A Novel Approach For Collision Avoidance In Collaborative Robotics Application

Ina this research, 3D robotic vision is implemented for pick and place tasks which uses a real-time collision avoidance algorithm that incorporates obstacle recognition and avoidance in collaborative robots. In conventional method of pick and place operation by collaborative robot (cobot), if in case any obstacle comes in between predefined path then cobot abort its operation and stops at same position. To continue pick and place operation operator need to remove obstacle and restart then operation. Due to this operational cycle time will increase, efficiency of cobot reduced. To overcome this problem we have implemented 3D vision system with machine learning algorithm in cobot. For safe human-robot collaboration, 3D vision technology has been implemented for obstacle detection and avoidance in pick-and-place operations. The Cognex IS2800 smart camera is used to take standard and depth images in a designated workspace. Object recognition is done using a deep neural network (DNN) along with point cloud segmentation, 3D object-pose estimation, and accurately identifying obstacle locations. A machine learning-based algorithm is used for collision avoidance based on obstacle coordinates received from the camera module. Due to the machine learning algorithm, the cobot can dynamically modify its trajectory, guaranteeing safe operation during the pick-and-place operation. This paper demonstrates the algorithm's performance under different operating conditions which results into optimize cycle time, avoid collision with obstacle and efficiency increased. By using proposed methodology will be able to achieve 84% of original cycle time with obstacle detection vision system and machine learning algorithm. The Techman TM5 6-DOF robotic arm is used for the demonstration of practical experimentation of the proposed method in detecting obstacles and avoiding collisions during pick-and-place operation.