Synthetic Data-Driven Exoskeleton Control via Contralateral Gait Fusion for Variable-Speed Walking

Data-driven exoskeletons offer the potential for adaptive augmentation of human mobility. Yet their widespread adoption is hindered by labor-intensive biomechanical data collection and manual tuning. Herein, this study presents a highly efficient synthetic data approach to facilitate data-driven pipelines. We leveraged an Adversarial Motion Priors (AMP) agent to learn stylized walking within a massively parallel, physics-based simulation. The resulting high-fidelity data were collected and validated against OpenSim inverse dynamics pipelines. Further, we trained an end-to-end torque prediction algorithm using the collected data. A novel CNN-Transformer architecture was developed to map contralateral swing-phase data to variable-length push-off torque profiles. This enabled real-time, adaptive torque assistance of exoskeletons for variable-speed walking. A custom ankle exoskeleton was used to demonstrate robust sim-to-real transferability. Our system achieved an average root mean square error of approximately 0.081 ± 0.015 newton-meters per kilogram and an average R2 of 0.836 ± 0.050 across speeds ranging from 0.6 to 1.75 m·s−1. The controller significantly reduced user-positive ankle mechanical work by up to 14 ± 6.30%. Finally, our multi-sensor configuration exhibited inherent fault tolerance, ensuring safe operation even under partial sensor failure. By taking a scalable, data-driven approach, this work offers a practical pathway toward deploying autonomous exoskeletons in versatile, real-world environments.