Research On Mobile Robot Self-localization Algorithm Fusion With Laser And RGB-D Camera

With the growing development of computer science and sensor technology, the robotics which increasingly tend toward intelligent and autonomous start into the family for the human services. The robot’s self-positioning, building a map and real-time path planning lie at the core of an intelligent and autonomic robot. Meanwhile, being capable of accurately positioning, the robot may probably then navigate autonomously and formulate a useful map. Consequently, to improve the positioning accuracy of a robot by developing new methods is an urgent problem, which means significantly.At present, many scholars have studied the robot position both at home and abroad, most of them leverage sensors such as odometer, laser range finder, GPS and inertial sensor. Recently, visual sensor based positioning methods which employs algorithms in computer vision and image processing technology has been developed and works well in practical applications. The Kinect is one of the Microsoft products in 2010, which is a RGB-D camera and has been well applied in the field of robots. The recent and most popular positioning algorithm is RGB-D SLAM(Simultaneous Localization and Mapping), but it depend largely on image information and has a big matching error under some special scene, additionally the positioning error can not be updated even eliminate it.In this thesis, we studied two traditional positioning methods both in theory and practical experiments in order to improve the accuracy of robot self-localization. We propose a novel positioning algorithm which fusing the information of both laser ranging and RGB-D camera visual orientation by using a dynamic weighting factor at indoor environment, which experimentally improves the positioning accuracy of robot. The main contents in this thesis are illustrated as follows.1. Models of three robots sensors that odometer, laser rangefinder and RGB-D camera are formulated respectively, and then the sources of their self-positioning errors are analyzed.2. The EKF algorithm which utilized the data of odometer and laser ranging is used to evaluate the position of a robot. Both the advantages and disadvantages of this algorithm are analyzed based on experimental results of MATLAB simulation and actual robot movement test.3. The current popular robot positioning algorithm, RGB-D SLAM algorithm is studied, based on which the improved algorithm is proposed and it has been used in the robot localization.4. The position of a mobile robot is localize accurately by using an algorithm that fusing the location result of laser ranging and the result of RGB-D with a dynamic weight factor. And the laboratory robot self-localization software system is established, and experimental verification is conducted with a three-wheel full direction family service robot of our laboratory. Experimental results demonstrated that the proposed fusing method with a dynamic weight factor has improved the position accuracy of a robot.