No abstract

For the purpose of implementation of human smooth and fast movement to a pneumatic robot, the authors used a humanoid robot arm with pneumatic agonistantagonist actuators as endoskeletons which has control mechanism in the stiffness of each joint, and the controllability was experimentally discussed. Using Kitamori's method to experimentally decide the control gains and using I-PD controller, three joints of the humanoid robot arm were experimentally controlled. The results showed that the controllability to step-wise input was less than one degree in error to follow the target angles, and the time constant was less than one second. The simultaneous input of command to five joints was brought about the overshoot of about ten percent increase in error. The robot arm generated the calligraphic motion when a humanoid robot hand was attached to it. These results suggested that the I-PD controller and control gains obtained in this study were proper for the humanoid robot arm with pneumatic agonist-antagonist actuators to move smoothly, quickly and accurately. I 1-4244-1 184-X/07/$25.00©2007 IEEE

In this study, we propose a control algorithm for a pneumatic actuator that has dynamics and features similar to those of the human muscle, mainly with the aim of helping elderly persons communicate. The algorithm in this study can estimate gains by using a simple method with a double-acting air cylinder and can realize accurate speed control and position control. Specifically, we aimed to achieve quick response and less overshoot by providing a PD controller for common bias pressure control, that can generate passive stiffness, in addition to a PID controller capable of controlling disturbance and target tracking without any complicated control system. We performed gain estimation by first theoretically estimating the PID gain and then determining the optimum PD gain by actually moving an air cylinder. We tried controlling a system consisting of one air cylinder and a solenoid valve and found that the overshoot, which was nearly 30% with only the PID controller, was controlled to 4%, while the rise time was less than 200 ms of that when only the PID controller was used.