A Navigation Method Using the Mutual Feedback of Waypoints and Self-Positions

Abstract

The purpose of this article is to solve a significant problem faced by waypoint-based robot navigation: the positions of the waypoints can be located in unreachable areas due to errors in self-localization, the map, or in the waypoint data. Much research has focused on self-localization, but few studies have focused on estimating waypoints. The method presented in this paper estimates waypoints and self-positions using information from Light Detection and Ranging (LIDAR) processed using particle filters. If waypoints are placed in unreachable areas, the weights of the particles around the waypoint are decreased according to a geometric constraint, and the weights of other particles in reachable areas are increased. Because the waypoint is the expectation value of all particles, this process relocates the waypoint at a reachable location. Another characteristic of this method is the mutual feedback between the probability density functions (PDFs) of the waypoint and the self-position. This method first alters the PDFs of the self-positions using the PDFs of the waypoints, and then changes the PDFs of the waypoints using the PDFs of the self-positions, thus implementing mutual feedback and improving the accuracy of self-localization and the robustness of the waypoint navigation. This method was used for a trial run of the Tsukuba Challenge 2010, and a 240 m run was successfully finished using only dead reckoning information. Other experiments show that this simple, unique method is very useful in waypoint-based robot navigation.