Modified bug algorithm with supervisory control using machine learning to reduce human-cobot collisions

A collaborative robot, or cobot, is an efficient robot that can work alongside human workers in the same work cell. Cobots are popular in industry because they allow close collaboration between humans and robots. Although a cobot can stop when a collision occurs, it is not safe or pleasant for the human worker to be hit by the cobot multiple times a day. Therefore, it is important to reduce or eliminate collisions between cobots and human workers. This research uses inexpensive proximity sensors on the cobot to detect obstacles and generate motion decisions. First, a modified bug algorithm is utilized to maneuver the cobot to achieve a pick-and-place task. This algorithm applies a wrist yaw angle to move around the human workers and avoid collisions in the same work cell. Second, a long short-term memory (LSTM)-based artificial neural network was trained using a collision-free path dataset. The network model generates corrections when abnormal movements in the modified algorithm are detected due to interference in sensor readings