Estimated 15,000 Veterans suffer a stroke each year. Stroke is a leading cause of long-term disability in the US. New strokes cost an estimated $111 million for acute inpatient care, $75 million for post-acute inpatient care, and $88 million for follow-up care in the first six months post-stroke in VHA. Yet, more than two thirds of stroke survivors have persistent hand impairment that significantly diminishes their abilities to perform activities of daily living. While it is known that training for healthy movement patterns is critical especially early on in rehabilitation, precise control of multiple finger joints simultaneously is not possible in current therapy. Controlling finger joint movements is challenging since the human hand has more than 20 degrees of freedom (DOF) densely located in a small space. There exist robots for hand rehabilitation that train for gross grasping and finger individuation. However, current robots have limited DOF and cannot control finger joint torques, so these systems are unable to deliver training that ensures healthy movement patterns. Thus, there is lack of tools for delivering training that ensures healthy movement pattern and prevents compensatory movements. With a lack of joint-level training tool, patients are either left with compensatory patterns, or worse, have no recovery of hand movements. With the long-term goal of improving hand rehabilitation, we have designed a robotic tool called the Maestro hand exoskeleton. Maestro’s design features enable delivery of versatile interventions for patients with a wide range of impairments in various stages of recovery. This innovation represents a substantial advancement from current rehabilitation robotic tools by providing high-intensity, task-based training with real time modulation of assistance and difficulty level ensuring patient participation and task saliency. The objective of this project is to develop novel controllers with promising neurological basis for training correct movement patterns in stroke patients. Specifically, (1) compensation avoidance (CA) controller will apply joint torques to push the patients away from the compensatory joint coordination, only interfering with the movements once the subject initiates a compensatory movement strategy. (2) Task assistance (TA) controller will apply assistive joint torques to directly help stroke patients achieve finger tasks with correct coordination. For both controllers, the torques in the finger joints will be modulated to match the individual patient’s ability, impairment, and progression throughout the training via robot control program. Four and nine Veteran subacute stroke survivors with moderate to severe hand impairment and with some ability to move fingers will participate in the testing for Aim1 and Aim 2, respectively. Aim 1 will involve one session and Aim 2 will involve four sessions of experimentation for each participant. Aim 1: Develop and determine feasibility of CA and TA controlle...