Advanced Adaptive Critic Designs
Abstract
We present a unified approach to a family of Adaptive Critic Designs (ACDs). ACDs approximate dynamic programming for optimal control and decision making in noisy, nonlinear, or nonstationary environments. This family consists of Heuristic Dynamic Programming (HDP), Dual Heuristic Programming (DHP), and Globalized Dual Heuristic Programming (GDHP), as well as their Action-Dependent forms (the prefix AD denotes these)[1]. The most powerful of these designs reported previously is GDHP [2,3]. After pointing out problems of the simple ACDs, we describe advanced ACDs and introduce ADGDHP. We also propose a general training procedure for ACDs and discuss some important research issues.
Recommended Citation
D. V. Prokhorov and D. C. Wunsch, "Advanced Adaptive Critic Designs," World Congress on Neural Networks. 1996 International Neural Network Society Annual Meeting (1996, San Diego, CA), pp. 83 - 88, Lawrence Erlbaum Associates, Jan 1996.
Meeting Name
World Congress on Neural Networks, International Neural Networks Society Annual Meeting (WCNN'96) (1996: Sep. 15-18, San Diego, CA)
Department(s)
Electrical and Computer Engineering
International Standard Book Number (ISBN)
805826084
Document Type
Article - Conference proceedings
Document Version
Citation
File Type
text
Language(s)
English
Rights
© 1996 Lawrence Erlbaum Associates, All rights reserved.
Publication Date
01 Jan 1996
