Iterative Learning Procedure with Reinforcement for High-Accuracy Force Tracking in Robotized Tasks

The paper focuses on industrial interaction roboticstasks, investigating a control approach involving multiples learninglevels for training the manipulator to execute a repetitive(partially) changeable task, accurately controlling the interaction.Based on compliance control, the proposed approach consists intwo main control levels: i) iterative friction learning compensationcontroller with reinforcement and ii) iterative force-trackinglearning controller with reinforcement. The learning algorithmsrelies on the iterative learning and reinforcement learning proceduresto automatize the controllers parameters tuning. Theproposed procedure has been applied to an automotive industrialassembly task. A standard industrial UR 10 Universal Robot hasbeen used, equipped by a compliant pneumatic gripper and aforce/torque sensor at the robot end-effector.