Intelligent Control of Uncertain Systems.
Abstract
This project has developed provably correct architectures and reconfiguring algorithms for controlling processes whose dynamical models may change drastically due to aging, component failure or other unpredictable events. With AFOSR support, the devising, testing and analysis of a provably correct, 'smart. high-level controller called a supervisor has been completed. The supervisor is capable of controlling the set-point of a very poorly modelled process by orchestrating the of switching a sequence of candidate, off-the-shelf, linear set-point controllers into feedback with the process. The provable features of the overall supervisory control system include robustness to unmodelled dynamics, noise and disturbances, as well as exponential convergence in the absence of noise. With the ultimate goal of extending these ideas to the supervision of families of nonlinear regulators, it has been shown that the any certainty equivalence control causes the familiar interconnection of a controlled process and associated output estimator to be detectable through the estimator's output error, for every frozen value of the index or parameter vector upon which both the estimator and controller dynamics depend. The concept of supervisory control has been successfully applied, both in simulations and in laboratory experiments, to the problem of auto-calibrating stereo-vision based system for driving a rigid mobile robot to a prescribed target.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1997
- Accession Number
- ADA332017
Entities
People
- A. Stephen Morse
Organizations
- Yale University