基于智能优化箱粒子滤波的移动机器人FastSLAM

针对传统FastSLAM算法需要大量粒子构建地图导致计算复杂度高、难以提高估计精度等问题，研究构建了一种基于智能优化箱粒子滤波（IOBPF）的移动机器人FastSLAM算法。首先，将萤火虫算法（FA）的动态寻优机制引入箱粒子滤波（BPF），建立了箱粒子的荧光亮度更新公式、吸引度计算公式和位置更新公式，使箱粒子集智能化地向高似然区域移动，避免了箱粒子的退化现象。然后，以改进的智能优化箱粒子滤波进行机器人位姿估计，并采用扩展区间卡尔曼滤波（EIKF）完成地图的构建和更新。移动机器人的模型仿真和实体实验结果表明:所提智能化FastSLAM算法可有效提升箱粒子的性能，并降低地图构建所需粒子数，从而显著提高FastSLAM的定位精度和地图构建的鲁棒性。 

FastSLAM for mobile robot based on box particle filter with intelligence optimization

The traditional FastSLAM algorithm requires a large number of particles to build the map, thus resulting in high computational complexity and difficulty in improving the estimation accuracy. In view of these problems, an algorithm of FastSLAM for mobile robot is presented based on box particle filter with intelligence optimization (IOBPF). First, the dynamic optimization mechanism of firefly algorithm (FA) is applied to the box particle filter (BPF), and the formulas of fluorescence brightness updating, attraction calculation and position updating of box particle are constructed, which makes the box particles move toward the high-likelihood region intelligently and avoid the phenomenon of box particle degeneracy. Then, the improved BPF with intelligence optimization is utilized to estimate the pose of robot, and the extended interval Kalman filter (EIKF) is employed to complete the map building and updating. The results of model simulation and entity experiment of mobile robot show that the intelligent FastSLAM algorithm in this paper can effectively improve the performance of box particles and reduce the number of particles required for map construction, thus significantly improving the positioning accuracy and robustness of map construction.