Research on Robot SLAM of RBPF Improved with Weight Accumulation

The research of mobile robots is progressing rapidly. The use of lidar to realize simultaneous positioning and mapping (SLAM) is a hot spot in the field of robotics research.. It is a hot spot in the field of robotics research. RBPF-SLAM is a SLAM algorithm based on particle filtering. It is an important algorithm in the field of laser SLAM. When creating a map, each particle represents a separate environmental map. During the iteration process, errors will continue to accumulate, causing serious particle degradation problems. The mapping result is far away from the true value. Although many algorithms that optimize particle resampling alleviate the problem of particle degradation, the process of resampling will cause loss of particle diversity and loss of mapping information. In response to this problem, this paper proposes an improved RBPF-SLAM algorithm for weight accumulation. The sampling area is equally divided by the set boundary, and then the continuously accumulated weight information is used to determine whether the cumulative weight breaks through the partition boundary as the judgment condition, and then the particles to be sampled are resampled in stages, and finally the new Monte Carlo method is used to construct a new Particles, rather than simply copying particles with significant weight. The weight accumulation improvement algorithm in this paper makes the RBPF-SLAM algorithm need not sample a large number of particles, alleviating the problem of particle degradation. After resampling, the particles can better reflect the posterior probability density function of the true state, which enhances the accuracy of the algorithm's estimation and can maintain the diversity of particles and improve the quality of mapping.The experimental results show that the method in this paper still improves the completeness of the mapping and reduces the computing time when the number of particles is reduced.