Neural Network Control of DoD and Industrial Motion Systems
Abstract
Actuator dead zones, backlash, and saturation impose severe performance limitations in industrial and DoD systems. Modern Battle Information Systems need improved dynamic decision-making control systems to avoid NP-complexity problems and properly assign resources. The goals of this grant were to develop neural network (NN) compensators for industrial actuator nonlinearities, to develop high-level NN architectures for control, to implement NN controllers on actual devices, and to design and implement discrete event decision controllers. A family of NN and fuzzy logic (FL) controllers was developed for dead zones and backlash. Rigorous analytical techniques were given for the design of NN controllers for actuator compensation that guarantee stability. High-level NN adaptive critic controllers were designed. Intelligent controllers were implemented on industrial test beds. The authors designed a remote site control system allowing monitoring and control of systems over the internet. A supervisory controller was designed based on matrices that allows for fast changing of goals and priorities. Matching funds have allowed them to work with small companies and transfer the ARO technology to them. Numerous students graduated and received many awards. The report includes a list of 3 books, 3 book chapters, 1 journal, 19 journal articles, 24 conference papers, and 3 patents that were produced as a result of this grant. (10 figures)
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 2003
- Accession Number
- ADA429174
Entities
People
- Frank L. Lewis
Organizations
- University of Texas at Arlington