Real-Time Self-Localization Method by Using Measurements of Directions of Two Landmarks and Dead Reckoning

Abstract

We propose a new real-time self-localization method for an autonomous mobile robot amidst dynamically moving multiple obstacles. This method uses direction of two landmarks and dead reckoning. Conventional method uniquely determines the pose of the robot if at least three landmarks are observed. However, robots often fail to simultaneously recognize three different landmarks in such environment: landmarks are easily occluded. When a robot detects two landmarks and measures angles of their directions, the robot is constrained to be on a circle, so we obtain a set of candidates of robot's pose. Successively the robot moves for a short period, and then observes landmarks. Displacement obtained by dead reckoning is added to pose candidates. Localization error of candidates is evaluated by difference of direction angle to the landmarks: comparison is made between measured direction of landmarks and computed ones. The robot's motion and measurement of landmarks is repeated, and the evaluation of candidates is updated. Finally correct pose is determined uniquely as the one having the smallest error accumulation. Multiple localization process in parallel results robust and accurate localization. The proposed method applies to the soccer robot in the RoboCup Middle-size league and experimental results indicate that the approach is reliable.