基于鸟群算法的交通信号控制

发布时间：2018-05-21 06:02

本文选题：多目标控制 + 鸟群算法　；参考：《中国科学院大学（工程管理与信息技术学院）》2014年硕士论文

【摘要】：交通信号控制系统是智能交通系统的基础子系统,能够协调控制区域内交通信号灯的配时方案,均衡路网内交通流运行,充分发挥道路系统的交通效益。然而,目前交通信号控制方法单一,灵活性差,不能有效缓解城市复杂路网的交通问题。因此有必要优化控制算法,找到具有针对性的交通信号控制解决方案。线性加权法、目标规划、约束方法等传统的多目标控制算法将各种客观功能集成到一个单一的目标函数,通过决策者或优化方法本身设定的系数的值自适应调整。这些传统方法虽然简单容易实现,但由于多目标控制问题的目标函数可能非线性、连续或不可微,需要事先充分掌握先验对优化问题的知识。因此,这些传统方法往往无法解决更复杂的多目标控制问题。而相对于控制系统的传统优化算法,进化算法是一种模仿生物自然选择和进化过程的随机搜索算法,更适用于处理多目标控制的实际问题。同时,由于鸟群算法思想新颖,且在智能交通控制方面的应用研究相对较少,对于高维的复杂问题,鸟群算法可以在尽可能降低计算量的同时保证较为理想的收敛结果,既克服了基于梯度的算法不易跳出局部最优解的问题,又克服了穷举法计算量过于巨大的缺点。因此,通过对已有算法进行适当改进后用于智能交通控制,可以取得突破性的进展。本文主要采用鸟群算法进行寻优,首先,通过研究层次分析法和鸟群算法的相关理论,重点研究鸟群算法的基本原理、数学模型、参数分析从而提出了针对交通控制系统参数寻优的改进办法；同时对鸟群算法的应用进行分析,通过对已有相关理论的研究对比,进一步加深对该算法的认识。其次,在研究鸟群算法基本理论的基础上,对基本鸟群算法进行改进,期望能够避免算法早熟收敛的问题,使其性能在基本鸟群算法基础上能有明显提高。最后,通过使用VISSIM交通仿真软件,完成路网模型的绘制和交通仿真参数的设置,并将优化改进后的鸟群算法用于交通信号控制的仿真实验,以验证在多目标期望下优化后的鸟群算法在控制效果方面的效果。结合以上研究和试验工作,本文引入多目标控制思想,将层次分析法和鸟群寻优算法相结合,用于解决交通信号灯控制问题：通过采用多目标方法,对指标层参数进行控制加权,得到不同目标下的交通通畅程度的评价函数；进而使用鸟群算法,以可以接受的速度和准确度优化出交通灯信号参数。层次分析法的引入使得评价函数更加合理有效,而鸟群算法的运用使得在保证寻优结果可靠性的同时极大地减小计算量。总体来说,本文主要完成了以下几个方面工作： 1、学习并研究了层次分析法及鸟群算法的相关理论,重点研究鸟群算法基本原理、数学模型、参数分析及其改进办法。对鸟群算法的应用进行综述,通过对已有相关理论的研究对比,进一步加深对该算法的认识。 2、在以上学习研究的基础上,改进基本鸟群算法,较为有效的避免算法早熟收敛的问题,使其性能比基本鸟群算法有明显提高。 3、学习使用VISSIM交通仿真软件,完成路网模型的绘制和交通仿真参数的设置。 4、论文的最后将优化改进后的鸟群算法用于交通信号控制的仿真实验,实验表明,在多目标期望下优化鸟群算法在控制效果方面有不错的表现。
[Abstract]:The traffic signal control system is the basic subsystem of the intelligent traffic system. It can coordinate the timing scheme of traffic signal in the region, balance the traffic flow in the road network and give full play to the traffic efficiency of the road system. However, the traffic signal control method is single and the flexibility is poor, and it can not effectively alleviate the traffic problems of the city complex road network. Therefore, it is necessary to optimize the control algorithm and find a targeted solution for traffic signal control.
The traditional Multiobjective Control Algorithms, such as linear weighting, target programming and constraint methods, integrate various objective functions into a single target function and adjust themselves by the values of the coefficients set by the decision-makers or optimization methods. These traditional methods are simple and easy to implement, but the objective function of the multiobjective control problem is possible. Nonlinear, continuous or non differentiable, it is necessary to fully grasp the knowledge of priori optimization problems in advance. Therefore, these traditional methods are often unable to solve more complex multi-objective control problems. Compared with the traditional optimization algorithms of control systems, evolutionary algorithm is a kind of random search algorithm that mimics biological natural selection and evolution process, and it is more applicable. At the same time, because of the novel idea of the bird swarm algorithm and less research on the application of intelligent traffic control, the bird swarm algorithm can reduce the computational complexity as much as possible while guaranteeing more ideal convergence results, which can not only overcome the gradient based algorithm, but also overcome the gradient algorithm. The problem of local optimal solution also overcomes the shortcomings of the immense computation of exhaustion method. Therefore, through the proper improvement of the existing algorithms for intelligent traffic control, breakthrough progress can be achieved.
This paper mainly uses the bird swarm algorithm to optimize. First, by studying the theory of AHP and the bird swarm algorithm, we focus on the basic principle, mathematical model and parameter analysis of the bird swarm algorithm, and propose an improved method for optimizing the parameters of the traffic control system. Secondly, on the basis of the basic theory of bird swarm algorithm, the basic bird swarm algorithm is improved to avoid the premature convergence of the algorithm, so that its performance can be obviously improved on the basis of the basic bird swarm algorithm. Finally, through the use of VISSIM traffic simulation The software has completed the drawing of road network model and the setting of traffic simulation parameters, and the improved bird swarm algorithm is applied to the simulation experiment of traffic signal control to verify the effect of the bird swarm optimization on the control effect after the multi target expectation optimization.
Combined with the above research and experiment work, this paper introduces multi-objective control thought, combines AHP and bird swarm optimization algorithm to solve the problem of traffic signal control: the evaluation function of traffic patency under different targets is obtained by using multi target method, and the evaluation function of traffic patency under different targets is obtained; and then the bird is used. The group algorithm optimizes the traffic light signal parameters with the acceptable speed and accuracy. The introduction of AHP makes the evaluation function more reasonable and effective, and the application of the bird group algorithm greatly reduces the computation while ensuring the reliability of the optimization results.
Generally speaking, this paper mainly completed the following aspects:
1, we study and study the related theories of AHP and bird swarm algorithm, and focus on the basic principle, mathematical model, parameter analysis and improvement methods of bird swarm algorithm. The application of bird swarm algorithm is summarized, and the understanding of the algorithm is further deepened through the comparison of the related theories.
2, on the basis of the study above, the basic bird swarm algorithm is improved, and the premature convergence of the algorithm is effectively avoided, and the performance of the algorithm is obviously improved than the basic bird swarm algorithm.
3, learn to use VISSIM traffic simulation software to complete the road network model drawing and traffic simulation parameters settings.
4, at the end of this paper, the improved bird swarm optimization algorithm is used to simulate the traffic signal control. The experiment shows that the optimal bird swarm algorithm has a good performance in the control effect under the multi target expectation.
【学位授予单位】：中国科学院大学（工程管理与信息技术学院）
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：U491.54

【参考文献】