Development of an adaptive, AI-guided, closed-loop control system for Functional Electrical Stimulation Cycling in Spinal Cord Injury

Abstract

Functional Electrical Stimulation (FES) cycling is an existing rehabilitation strategy for individuals with spinal cord injury (SCI), with demonstrated potential to preserve muscle strength. There is not enough data to substantially conclude that it also promotes bone health. Many existing FES cycling systems operate using open-loop or pre-timed stimulation patterns that do not dynamically respond to real-time crank kinematics, potentially leading to inefficient torque generation and premature muscle fatigue.

This project proposes a closed-loop control architecture that modulates stimulation delivery based on crank velocity and pedal angle on a rat bicycle. Encoder-derived crank position is used to compute angular velocity, which is continuously compared to a predefined target velocity to generate a real-time error signal. Electrical nerve stimulation is applied to either the femoral nerve or sciatic nerve to produce muscle force and crankshaft torque to minimize the error between the measured velocity and the desired velocity. A phase-dependent decision subsystem gates stimulation according to biomechanically appropriate crank angle ranges, directing corrective output to femoral stimulation (30°–150°) or sciatic stimulation (150°–270°). In the zone where unilateral muscle contraction does not contribute to rotation (“dead zone”; <30°, >270°), rotation is achieved with motor assistance.

The architecture was developed and simulated in MATLAB/Simulink to validate logical signal flow and conditional activation pathways. This structured framework establishes a foundation for future implementation of adaptive control strategies that can tune stimulation parameters based on defined performance metrics without requiring large training datasets.