| Summary: | Vision is a useful information for robot manipulation tasks because through it, a large spectrum of details about the environment can be provided. In this thesis, by introducing a YOLO v3 object detection algorithm, we propose a vision based robotic manipulation method to handle different objects detection tasks in real time and output position data of target center points with a RealSense D435i RGB-D camera. The proposed method is very flexible to environment changing and has the fast reaction speed and small computational cost. Furthermore, in order to test the proposed method performance, through socket communication, we successfully integrate the method with a UR5-e robot as a robot work cell to perform a specific pick-and-place task, which can be assistant for physically challenged people. Simulation results are presented to illustrate the performance of the proposed method.
Keywords: Convolution Neural Network (CNN), YOLO, pick-and-place task, RGB-D
camera
|
|---|