Our research aims to develop machine learning models for estimating biological joint moments in stroke populations. Accurate joint moment estimation is critical for controlling assistive exoskeletons, which can significantly improve mobility in individuals with impaired gait.

While existing models are effective for able-bodied populations, they fall short in clinical groups due to the variability in gait patterns post-stroke and the challenges in collecting large-scale biomechanical data in clinical populations. By leveraging transfer learning and other cutting-edge techniques, our goal is to overcome data limitations and develop intelligent, adaptable algorithms that can provide assistance via personalised exoskeleton control. This could eventually contribute to the design of more effective assistive devices, fostering improved mobility and independence for stroke survivors and other clinical populations.