2D-Laser-Based Odometer Construction And Autonomous Navigation For Service Robots

In recent years, service robots are gradually going mass market, and also constantly changing people's daily life style. Autonomous navigation is the premise of the service robot to complete other complex tasks. In this paper, the problem of autonomous navigation based on 2D laser ranging for a service robot in unknown environment is studied including pose tracking, autonomous localization, grid map construction and path planning.In order to track the pose of the mobile robot, the comparative analysis of three current typical scan matching algorithms is conducted, namely Mb-ICP, PSM and PL-ICP. To reduce the impact of raw laser data towards the algorithm, median filtering and segmentation is conducted respectively, which get rid of disorderly and invalid data and improve robustness of the algorithm. This paper takes advantage of key frame technique to tackle pose drift owing to dynamic error of laser data. Based on Rao-Blackwellized particle filter, the problem of robot current localization and mapping (SLAM) is decomposed to robot's pose estimation and grid map building. Encoder odometry is substituted by laser odometry to improve the proposed distribution without losing algorithm accuracy. Resample decided by dispersion of the importance weights avoids particle impoverishment. Under the assumption that the state of individual grid is independent with each other, grid map construction is converted to estimating posterior of every grid.Considering the influence of the robot's size on the actual navigation, the static obstacles and the dynamic obstacles detected in real time are inflated according to size of the robot. Based on cost map and original pose and goal, A* algorithm is employed to conduct global path planning, which generates a path form original pose to goal position. Based on dynamic window approach, an optimal local path is selected in velocity space. The search space is reduced by incorporating dynamics of the robot and a final path is obtained according to the evaluation function.In large scale environment, Service robot autonomous navigation system is built based on Pioneer 3 mobile robot within ROS software framework. Experiments conducted in different scene prove the practicability and validity of the algorithm.