Self-localisation of indoor mobile robots using multi-hypotheses and a matching algorithm

Abstract

This paper proposes a new, fast and computationally light weight methodology to pinpoint a robot in a structured scenario. The localisation algorithm performs a tracking routine to pinpoint the robot's pose as it moves in a known map, without the need for preparing the environment, with artificial landmarks or beacons. To perform such tracking routine, it is necessary to know the initial position of the vehicle. This paper describes the tracking routine and presents a solution to pinpoint that initial position in an autonomous way, using a multi-hypotheses strategy. This paper presents experimental results on the performance of the proposed method applied in two different scenarios: (1) in the Middle Size Soccer Robotic League (MSL), using artificial vision data from an omnidirectional robot and (2) in indoor environments using 3D data from a tilting Laser Range Finder of a differential drive robot (called RobVigil). This paper presents results comparing the proposed methodology and an Industrial Positioning System (the Sick NAV350), commonly used to locate Autonomous Guided Vehicles (AGVs) with a high degree of accuracy in industrial environments.