A new paper on arXiv proposes a multimodal control framework for upper-limb rehabilitation that combines mental and physical engagement. The work targets limitations in existing Assist-as-Needed (AAN) control, which adjusts robot assistance based on user performance but typically uses only physical measures.

The authors argue that physical metrics alone cannot tell whether a user is failing due to muscular weakness or because they are mentally disengaged. Their flow-theoretic approach integrates both types of engagement, potentially allowing the robot to respond more appropriately to the user's actual state.

The abstract does not provide experimental results or implementation details, so the practical benefits remain to be demonstrated. Still, the conceptual shift toward including mental engagement in AAN control could address a known gap in rehabilitation robotics.