| Robotic technology has the potential to deliver therapy activities for rehabilitation of arm and hand function after stroke more efficiently and effectively than conventional rehabilitation, as it can objectively dose the prescribed intensive amount of therapy in automated design with less cost and effort, and can provide highly reliable measurement of patients' progress. The primary goal of this dissertation is to develop a robotic rehabilitation system that fulfills current guidelines for the stroke rehabilitation: motor training focus on realistic tasks that require reaching and manipulation and engage patients post stroke intensively, actively, and adaptively.Firstly, we presented a novel robotic task-practice system, i.e., adaptive and automatic presentation of tasks (ADAPT), which was designed according to the guidelines. A modular and reconfigurable robot with the configuration of a 3 degree-of-freedom (DOF) wrist mounted on a 1-DOF linear actuator simulates the dynamics of functional tasks and presents the functional tasks to patients post stroke. A novel tool-changing system enables ADAPT to automatically switch between the tools corresponding to the functional tasks. The control architecture of ADAPT is composed of three main components: a high-level task scheduler, a functional task model, and a low-level admittance controller. The high-level task scheduler adaptively selects the task to practice and sets the task difficulty based on the previous performance of the patients. The functional task model generates desired trajectories based on learned models of task dynamics. Tasks dynamics are modeled with receptive field weighted regression (RFWR), such that the feel of the task tools is accurately modeled, and the task difficulty can be easily adjusted. The low-level admittance controller, which is also learned with RFWR, implements the selected task trajectory for robot--patient interaction.Secondly, we introduced new adaptive schedules for the high level adaptive task scheduler of ADAPT in an attempt to maximize the relearning of multiple functional tasks and balancing learning among tasks in a limited training time. Although random scheduling of several tasks has been shown to enhance learning more than blocked scheduling does, the advantages of random scheduling may be limited because it does not take into account the nominal difficulty of each task, the difference in difficulty between tasks, and the skill level of the learner in that type of schedule. We proposed two new algorithms for adaptively determining the nominal difficulty and the number of trials for each task on the basis of both current and delayed performance of the learner (N = 48). We tested the adaptive algorithms in a 2 x 2 factorial design, and they show that the algorithms outperform random scheduling when performance is measured on a delayed retention test.Finally, we investigated the feasibility of ADAPT to patients post stroke by evaluating safety, system utility, fidelity of simulated tasks and patient acceptance. Five patients with chronic stroke participated in approximately one hour training session with an adaptive difficulty schedule. Additional pre and post test sessions lasted approximately 10 minutes each, and questionnaire were administered after all the sessions. All participants completed the presented sessions, functional measurements, and questionnaires without any adverse event or report from the participants. ADAPT provided adaptive training tailored to patients' performance by modulating task difficulty in the training session. The results from this study validate the feasibility of ADAPT for rehabilitation of arm and hand function after stroke, and provide justification for continued investigation of clinical efficacy. |
