An improvement of self-localization for omnidirectional mobile robots using a new odometry sensor and omnidirectional vision

Abstract

The paper describes a new method to reduce the positioning error of 3-degree-of-freedom (3Dof) mobile robots. The 3Dof robot uses 3 omni-wheels as the motion system. A combination of a modified odometry system and a vision-based self-localization algorithm and a two-channel complementary filtering algorithm is used. The low frequency stability of the vision sensors with the high frequency tracking of the odometry sensor leads us to achieve both dynamic and static stability of positioning. Two new algorithms are introduced in order to calculate the position and orientation of the robot using the results from the odometry sensors. An omni-directional camera provides vision data for the self-localization algorithm, which calculates the robot position according to the angles under which the known obstacles around the robot are viewed. In order to check the improvement of the proposed positioning system compared to existing systems, a test condition is suggested to measure the position and orientation errors. The results of this test show that an improvement of 80% in position error and 86% in orientation error is achieved.