Distributed parallel processing of mobile robot PF-SLAM

Real-time property is a fundamental requirement for a practical robot system. For this purpose, this article proposes an implementation architecture of robot SLAM by adopting two parallel threads processing. Since the dominant factor which determines the computational complexity is the employed particle number, two distributed threads with different particle set size are executed simultaneously. Conventional PF-SLAM algorithm occupies one of threads, and the other thread which hires more particles is activated whenever robot has significant motion changes. Advantages of this presented idea are validated by experiment carried on Pioneer robot.