Iterative Learning Scheme for a Redundant Musculoskeletal Arm: Task Space Learning with Joint and Muscle Redundancies

This paper proposes an iterative learning control scheme in a task space for a musculoskeletal redundant planar arm model to accomplish a desired time dependent trajectory tracking task. In our previous work, we have proposed the iterative learning control scheme in a muscle length space for a two-link six-muscle planar arm model. This proposed method has been effective for performing a time dependent desired trajectory tracking task even under the existence of strong nonlinearity of muscles dynamics. However in the previous work, a muscle redundancy only treated, and a joint redundancy has not yet been considered. Also a solution of inverse kinematics from the task space to the joint angle space must be calculated in real-time. This paper considers both muscle and joint redundancies, and the task space iterative learning scheme is newly exploited. By introducing the task space controller, it is unnecessary to compute inverse kinematics from the task space to the joint space in real-time. Firstly, a three-joint nine-muscle redundant planar arm is modeled. Secondly, the task space iterative learning control signal is designed. Then finally, the effectiveness of our proposed controller is illustrated through some numerical simulation results even under the existence of both redundancies and the nonlinear muscle dynamics.