Progress based assist-as-needed control strategy for upper-limb rehabilitation

Assist-as-Needed (AAN) strategy is one strategy of administering robotic assistance which has gained wide acceptance in recent time and has been clinically demonstrated to provoke motor recovery in neurologically impaired patients. The underlying principle in this scheme is to provide assistance to a patient to complete desired movement only when the patient is in need of that assistance, which invariably encourages patient's active participation to provoke neuroplasticity. Effective deployment of this scheme in robot aided therapy still pose some challenges. One current difficulty is finding an appropriate method of estimating patients' functional input, or movement ability, by means on which the robot judges whether to assist or to relax the assistance. Another concern is the repeatability or consistency of the estimated subject's movement ability across a wide range of subjects with different disability level that can warrant a broad clinical adoption. Also, generally overlooked is the influence of subject's recovery progress on the modality of the AAN assistance. In this study, we propose a progress-based assist-as-needed algorithm (pAAN) which rely on subject's real-time functional ability record to drive robotic assistance. We define a novel functional ability index (FAI) on a baseline adaptive inertia related controller which decays the robotic torque assistance according to patients' progress or active participation during therapy session. Simulation studies carried out for the task of lifting a can from table to the mouth, designed in accordance with the Wolf motor function test (WMFT), have demonstrated the feasibility of the pAAN strategy.