Artificial Intelligence Theory and Reconfigurable Control Systems

摘要

The principal topic of research in this effort was the development of analysis and design methods for fault-tolerant control systems, using artificial intelligence concepts to screen failure hypotheses, and more conventional control theory for reconfiguration and actual control. Methods developed include new graphical presentations of logic trees and signal dependency graphs, procedures for translating expert systems from LISP to Pascal, the definition of performance metrics for real-time expert systems, a Control Equation Parser computer program for computer-aided control system design, and a control-reconfiguration strategy based on proportional-integral implicit model-following. A Rule-Based Flight Control System for a tandem-rotor helicopter, implemented with parallel microprocessors and operating in real time, was developed in the course of this research. This task made use of the Princeton Rule-Based Controller, a unique software architecture for combining procedural and symbolic processing. The knowledge base is developed in the LISP computer language and is translated with the inference engine into the Pascal language. The ease with which Pascal code can be embedded in the knowledge-base structure makes this a highly suitable tool for task and algorithm scheduling. (kr)