Learning a Shape-Adaptive Assist-As-Needed Rehabilitation Policy from Therapist-Informed Input
Zhimin Hou, Jiacheng Hou, Xiao Chen, Hamid Sadeghian, Tianyu REN, Sami Haddadin
AI summary
Problem
Current therapist-in-the-loop rehabilitation systems lack effective mutual interaction and adaptation, making it difficult to deliver intuitive, minimal-assistance therapy without overwhelming patients or requiring constant reprogramming.
Approach
The system uses two connected robots to capture therapist corrective forces, encoding them as sparse via-points in a latent space that partially and progressively deforms the reference trajectory while preserving the patient’s natural motion preferences.
Key results
- Encodes therapist corrective forces into sparse latent via-points to minimize assistance
- Learns a shape-adaptive policy to partially deform reference trajectories based on patient preferences and therapist input
- Validates the framework on a dual-robot telerobotic system across two representative tasks
- Outperforms two state-of-the-art methods in reducing corrective force and improving movement smoothness
Why it matters
Enables precise, remote therapist-guided rehabilitation that maximizes patient engagement while bridging the gap between human expertise and robotic execution.
Abstract
Therapist-in-the-loop robotic rehabilitation has shown great promise in enhancing rehabilitation outcomes by integrating the strengths of therapists and robotic systems. However, its broader adoption remains limited due to insuffi- cient interaction and limited adaptation capability. This article proposes a novel telerobotics-mediated framework that enables therapists to intuitively deliver assist-as-needed (AAN) therapy based on two primary contributions. First, our framework encodes the therapist-informed corrective force into via-points in a latent space, allowing therapists to provide only minimal assistance while encouraging patients maintaining their own motion preferences. Second, a shape-adaptive AAN rehabilita- tion policy is learned to partially and progressively deform the reference trajectory for movement therapy based on encoded patient motion preferences and therapist-informed via-points. The effectiveness of the proposed shape-adaptive AAN strategy was validated on a telerobotic rehabilitation system using two representative tasks. The results demonstrate its practicality for remote AAN therapy and its superiority over two state- of-the-art methods in reducing corrective force and improving movement smoothness.