Reinforcement Learning Fractional Order PID Controller For Upper Limb Rehabilitation Robot

Nowadays, it has been demonstrated that individuals with motor dysfunction can benefit significantly from robot-assisted rehabilitation therapy in terms of their everyday behavior and upper limb motor performance. This study discussed the use of iTbot (intelligent therapeutic robot), a two-degrees-of-freedom (DoFs) end-effector type robot, to deliver therapy for upper limb rehabilitation. For this purpose, a Reinforcement Learning (RL) agent is trained using the Deep Deterministic Policy Gradient (DDPG) algorithm to tune a Fraction order-PID (FOPID) control parameters. The suggested method aims to enable the robot to monitor any therapeutic exercise trajectory without being aware of the controller parameters. Moreover, the joint angles are always within a safe range to ensure the patient’s safety and the efficacy of the therapy exercise. The obtained results demonstrate the efficiency and accuracy of the proposed approach in both learning stability and trajectory tracking competency.