A novel laser-vision combined approach to indoor robot pose estimation

Abstract

The pose of a robot is an essential information for manipulating, scheduling and monitoring tasks. This paper presents a novel system architecture and algorithm for indoor robot pose estimation. Previous robot pose estimation systems have limitations in terms of computational cost, accuracy, price or safety. Therefore, we propose a cheap and easy-to-use solution which needs only one or two laser pointers mounted on the robot pointing vertically towards the ceiling plane and a fixed low-cost camera. The proposed system is suitable for applications that demand very high localization accuracy. It can be potentially used for health care and surveillance applications. The experimental results clearly show that the proposed method is able to achieve high speed and high precision. The average pose estimation error is 14 mm in an area of about 24 m2 and the algorithm runs at 27 Hz.