively. The experimental results show that the improved strategy can further improve the performance of the transportation system.
【学位授予单位】：沈阳理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491.54

【参考文献】