EEG-Induced Adaptation of Controller Parameter for Closed-Loop Position Control of the End-Effecter in a Robot Arm

摘要

Existing brain-computer interfaces for position control of robot arms usually attempt to control the motion of the individual links of the arm in order to control the position of the end-effecter. The above position control schemes suffer from one fundamental limitation in mental pre-selection of the target positions. This paper provides an interesting approach to overcome the above limitation by attempting to directly control the position of the end-effecter in the desired location by mental imagination of its movement rather than attempting to position the individual links of the robot arm. This has been realized by utilizing the benefits of inverse kinematics of the robot arm using neural network based function approximation to serve the computation of inverse kinematics in real time. Apart from the above, the other notable point of interest of the present work lies in controlling the step-size of motion of the end-effecter by brain-inspired position control. An increase in step-size is allowed until the end-effecter crosses any one of the three reference planes of the target position, whereas a decrease in step-size is selected when the end-effecter crosses a reference plane. The decision about crossing the reference plane by the end-effecter is subjective and is also dependent on subject's training. The proposed 3-dimensional position control scheme offers low settling time (0.5 sec), small (<; 1 %) peak overshoot and also small steady-state error (0.02 %). It can readily be used as an artificial appendage for the people with neuro-motor disability.