2016 Volume 5 Issue 2 Pages 123-131
In the last few decades, the increase in the worldwide elderly population and the progress in the treatment of severe and chronic pathologies have led to a growing demand for rehabilitation therapies. Meanwhile, rehabilitation robotics has started to grow and to evolve, in order to develop suitable robotics devices and control strategies to better assist patients during training and to promote rehabilitation processes. In particular, some control strategies are designed to assist patients in completing the desired movements while applying the minimum force necessary. As a result, an “assist-as-needed” behavior can be achieved. A novel nonlinear adaptive compliance controller, that aims to achieve such “assist-as-needed” behavior, has been developed and is presented in this paper. In addition to promote the active participation of patients, the proposed control also provides a tool to estimate and evaluate patient's state and therapeutic improvements. The proposed controller is obtained by appropriately merging a PD (proportional and derivative) control and an adaptive learning control. The latter is driven by the errors made by patients while performing the assigned exercise. As a result, the PD controller parameters are adapted according to different patient injuries and degrees of impairments and may be used to evaluate the improvements during training sessions. The paper presents an overview of the novel control algorithm and some preliminary clinical trials with real patients, demonstrating benefits of the controller.