Data-Driven Predictive Control of a Pneumatic Ankle Foot Orthosis

Abstract

We present the design and control of a pneumatic ankle-foot orthosis (P-AFO) device powered via bi-directional pneumatic rotary actuator and a pneumatic artificial muscle for rehabilitation assistance and treatment of neuromuscular disorders. The rotary actuator and the pneumatic muscle assist with dorsiflexion and plantar flexion, respectively. The prototype is also equipped with simple sensor system for gait pattern analysis. The P-AFO has the capability of 20 degrees dorsiflexion from the plantar flexion and 12 degrees dorsiflexion from the neutral position of an ankle joint. The data-driven predictive control (DDPC) algorithm has been designed for P-AFO to follow desired gait cycle trajectories while rectifying the nonlinearity and uncertainties of the pneumatic actuators. The design of DDPC is realized from the subspace identification matrices acquired by the input-output values obtained as a result of an open-loop operation. The control structure is completely data-based without certain use of a model in the control implementation. In order to control the developed P-AFO prototype device, the suggested controller was implemented in a real-time operating system. Experimental studies are performed to compare the proposed controller with a three-term controller (PID) in trajectory tracking of the P-AFO.

Keywords

• rehabilitation assistance
• ankle-foot orthosis
• subspace identification
• pid
• data-driven predictive control