Letters: Adaptive biomimetic control of robot arm motions
Neurocomputing
A biped static balance control and torque pattern learning under unknown periodic external forces
Engineering Applications of Artificial Intelligence
| Hi-index | 0.00 |
In studies of robots, intended for collaborative operation with humans, there have been intensive efforts to develop and apply a muscle-driving system with mechanically soft properties, to reduce the danger to humans. Such a muscle-driving system has the advantage that the danger to humans in collaborative operation can be reduced, but has the disadvantage that accurate control is difficult due to the soft properties. The soft and accurate motion of the limbs in biological systems is realized by skillful control of the muscles by the control system composed of the spinal cord and the cerebellum. Consequently, inspired by the motion control mechanism in biological systems, this paper tries to develop a control system that provides accurate trajectory tracking control of the robot, by using a muscle-driving actuator with soft mechanical properties. The proposed control system consists of the computational models of the stretch reflex system in the spinal cord, which compensates the external load torque exerted on the joint, and the cerebellum, which compensates the inertial torque with learning the internal model of the limb dynamics. Using the indirect Lyapunov method, it is shown theoretically that the posture of the robot asymptotically approaches the desired trajectory when driven by the proposed control system. A computer simulation is performed in order to demonstrate the effectiveness of the control system, and it is verified that the behavior converges satisfactorily to the desired target. © 2004 Wiley Periodicals, Inc. Syst Comp Jpn, 35(11): 1–13, 2004; Published online in Wiley InterScience (). DOI 10.1002/scj.10646