Designing a Control Architecture for the Knee Prosthesis Using ANFIS and Supervisory PI Controller

Abstract

this paper presents a control architecture for knee joint actuator in one gait cycle. The control architecture includes Fuzzy subsystem that is designed. Recursive Least Squares (RLS) and Adaptive Network-based Fuzzy Inference System (ANFIS) methods are applied and compared to choose the best method for identification of the fuzzy system. After that PI and Supervisory PI controllers are utilized to achieve the desired force. The coefficients of PI supervisory controller are designed with suitable and compatible membership functions for the best results. External disturbance is exerted to the output of our closed-loop control system to simulate uneven surface for walking. Results indicate that the robustness of PI and PI supervisory controller is acceptable.