Research On SLAM Technology Of Service Robot Based On Vision Sensor

With the development of robot industry being included in the national key research projects,SLAM technology as one of the core issues in the field of robotics,getting more and more attention.This paper focuses on the localization and navigation of indoor service robots.Firstly,the motion model of the robot,the mathematical model of depth camera and the model of environment are introduced,preparing for the rest of the paper.Then,a point cloud based visual odometry will be introduced.We predict the depth uncertainty of the camera using a Gaussian mixture model and obtain sparse 3D point cloud from the RGB image and depth image.The position and orientation transformation matrix is calculated by the ICP algorithm.Then a data fusion is designed using nonlinear extended Kalman filter(EKF)for sensor-data fusion of odometry,IMU and visual odometry.The simultaneous localization and mapping(SLAM)is designed in particle filter theory,using a large number of weighted particle assumptions to approximate the real pose of the robot and creating a occupancy grid map at the same time.Then,a mixed method of global planning and local planning is implemented based on the occupancy grid map to drive the robot form the current pose to the goal in a dynamic environment.At last,we build a robot hardware platform and coding to realize the algorithm.Experiments are conducted to verify the effectiveness of the algorithm and get good performance.