Optimal Trajectory Control of Arms Using a Neural Network Model
Faculty of Engineering University of Tokyo 7-3-1, Hongo, Bunkyo-ku Tokyo 113, Japan
Published:August 20, 1990
The author et al. have proposed a minimum torquechange model as a computational model for predicting and reproducing movement trajectories of human arms. The model proposes that in voluntary movements, trajectories should be designed so that changes with time in motor command diminish as much as possible while satisfying conditions given for movement. This paper propose a neural network model capable of autonomously generating movement trajectories on the basis of the minimum torque-change model. The basic operations of the neural network model are based on learning and relaxation. In other words, (1) the dynamics model of a controlled object is obtained in the neural network by learning and (2) movement trajectories are generated by relaxation of the neural network on the basis of the model obtained. For generating movement trajectories, two methods can be considered: one using the forward dynamics model of a controlled object, and the other using the inverse dynamics model. This paper proposes two different types of neural network models corresponding to these.
Cite this article as:Y. Uno, “Optimal Trajectory Control of Arms Using a Neural Network Model,” J. Robot. Mechatron., Vol.2 No.4, pp. 266-272, 1990.Data files: