TVSLAM: An Efficient Topological-Vector Based SLAM Algorithm for Home Cleaning Robots

Abstract

The Simultaneous Localization and Mapping problem limits the promotion of home cleaning robots in practical domestic environments. In this paper, a novel topological-vector based simultaneous localization and mapping (TVSLAM) algorithm is proposed to solve the problem. The algorithm involves four aspects. First, the ultra-wideband localization and dead reckoning localization are selected to develop a new combined localization algorithm which can improve the localization accuracy. In addition, a data acquisition algorithm which simplifies the process of data collection and demands much smaller memory size is proposed. Furthermore, a partitioning algorithm is developed to adapt to the various change rates of different rooms. Finally, an autonomous learning algorithm based on the regular and repetitive cleaning task is put forward. It makes the constructed map approach to the real environment with the increase of cleaning times. Overall, a novel topological-vector map is generated according to the above process of the algorithm. Simulation results show that the TVSLAM is an efficient and robust localization and mapping algorithm.