Title :
Learning By Doing; Synthesis and Refinement of Control Strategy
Author :
Bernstein, Mark ; Miller, Donald L. ; Schnelle, P.David ; Morganstein, Lester I.
Author_Institution :
E. I. du Pont de Nemours & Co. Inc., Experimental Station 356/203, Wilmington, DE 19898
Abstract :
This work describes an application of machine learning to the synthesis and refinement of control strategy. Given a model of the target system´s dynamic behavior, we describe a program that proposes possible control strategies and tests their behavior by exploring a curriculum of training problems. By generalizng from its successes and failures, the program learns to concentrate its efforts on control structures which are most likely to reward study.
Keywords :
Control system synthesis; Control systems; Feeds; Learning systems; Machine learning; Mechanical variables control; Process control; Sensor systems; System testing; Temperature sensors;
Conference_Titel :
American Control Conference, 1987
