一种改进天牛群算法优化的RBPF-SLAM模型研究

Research on RBPF-SLAM model optimized by improved beetle swarm algorithm

针对传统基于粒子滤波器的Fastslam(RBPF-SLAM)算法由于粒子退化及粒子多样性缺失,造成定位精度低和建图一致性差的问题,该文提出了一种基于改进天牛群算法优化的RBPF-SLAM算法。该算法将天牛个体视为粒子,利用天牛觅食特性使其不断向高似然区域移动;在天牛群算法中引入差分进化策略,对迭代产生的天牛群体进行变异、交叉、选择,以增强群体多样性。实验结果表明,改进算法较RBPF算法在状态跟踪上精度提高了50%以上。在地图构建上,改进算法仅用RBPF-SLAM算法一半粒子数,就能得到精度更高的地图。

Due to particle degradation and the loss of particle diversity,the traditional Rao-Blackwellized particle filter(RBPF-SLAM)algorithm has the problems of low positioning accuracy and poor consistency in mapping.A RBPF-SLAM algorithm optimized by the improved beetle swarm algorithm was proposed.In this algorithm,the individual beetle was regarded as a particle,and its foraging characteristics were used to make it move to the high likelihood region;the differential evolution strategy was introduced into the beetle population algorithm to mutate,cross and select the iterated beetle population to enhance the population diversity.Experimental results showed that the state tracking accuracy of the improved algorithm was more than 5o%of RBPF algorithm.In map construction,the improved algorithm only used half of the particle number of RBPF-SLAM algorithm to get a map with higher accuracy.