Learning Therapist Policy from Therapist-Exoskeleton-Patient Interaction
Grayson Snyder, Lorenzo Vianello, Levi Hargrove, Matthew Elwin, Jose L. Pons
AI summary
Problem
Current exoskeleton-based stroke rehabilitation relies on fixed control strategies or exhausting direct physical therapist guidance, which limits therapy intensity and prevents therapists from continuously monitoring patient progress.
Approach
The authors use a Variational Autoencoder and Gaussian Mixture Model to visualize therapist-patient interaction dynamics, and train an LSTM network on recorded therapist-patient dyad data to predict real-time therapist-applied joint torques from patient kinematics.
Key results
- PTFF visualization of therapist corrective strategies in a low-dimensional latent space
- LSTM-based Synthetic Therapist model accurately predicts therapist-applied joint torques from patient kinematics
- Leave-one-out cross-validation across eight post-stroke patients demonstrates model generalizability
- Real-time ROS integration enables exoskeleton torque assistance without direct therapist physical contact
Why it matters
This approach reduces therapist physical strain and enables continuous patient monitoring, potentially accelerating the clinical adoption of adaptive exoskeleton therapy for stroke rehabilitation.
Abstract
Post-stroke rehabilitation is often necessary for patients to regain proper walking gait. However, the typical therapy process can be exhausting and physically demanding for therapists, potentially reducing therapy intensity, duration, and consistency over time. We propose a Patient-Therapist Force Field (PTFF) to visualize therapist responses to patient kinematics and a Synthetic Therapist (ST) machine learning model to support the therapist in dyadic robot-mediated physi- cal interaction therapy. The first encodes patient and therapist stride kinematics into a shared low-dimensional latent manifold using a Variational Autoencoder (VAE) and models their inter- action through a Gaussian Mixture Model (GMM), which learns a probabilistic vector field mapping patient latent states to ther- apist responses. This representation visualizes patient–therapist interaction dynamics to inform therapy strategies and robot controller design. The latter is implemented as a Long Short- Term Memory (LSTM) network trained on patient–therapist interaction data to predict therapist-applied joint torques from patient kinematics. Trained and validated using leave-one-out cross-validation across eight post-stroke patients, the model was integrated into a ROS-based exoskeleton controller to generate real-time torque assistance based on predicted therapist re- sponses. Offline results and preliminary testing indicate the potential of their use as an alternative approach to post-stroke exoskeleton therapy. The PTFF provides understanding of the therapist’s actions while the ST frees the human therapist from the exoskeleton, allowing them to continuously monitor the patient’s nuanced condition.