Adaptive critic (AC) neural network solutions to optimal control designs using dynamic programming has reduced the need of complex computations and storage requirements that typical dynamic programming requires. In this paper, a "single network adaptive critic" (SNAC) is presented. This approach is applicable to a class of nonlinear systems where the optimal control (stationary) equation is explicitly solvable for control in terms of state and costate variables. The SNAC architecture offers three potential advantages; a simpler architecture, significant savings of computational load and reduction in approximation errors. In order to demonstrate these benefits, a real-life micro-electro-mechanical-system (MEMS) problem has been solved. This demonstrates that the SNAC technique is applicable for complex engineering systems. Both AC and SNAC approaches are compared in terms of some metrics.

Meeting Name

American Control Conference, 2004


Mechanical and Aerospace Engineering

Keywords and Phrases

Adaptive Control; Adaptive Critic Neural Network; Approximation Error Reduction; Approximation Theory; Complex Engineering Systems; Computational Complexity; Computational Load; Control System Synthesis; Costate Variables; Dynamic Programming; Large-Scale Systems; Microelectromechanical System; Micromechanical Devices; Neural Net Architecture; Neurocontrollers; Nonlinear Control Systems; Nonlinear Systems; Optimal Control; Optimal Control Equation; Optimal Control Synthesis; Single Network Adaptive Critic Architecture; State Variables; Stationary Equation; Storage Requirements

International Standard Serial Number (ISSN)


Document Type

Article - Conference proceedings

Document Version

Final Version

File Type





© 2004 Institute of Electrical and Electronics Engineers (IEEE), All rights reserved.

Full Text Link