Research On SLAM, Localization And Navigation Of An AGV Robot Based On Laser Sensors

SLAM,localization and navigation technology are one of the core technologies behind robot intelligence.This thesis focuses on the research of SLAM,localization and navigation of an AGV robot platform in unknown environments.The Turtlebot is used as the experimental AGV platform in this thesis while the RPLIDAR laser sensor is used as the external environmental sensor.The encoder(odometer)of the robot driving wheel is exploited for the purpose of internal sensing.The experimental framework is built on the Robot Operating System(ROS)platform,and multiple SLAM and navigation algorithms are simulated and verified.The main contributions of this thesis is as follows:Firstly,the hardware and software platform of the AGV robot platform based on the ROS framework are introduced.The system model of robot is analyzed,including coordinate system modelling,motion modelling and laser sensor modelling.The theoretical background of SLAM is expounded and the particle filter is verified through Matlab simulation.The algorithm is shown to achieve superior robustness and accuracy compared with the extended Kalman filter algorithm,and can better adapt to nonlinear systems.This shows that the algorithm is more suitable for robot positioning.Secondly,the SLAM method of the AGV platform is studied,and the FastSLAM method based on the Rao-Blackwellized particle filter and the Hector SLAM method based on scan matching are studied.In view of the shortcomings of these algorithms,this thesis proposes an improved RBPF-SLAM method: based on the odometer,combined with RPLIDAR laser measurements and its uncertainty during updates and adjusts to the weights of the particles.The proposed algorithm also adopts adaptive resampling,applying updates.The particles are then used to perform position estimates.The laser SLAM experiments in the unstructured indoor room and long corridor environments shows that the improved RBPF-SLAM algorithm can effectively reduce the particle degradation problem,improve the positioning accuracy,and successfully construct a large-scale two-dimensional grid map.Finally,in view of the wheel slippage problem encountered during the navigation of the robot and the inaccuracy leading the robot to “drift”,this paper proposes an improved AMCL global positioning algorithm,which incorporates the point-to-line ICP(PLICP).Scanning matching positioning enables global relocation.When the positioning error of the odometer is large,the improved global positioning PLICP-AMCL is used as the positioning input of the navigation frame to realize the pose estimation during the robot navigation process.The experimental results show that the improved positioning and navigation framework enable the robot to adapt to inaccurate odometer conditions in the long-term positioning and navigation experiments,and successfully reaches the designated target point with high robustness.