Determination of Angular Status and Dimensional Properties of Objects for Grasping with Robot Arm

In any task, robot arms can work more effectively without human control. With components such as imaging devices, it is possible to program robots to control autonomously. In this study, the problem of grasping moving objects with the robot arm is realized fully automatically. Deep learning-based You Only Look Once(YOLO) recognizes the objects moving at unknown speeds on a conveyor belt. The velocities of the detected objects are calculated by image processing methods with 2D camera frames. An Artificial Neural Network (ANN) was trained to output the angles required for the robot arm to grasp. According to the got values, the robot arm waits for the object to arrive and then realizes the grip. In the trials, the robot arm achieved successful gripping of over 93% without knowing the sizes, speeds, and positions of the objects.