Title :
Second-order training of adaptive critics for online process control
Author :
Govindhasamy, James J. ; McLoone, Sean F. ; Irwin, George W.
Author_Institution :
Res. Group, Queen´´s Univ. Belfast, UK
fDate :
4/1/2005 12:00:00 AM
Abstract :
This paper deals with reinforcement learning for process modeling and control using a model-free, action- dependent adaptive critic (ADAC). A new modified recursive Levenberg Marquardt (RLM) training algorithm, called temporal difference RLM, is developed to improve the ADAC performance. Novel application results for a simulated continuously-stirred-tank-reactor process are included to show the superiority of the new algorithm to conventional temporal-difference stochastic backpropagation.
Keywords :
intelligent control; learning (artificial intelligence); multilayer perceptrons; neurocontrollers; optimisation; process control; RLM training algorithm; action-dependent adaptive critic; intelligent control; multilayer perceptrons; neural networks; online process control; process modeling; process optimization; recursive Levenberg Marquardt; reinforcement learning; second-order training; simulated continuously-stirred-tank-reactor process; temporal-difference stochastic backpropagation; Adaptive control; Backpropagation algorithms; Intelligent control; Multi-layer neural network; Multilayer perceptrons; Neural networks; Neurocontrollers; Process control; Programmable control; Stochastic processes; Action-dependent adaptive critic; intelligent control; multilayer perceptrons; neural networks; nonlinear process control; process optimization; reinforcement learning; Algorithms; Artificial Intelligence; Bioreactors; Computer Simulation; Feedback; Models, Theoretical; Online Systems;
Journal_Title :
Systems, Man, and Cybernetics, Part B: Cybernetics, IEEE Transactions on
DOI :
10.1109/TSMCB.2004.843276
