Robust Planning and Control Using Neural Networks
Abstract
During the past five years, the Robotics Laboratory of the Department of Electrical and Computer Engineering at the University of New Hampshire has been studying the application of locally generalizing neural networks to difficult problems in control. In a series of theoretical and real time experimental studies, learning control approaches have been shown to be effective for controlling the dynamics of multidimensional, nonlinear robotic systems during repetitive and nonrepetitive operations. This project involves the extension of our work in learning control, with the combined goals of expanding our theoretical understanding of neural network based learning control systems and of extending our experimental work to include hierarchical learning control structures. Our work will consider the efficacy of locally generalizing versus globally generalizing neural network architectures in control applications, as well as developing and analyzing learning control paradigms which are not restricted to specific network architectures. Various robotic systems within the laboratory will form the basis for the real time experimental portions of the research. The concepts explored, however, will be applicable to a wide variety of control problems in addition to robotics.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 31, 1989
- Accession Number
- ADA217216
Entities
People
- Filson H. Glanz
- L. G. Kraft Iii
- Michael J. Carter
- W. T. Miller Iii
Organizations
- University of New Hampshire