Reinforcement Learning Human Inverse Kinematics of an Upper Limb Exoskeleton Robot
- Resource Type
- Conference
- Authors
- Abdallah, Mahmoud; Saad, Maarouf; Fareh, Raouf; Kali, Yassine; Bettayeb, Maamar
- Source
- 2023 Advances in Science and Engineering Technology International Conferences (ASET) Advances in Science and Engineering Technology International Conferences (ASET), 2023. :1-6 Feb, 2023
- Subject
- General Topics for Engineers
Training
Simulation
Exoskeletons
Estimation
Reinforcement learning
Kinematics
Trajectory
Inverse Kinematics
Exoskeleton
Rehabilitation
Reinforcement Learning
Upper Limb
- Language
- ISSN
- 2831-6878
This paper presents a Reinforcement Learning (RL)-based swivel angle estimation for an upper-limb 7-DoF exoskeleton robot. Choosing the best swivel angle in the rehabilitation process helps in ensuring the safety and the comfortability of the patient's upper-extremity movement. Also, fixing the swivel angle will result in having one unique IK solution. The RL agent is trained to estimate the swivel angle that minimizes the discomfort index of the patient's upper limb. After training the agent, it is used to simulate a workspace trajectory to validate the IK solution with the fixed swivel angle value. Simulation results are showing the feasibility and effectiveness of using RL in solving the IK for 7-DoF exoskeleton robots.