摘要
粒子滤波算法中存在粒子退化、多样性缺失以及粒子数自适应问题。针对上述问题,首先,用先验转移概率密度和观测似然概率密度的混合分布作为重要性密度函数,用退火参数调控两者的比例,根据自适应参数优化控制机制对退火参数进行优化。其次,采用JS距离对采样粒子数进行动态调控,增强算法实时性。最后,通过遗传变异方法对粒子集进行调整,在一定程度上保持粒子多样性。仿真结果表明:与基于KL距离采样的蒙特卡罗定位算法相比,改进算法的精度提高了63.48%,平均定位误差为19.051 cm,用时减少了48.92%,达到39.268 s,采样粒子数维持在80个,改进算法的有效性得到验证。
Problems like particle degeneration,Diversity Loss and adaptive particle number exist in particle filter algorithm.To solve these problems,firstly,the hybrid distribution of prior transfer probability density and observation likelihood probability density is taken as the importance density function,The mixed proportion of the two probability density functions is adjusted by annealing parameters,and the annealing parameters are dynamically optimized by using the adaptive parameter optimization control mechanism.Secondly,the JS distance is used to dynamically adjust the number of sampling particles to enhance the real-time performance of the algorithm.Finally,the genetic mutation operation was used to optimize the particles,which maintained the diversity of the particles to a certain extent.The simulation results show that compared with the Monte Carlo localization algorithm based on KL distance sampling,the accuracy of the improved algorithm is enhanced by 63.48%,the average positioning error is 19.051 cm,the time consumption is reduced by 48.92%,reaching 39.268 s,and the number of sampled particles is maintained at 80.The validity of the improved algorithm is verified.
作者
刘红林
凌有铸
陈孟元
LIU Honglin;LING Youzhu;CHEN Mengyuan(Anhui Key Laboratory of Electric Drive and Control,Anhui Polytechnic University,Wuhu 241000,China)
出处
《安徽工程大学学报》
CAS
2019年第4期56-62,共7页
Journal of Anhui Polytechnic University
关键词
粒子滤波
蒙特卡罗定位
混合提议分布
JS距离
遗传变异
particle filtering
monte carlo localization
hybrid proposal distribution
js distance
genetic variation