Decentralized EIF-based Global Localization Using Dead-Reckoning, KINECT and Laser Scanning for Autonomous Omnidirectional Mobile Robot

This paper presents a global pose localization method using decentralized extended information filter (DEIF) for an autonomous omnidirectional mobile robot (AOMR) driven by omnidirectional Mecanum wheels in indoor environments by fusing measurements from one KINECT sensor, one laser scanner and four encoders mounted on the omnidirectional Mecanum wheels. The KINECT sensor is used to recognize two landmarks in the working environments of the robot, and then obtain the azimuth angles and distances between these two landmarks and the robot. Based on these measured information, a static global pose initialization algorithm is applied to estimate both start-up position and orientation of the robot. Once the initial pose is determined, a DEIF approach is proposed to fuse the measurements from the four encoders, Kinect and laser scanner, in order to continuously keep track of the accurate moving poses of the robot at slow speeds less than 100 cm/sec. Simulations are conducted to show that the proposed methods provides accurate estimation of both unknown initial and continuous moving poses of the robot.