Systems and methods for reinforcement learning control of a powered prosthesis

Abstract

Systems and methods for tuning a powered prosthesis are described herein. A system includes a powered prosthesis including a joint, a motor mechanically coupled to the joint, a plurality of sensors, a finite state machine, and an impedance controller. The sensors are configured to measure a plurality of gait parameters, and the finite state machine is configured to determine a gait cycle state. The impedance controller is configured to output a control signal for adjusting a torque of the motor, where the torque is adjusted as a function of the measured gait parameters and a plurality of impedance control parameters, and where the impedance control parameters are dependent on the gait cycle state. The system also includes a reinforcement learning controller operably connected to the powered prosthesis. The reinforcement learning controller is configured to tune the impedance control parameters to achieve a target gait characteristic using a training data set.

Dependencies are inferred from the claim text.