基于快速MH变异的遗传粒子滤波的WSN目标跟踪

基于快速MH(Metropolis-Hastings)变异的遗传粒子滤波跟踪算法,提出在遗传粒子滤波算法上,加入遗传进化思想,利用快速移动粒子交叉次数和变异算子,同时与赌轮选择一起产生了一种新的遗传算法,更快地提取到反映目标概率特征的典型粒子;实验证明,基于快速MH变异的遗传重采样方法可以避免粒子退化,快速提高粒子的多样性,减小跟踪误差,减少算法的运行时间,实时跟踪性能得到了大大提高;全面比较后,MHGAPF算法计算时间最短.跟踪精度用均方根误差表示,与GAPF算法和PF算法相比较,MHGAPF算法位置和算法速度的中均方根误差最低(位置为0.031 3,0.027 0,速度0.020 21,0.010 2),其中PF算法的跟踪精度最低,MHGAPF跟踪精度最高,进一步表明MHGAPF算法具有良好的跟踪性能.

WSN Target Tracking Based on Fast MH Mutation and Genetic Particle Filter

Abstract: Based on the genetic particle filter tracking algorithm of fast MH mutation, it is proposed to add the genetic evolution idea to the genetic particle filter algorithm, using the number of fast-moving particle crossovers and mutation operators, and at the same time produce a new genetic algorithm together with the wheel selection. Extract the typical particles that reflect the target probability feature faster. Experiments show that the genetic resampling method based on rapid MH mutation can avoid particle degradation, quickly increase the diversity of particles, reduce tracking errors, reduce the running time of the algorithm, and greatly improve the real-time tracking performance. After comprehensive comparison, MHGAPF algorithm has the shortest computing time. The tracking accuracy is expressed by root mean square error. Compared with GAPF algorithm and PF algorithm, MHGAPF algorithm has the lowest root mean square error in position and speed (position is 0.0313, 0.0270, speed is 0.02021, 0.0102). Among them, PF algorithm has the lowest tracking accuracy and MHGAPF algorithm has the highest tracking accuracy, which further shows that MHGAPF algorithm has good tracking performance.