A recent arXiv paper proposes an interaction-dynamics approach to controlling knee rehabilitation exoskeletons, arguing that safe rehabilitation is fundamentally a disturbance-rejection problem. The controller must regulate a prescribed motion while simultaneously absorbing involuntary spasms, voluntary effort from the patient, actuator compliance, and model mismatch.
The authors describe a closed-loop series elastic actuator (SEA) outer-loop study using model predictive control (MPC). By explicitly treating these interaction forces as disturbances, the framework aims to maintain safe, prescribed movement without fighting the patient or the hardware's inherent flexibility.
The paper is a single-source contribution, and no conflicting findings are noted. It focuses on the control formulation and the SEA outer-loop architecture, though the abstract stops short of reporting specific experimental results in the provided text.