From Simulation to Stimulation: A Predictive Synergy-Based Control Strategy for FES Cycling in a Person with SCI

Background: Functional electrical stimulation (FES) cycling stimulation strategies are often derived from heuristics, or able-bodied kinematics and electromyography, which may be unsafe and non-representative for people with complete SCI. Optimally controlled predictive musculoskeletal simulations with synergy implementation offer a way to generate individualized, physiologically plausible muscle activation patterns without prior experimental data. Building on a previous proof of concept, this study developed and experimentally validated a subject-specific simulation-derived stimulation strategy for FES cycling. Methods: A two-stage musculoskeletal predictive simulation framework was implemented using OpenSim Moco. First, a speed-tracking optimal control problem optimized the model's kinematics. Second, muscle excitations were decomposed into synergies using non-negative matrix factorization, producing coordinated activation patterns for quadriceps and hamstrings. Subject-specific constraints were incorporated into the model. The resulting activation profile was then processed and converted to a crank-angle lookup table for real-time control. Experimental validation was conducted with an experienced FES cycling pilot with complete SCI, comparing the simulation-derived Control Signal (CS) with a refined empirical bang-bang (BB) protocol used in the Cybathlon 2024 edition. Results: Simulations generated stable synergy-based activation patterns aligned with expected cycling dynamics. Experimentally, performance varied with trial order, with both CS and BB enabling unassisted FES cycling. BB produced higher average power and cadence, while CS delivered more symmetrical left–right balance and showed partial recovery in performance as pulse width increased. Discussion: Although BB outperformed CS in power and cadence, the CS approach proved to be feasible and has methodological advantages. The consistent and physiologically structured behavior of the CS signal demonstrates that predictive simulation can provide a safe, reproducible baseline for individualized FES-cycling control, particularly for new users or research groups. Conclusion: This was the first implementation in cycling of optimal control-driven FES combined with muscle synergies, generating feasible, physiologically grounded stimulation profiles for FES cycling in individuals with complete SCI. While not surpassing a highly refined empirical strategy, the simulation-derived signal performed reliably and represents a promising foundation for model-based personalization of FES control. Trial Registration: Universidade de Brasília Ethics Committee, 75441323.4.0000.8093.