Estudo sobre interfaces cérebro-máquina e Interação humano-robô

Abstract

Brain-machine interfaces (BMIs) are systems capable of extracting information from the brain activity of a user to allow the control of any device. The development of BMIs is a goal pursued by various research laboratories around the world given its potential applications in the assistance of individuals su ering from motor disabilities. This dissertation seeks the development of a BMI based on the intention of movement of a person. The intention of movement is detected from the user's electroencephalogram (EEG) signals that can trigger the execution of an assistance task performed by a manipulator robot. The research consisted of three stages: (1) the study of the electrophysiological brain responses related to the intention of a movement, (2) the implementation of a detector of movement intention, and (3) the de nition of an assistance task using a manipulator robot. Electroencephalogram (EEG) signals from six volunteers were collected during the execution of a motor task consisting of a exion and an extension movement that was mimicked by an anthropomorphic robot manipulator. The volunteer was asked to stare, replicate, and imagine the movement performed by the robot. A detector of movement intention based on the spectral F test was implemented by comparing the instantaneous frequency spectrum of the EEG signals with a mean spectrum estimate. The assistance task de ned for the robot manipulator consisted in reaching a drink located in a xed position on a table, serving it to a person, and nally returning it to the table. The position of the person was tracked using a sensor, providing a closed-loop control at the human-robot interaction level. Given the underactuated structure of the robot, which provides only ve degrees of freedom for a task that requires six, a hierarchical controller that gives priority to position was used along the execution of the task. Orientation was controlled with less priority in the null space of the position task. EEG signals showed an energy desynchronization in the alpha band related to the motor task over the central, parietal, and occipital areas. The duration of the desynchronization corresponded to that of the movement. Detection rates ranging from 53 to 97 %, when using four realizations of the task, were found in ve of the volunteers when the movement was executed, in three of them when the movement was imagined, and in two of them when the movement was observed. Detection when the movement is observed rises the question of how the visual feedback may a ect the performance of a working BMI. The robot manipulator proved its capability to assist the user with the task of reaching and grasping for taking a drink. The robot was able to interact with people by getting close to their mouth and making it easy for them to take the drink. Overall results showed that the development of BMIs and its continuous improvement may lead, in the near future, to clinical and commercial applications to support and assist individuals who su er from motor disabilities. It is expected that the development of robotics technology will bring deeper interactions with the human being, leading to solutions that are comfortable, safe and functional.