Obstacle Detection And Self-localization For Mobile Robots By Using Laser Range Finders

In recent years people have seen an increasing interests on the research of mobile robots, a intelligent machine which is capable of performing some disgusting, risky or monotonous tasks historically assigned to human beings. A simple classification can be considered with the different working environments: indoor mobile robots and outdoor mobile robots. The focus of this dissertation can be divided into two parts: obstacle detection for Autonomous Land Vehicle (ALV) and self-localization for indoor mobile robots.On the first subject of this dissertation, a concise review of the relationship between range perception sensors and obstacle detection methods was presented. Comparing results show that Laser Range Finder (LRF) is a preferred sensor for obstacle detection. A detailed geometric analysis of obstacle detection system for ALV was presented by considering the safety requirements of ALV. The ranging error caused by various environmental factors and inevitable vibration of ALV was discussed as well. All of the above analysis show that although 2D LRF is not capable of scan in 3-dimension, when considering the flat road, its high sampling rate can do great good to the detection of even small obstacles. Then the two 2D LRF based obstacle detection system was carefully designed and consequently the algorithm based on multi-sensor data fusion was presented.In the algorithm, the theory of Multi-Targets Tracking (MTT) was induced to recognize and tracking obstacles. Due to the two different tasks of obstacle detection: recognition of obstacles from roadsides and accurate positioning the obstacles, the Dempster-Shafer evidence theory based identification and extend Kalman filtering based target tracking technique were adopted respectively. The algorithm features full utilization of different installation positions and different angular resolutions of the two LRFs by assigning different tasks to them. The top LRF was used to set up the initial trace of new obstacles recognized from the result of D-S evidence theory based multi-period data fusion and the bottom one was used to maintain tracking of obstacles, providing relatively accurate positions of the obstacles. In the hardware solution, a dual-port high speed communication card was designed to resolve the "data' losing" problem in the communication between the LRF and PC.On the second subject of this dissertation, we focus on the localization problem of the indoor mobile robots. A new approach for initial localization of the mobile robots, namely Complete Line Segments (CLS) based localization was proposed. The definitions of CLS as well as its properties and decision rules were given. The experimental results show that this method has much higher computational efficiency and better localization accuracy. Then the traditional position tracking method was improved by considering the sensor error spreading during positioning. Sub-centimeter position accuracy was achieved in the experiment. By combining the CLS method and the position tracking, robust global localization was realized for indoor mobile robots.Finally, as a creative application of the proposed localization techniques, a mobile robot based 3D scene reconstruction system was proposed. The platform avoids the most complex "registration" which is necessary for most 3D scene reconstruction systems and therefore an optimistic future isexpected.