An improved FastSLAM algorithm based on stream computing

Simultaneous localization and mapping (SLAM) is a key problem for mobile robots to realize autonomous exploration. SLAM is a typical computing intensive task, and has high computing requirements for mobile robots. To improve the efficiency and accuracy for SLAM process, a high real-time computing mode based on stream computing is built in the cloud edge. In this mode, the distributed parallel SLAM is carried out. Based on 5G communication, Kafka component is used for message transmission between the cloud edge and mobile robot. In the cloud edge, the SLAM process is divided into four steps, data reading, particle updating, particle sampling and result pushing. Comparing the traditional FastSLAM algorithm, the proposed algorithm has the better estimation quality and shorter execution time. Experiments verify the feasibility and effectiveness of this algorithm.