Research And Implementation Of Indoor SLAM Algorithm Based On Multi-sensor Fusion

In recent years,autonomous mobile robots with the ability of navigation and planning have been widely applied in various fields.SLAM(Simultaneous Localization And Mapping)is one of the core technologies for autonomous mobile robots,which is a important guarant for autonomous navigation and successful execution of specific tasks.In dynamic and complex indoor application scenarios,such as offices and shopping malls,autonomous robots must accurately know their location and avoid obstacles to complete the assigned tasks.This paper proposes a multiple sensor fusion SLAM algorithm.It estimates the accurate pose of robot by the fusion of lidar,IMU,encoder information.And it constructs the probability grid map of indoor environment by point clouds' scan matching.At the same time,this algorithm adds the constraint of loop closure in the back-end when the loop closure appears,and eliminates all the cumulative errors of mapping by the optimization method.It leads to a environment map with good precision for a wheeled robot and a solution for global localization based on the map.The multi-sensor fusion algorithm proposed in this paper has the advantages of real-time,low-power consumption,robustness and it can obtain a good result on the embedded platform with limited resources.The main work of this paper is as follows:(1)The motion model for two-wheel differential driven robot and the measurement models of sensors are established.This paper combines IMU and the wheel's odometry data fusion by the extended kalman filter,it determines jacobians of the motion model with respect to the state of the robot and the noise covariance matrix in the prediction stage and determines the Jacobian matrix of measurement model with respect to the state of the robot in the update phase.(2)The model and theoretical analysis of SLAM in wheeled robot are established.This paper focuses on the analysis and elaboration of the SLAM problem of a wheeled indoor robot as the form of factor graph,and turns this SLAM problem into a least square optimization problem.Finally,the LM algorithm is selected to solve this optimization problem,and the process of the algorithm is given.(3)Based on correlation scan matching algorithm are studied,and the loop-closing is detected by the multi-resolution correlative scan matching.And then it calculates all constraints for SLAM problem and adds constraints to the back-end optimization problem.It determines the specific expression of residual function with each constraints and solves the least squares problem by the nonlinear optimization algorithm.(4)Experiments are designed to verify the performance and robustness of closed-loop mapping and global localization.In this paper,the accuracy average error of the mapping and the success rate and the time needed for the global localization are statistically analyzed.