NON-DETERMINISTIC MODELLING AND ITS APPLICATION IN ADAPTIVE OPTIMAL CONTROL

Date
1982-04-01
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
A problem which occurs in every branch of system theory is the conflict between data acquisition for system identification and systeM control towards some objective. The appropriate actions for identification are rarely those most appropriate for control, and yet in most real system studies control actions cannot be determined properly without some prior identification. The need to balance the costs of further exploratory activity against the costs of poor system knowledge creates a basic conflict that cannot be avoided. This conflict is succinctly expressed in the two-armed bandit problem, which has therefore received a great deal of attention since it was formulated three decades ago. A useful toolkit of techniques has been accumulated to attack the problem under various realistic constraints, such as finite memory of past history, finite complexity in terms of state count, and finite time.
Description
Keywords
Computer Science
Citation