Finite state control of a variable impedance hybrid neuroprosthesis for locomotion after paralysis

Abstract

We have previously reported on a novel variable impedance knee mechanism (VIKM). The VIKM was designed as a component of a hybrid neuroprosthesis to regulate knee flexion. The hybrid neuroprosthesis is a device that uses a controllable brace to support the body against collapse while stimulation provides power for movement. The hybrid neuroprosthesis requires a control system to coordinate the actions of the VIKM with the stimulation system; the development and evaluation of such a controller is presented. Brace mounted sensors and a baseline open loop stimulation pattern are utilized as control signals to activate the VIKM during stance phase while simultaneously modulating muscle stimulation in an on-off fashion. The objective is twofold: reduce the amount of stimulation necessary for walking while simultaneously restoring more biologically correct knee motion during stance using the VIKM. Custom designed hardware and software components were developed for controller implementation. The VIKM hybrid neuroprosthesis (VIKM-HNP) was evaluated during walking in one participant with thoracic level spinal cord injury. In comparison to walking with functional neuromuscular stimulation alone, the VIKM-HNP restored near normal stance phase knee flexion during loading response and pre-swing phases while decreasing knee extensor stimulation by up to 40%.

Fig. 1

The five discrete states of the VIKM-HNP finite state machine. (a) The schematic indicates the damper state and knee position for the leading limb during each controller state. An open circle indicates the VIKM is off, a half-filled circle indicates the damper is active but allows flexion, and a black circle indicates the VIKM is active at a high level to prevent flexion. (b) A schematic of the VIKM-HNP finite state machine including knee extensor stimulation (on-off), VIKM activation level, and transitions between states. Note that FNS Psw feedback refers to the control signal generated from the stimulation pattern (see Fig. 2).

Fig. 2

The FNS pattern and stimulation based feedback control signals for a right heel strike (RHS) to RHS gait cycle. Vertical bar indicates approximate left heel strike. (a) 16 channel baseline stimulation pattern for walking. Dark areas represent high frequency (33.3 Hz), light areas low (16.7 Hz). See Table 3 for abbreviations. (b) Feedback signals derived from the stimulation pattern used to aid in transition to pre-swing phase and to provide supervisory control of the VIKM-HNP finite state machine. (c) States of the VIKM finite state machine. LR is loading response, Mst is mid-stance, Psw is pre-swing.

Fig. 3

Participant with SCI donning the prototype VIKM-HNP.

Fig. 4

A comparison of loading response (leading limb) and pre-swing (trailing limb) phases during walking with (a) FNS-only and (b) VIKM-HNP.

Fig. 5

The average (± 1 SD) (a) VIKM damper activation, (b) knee extensor pulse width, and (c) knee angle with respect to gait cycle for left and right limbs in walking with VIKM-HNP (solid) and FNS-only (dashed). Data is averaged over 30 and 38 strides for the VIKM-HNP and FNS-only conditions, respectively.