The role of a priori knowledge of plant dynamics in neurocontroller design

The authors modify an earlier neurocontroller architecture so as to guarantee the performance of the neurocontroller. This architecture uses a priori knowledge of the general structure of the system's dynamics. The knowledge is utilized for the selection of exploratory schedules to excite selected subsets of the dynamics. The controller does not require a priori knowledge of the exact system dynamics, as they are learned online, nor does it assume the existence of an explicit external teacher. The control architecture developed is not limited to tracking of a prespecified trajectory. The architecture is developed for the control of a robot manipulator.<<ETX>>

[1]  A. Guez,et al.  A trainable neuromorphic controller , 1988 .

[2]  A. Sideris,et al.  A multilayered neural network controller , 1988, IEEE Control Systems Magazine.

[3]  B. Widrow,et al.  An adaptive 'broom balancer' with visual inputs , 1988, IEEE 1988 International Conference on Neural Networks.

[4]  Jean-Jacques E. Slotine,et al.  Adaptive manipulator control: A case study , 1988 .

[5]  Allon Guez,et al.  Neurocontroller design via supervised and unsupervised learning , 1989, J. Intell. Robotic Syst..

[6]  Moshe Kam,et al.  Neuromorphic architectures for fast adaptive robot control , 1988, Proceedings. 1988 IEEE International Conference on Robotics and Automation.

[7]  R. Paul.  Robot manipulators : mathematics, programming, and control : the computer control of robot manipulators , 1981 .

[8]  Bernard Widrow,et al.  Adaptive Signal Processing , 1985 .

[9]  Richard S. Sutton,et al.  Neuronlike adaptive elements that can solve difficult learning control problems , 1983, IEEE Transactions on Systems, Man, and Cybernetics.

[10]  A. Guez,et al.  Neural network architecture for control , 1988, IEEE Control Systems Magazine.

[11]  Geoffrey E. Hinton,et al.  Learning internal representations by error propagation , 1986 .