基于禁忌免疫及权值选择的粒子滤波算法的设计与实现

发布时间：2018-08-14 15:17

【摘要】：随着系统规模变大,系统的复杂度不断增强,原有的粒子滤波算法在系统参数、状态估计以及目标跟踪方面存在不足,因此,设计新的粒子滤波算法,提高估计精度,减少计算复杂度显得至关重要。本文针对粒子滤波算法进行改进,主要工作如下： 1.介绍了粒子滤波算法的基本原理以及标准粒子滤波算法的计算流程,同时分析了其存在的主要问题,对并本文改进算法中要用到的一些智能算法如：禁忌搜索、人工免疫以及权值选择算法予以说明。 2.在粒子滤波算法基础上,针对粒子退化,样本集多样性低的问题,设计出基于禁忌免疫的粒子滤波算法。该算法利用人工免疫算法的寻优能力从众多粒子中挑选好的粒子,提高了样本集的多样性,并且通过禁忌搜索回避搜索陷入局部最优。利用该算法估计系统的参数和状态,并与人工免疫粒子滤波、标准粒子滤波算法进行对比,验证算法的估计性能。 3.针对粒子滤波算法计算复杂度高的问题,提出了基于权值选择的边缘化粒子滤波算法。该算法通过利其模型中的线性子结构降低从标准粒子滤波算法中得到的估计方差,并能边缘化处理相应的线性状态变量,同时能利用最优线性滤波进行估计,从而降低了计算量。并且,粒子间的相互独立性使得粒子集包含更多相异的粒子路径,提升粒子集的多样性,具有较好的优化效果。 4.以城市轨道列车制动模型为背景,将提出的两种改进算法用于列车制动率以及列车运行状态的联合估计,对两种改进算法进行了仿真对比。
[Abstract]:With the scale of the system becoming larger and the complexity of the system increasing, the original particle filter algorithm has some shortcomings in system parameters, state estimation and target tracking. Therefore, a new particle filter algorithm is designed to improve the estimation accuracy. It is very important to reduce computational complexity. In this paper, the particle filter algorithm is improved, the main work is as follows: 1. This paper introduces the basic principle of particle filter algorithm and the calculation flow of standard particle filter algorithm, analyzes its main problems, and improves some intelligent algorithms used in the algorithm, such as Tabu search, Tabu search, etc. Artificial immune and weight selection algorithm to explain. 2. On the basis of particle filter algorithm, aiming at the problem of particle degradation and low diversity of sample set, a particle filter algorithm based on Tabu immunity is designed. The algorithm uses the optimization ability of artificial immune algorithm to select good particles from many particles, which improves the diversity of sample set and falls into local optimum through Tabu search avoidance search. The algorithm is used to estimate the parameters and states of the system, and compared with the artificial immune particle filter and the standard particle filter algorithm, the estimation performance of the algorithm is verified. Aiming at the problem of high computational complexity of particle filter algorithm, an edge particle filter algorithm based on weight selection is proposed. The algorithm can reduce the estimated variance obtained from the standard particle filter algorithm by using the linear substructure in the model, and can marginalize the corresponding linear state variables, and at the same time, it can use the optimal linear filter to estimate. Thus, the calculation amount is reduced. Moreover, the mutual independence of particles makes the particle set contain more different particle paths, improve the diversity of particle sets, and have a better optimization effect. 4. Based on the braking model of urban rail train, the two improved algorithms are applied to the joint estimation of train braking rate and train running state, and the two improved algorithms are simulated and compared.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN713;TP18

【参考文献】