Design and Implementation of AMR Robot Based on RGBD, VSLAM and SLAM

In recent years, robots have gradually entered life from the factory. Since robots that can move autonomously can improve the productivity of personnel, mobile robots are even more commonly used. With the advancement of hardware technology, improving the performance of LIDAR and depth cameras and reducing the required cost, robots can be equipped with these devices to make positioning and movement more accurate. This thesis implements an AMR robot on TurtleBot3 Burger by combining 2D LIDAR, RGBD camera and related safety devices. The data of 2D LIDAR is the input of SLAM and path planning, build 2D map and detect whether there are obstacles so that the robot can dodge; RGBD camera is used as the input of VSLAM, build 3D map, landmark database, show trajectory of the robot in 3D space and verify whether the robot has moved to the correct place. In the VSLAM algorithm, object detection and OCR technology are added to extract semantic data from the image. The 3D map and landmark database can provide users with more meaningful information. The landmark database can be used to compare the similarity of SURF features and SSIM with the images obtained when the robot is moving, to verify whether there is movement to the correct place, to improve the accuracy of the robot’s movement. The state lights, anti-falling, power detection and warning voice implemented by Arduino and related sensors are used as the safety device of the robot, so that the robot can make higher safety during operation.