| 研究概要 |
The study of natural reaching movement is very important in the design of control systems for modern skillful robots. We start this research with analyzing human-like reaching movement and constructed two modes, the minimum variance model and the minimum effort model, describing and predicting the invariant features of these movements. It has been shown that these two different models predict the same outcome if the model of the control plant is stable. The theoretical models have been verified under experiments in virtual environments with the use of haptic interface and a simulator incorporating different holonomic constraints. In the second part we dealt with non-holonomic constraints and considered reaching movement for a rolling robot actuated either by internal rotors or a pendulum. This system is not differentially flat and not nilpotent and the conventional motion planning methods are not applicable. It was found that kinematically assigned trajectories may not be dynamically realizable and this requires the development of new techniques. It was established that the optimal control results in the Euler’s elasticae but the numerical implementation is sensitive to the initial guess. To alleviate these difficulties, a motion planning strategy employing a nilpotent approximation, iterative steering, and the geometric phase approach has been developed and verified under simulation. In addition, two spherical rolling robots, one actuated by internal rotors and one by pendulum, have been designed and their motion abilities have been verified under initial experiments.
|
