A Local Map Construction Method for SLAM Problem Based on DBSCAN Clustering Algorithm

Abstract

The accuracy of the local map of the robot is crucial to the accuracy of the real-time positioning of the robot and the construction of the global map. However, the current local map construction methods have problems such as low accuracy and long calculation time, which lead to the result that the robot cannot locate in real time and the global map construction is not accurate. Therefore, this paper proposes a DBSCAN (Density-Based Spatial Clustering of Applications with Noise) method for local map construction in SLAM problem. According to this algorithm, firstly, the pre-processed feature points are subjected to density clustering algorithm for large-area division. Then through a dynamic region re-division method based on two influencing factors of angle and distance, finally, the least squares method is used to fit the point set in each divided region. The experimental results show that, according to the data set in this paper, when the number of clustering is eps = 6, then in the dynamic area subdivision method based on Angle and distance, then make θ = 5%θi, d = 5%di, Finally, the point set is straight-line fitted by the least squares method to obtain the most accurate local map.