Desarrollo de estrategias de control biocooperativo para plataformas robotizadas de rehabilitación neuromotora

Abstract

Rehabilitation robotics has emerged as a promising solution to promote motor recovery and functional independence for patients with neurological disorders. Neurorehabilitation with robotic devices has demonstrated its potential to provide intensive and repetitive training to promote motor recovery. These robotic devices can provide different types of assistance and can be controlled by different input modalities such as electromyography (EMG), electroencephalography (EEG), and kinematic signals. Among the different input modalities, EMG has been widely used, as it can provide real-time information about the patient¿s muscle activation patterns and allow for a biocooperative control between the user and the robot. However, physiological acquisition systems are too expensive, and often bulky, to be used in clinical settings. Moreover, these systems typically lack processing capabilities, thereby impeding the development of real-time biocooperative control strategies and an efficient human-robot interaction. Additionally, biocooperative controls based on machine learning algorithms have not been implemented in real environments, mainly due to reliability issues or the need of a robot with a large number of degrees of freedom. In this context, the present Doctoral Thesis is focused on development of biocooperative control strategies, coupled with the design of low-cost embedded solutions for physiological data acquisition that contribute to provide a real use by individuals suffering from neuromotor impairments. The studies included in this compendium of publications are primarily concern with motor rehabilitation of the upper-limb. This is a significant area of focus, as upper-limb paresis is among the most frequently observed outcomes of stroke, having a profound effect on the quality of life and independence of stroke survivors. The contributions of this study are canalized in three different ways. First, affordable solutions have been created for the acquisition of physiological signals and implementation of biocooperative control in real-time embedded systems. An EMG recording system and a wearable multimodal physiological acquisition system have been designed to enhance accessibility and facilitate the use of upper-limb biocooperative control in the clinical settings. Secondly, a non-pattern recognition-based EMG-driven control has been developed for hand rehabilitation robot. The system operates on a real-time embedded platform and has demonstrated favorable performance in terms of both accuracy and latency. Additionally, we found that providing EMG-based visual feedback to subjects during EMG-driven bilateral therapies led to significant improvements in performance. These results suggest that the feedback enables subjects to increase their control over the movement of the robotic platform by assessing their muscle activation in real-time. Lastly, an EMG&IMU-based control using virtual reality-based therapy, along with an adaptive assistive control (AAN) using a wrist rehabilitation robot, have been proposed and employed to validate the performance of the embedded multimodal acquisition platform. The wearable system, which integrates multiple sensors, wireless communication, and a high-efficiency real-time microcontroller, is characterized by its high versability and configurability. It has been verified that its low-cost does not compromise the quality of the signals and it could promote the development of real-time biocooperative controls for a wide range of neuromotor rehabilitation applications. Overall, the findings of this Doctoral Thesis could pave the way for the development of more affordable and effective robotic devices for upper-limb neurorehabilitation and provide insights into the design and implementation of biocooperative controls for neurorehabilitation platforms.