Distributed Event-Sampled Approximate Optimal Control of Interconnected Affine Nonlinear Continuous-Time Systems

Abstract

In this paper, a novel distributed near optimal control of an interconnected affine nonlinear continuous-time system with known dynamics is presented by using event-sampled state vector via novel hybrid learning scheme. Neural networks (NN) using event sampled state vector are designed at each subsystem to learn the optimal value function in an online and forward-in-time manner to generate the solution to the infinite horizon Hamilton-Jacobi-Bellman (HJB) equation. The presence of strong interconnections and limited communication among subsystems complicates the controller design. In order to improve the learning rate without explicitly increasing the events during the learning phase, iterative weight updates combined with time driven learning at the event sampled instants is introduced. By using Lyapunov technique, it is shown that the state vector and the NN weights at each subsystem are locally uniformly ultimately bounded (UUB). Simulation results are provided to illustrate the effectiveness of the proposed analytical design.

Meeting Name

2016 American Control Conference, ACC (2016: Jul. 6-8, Boston, MA)

Department(s)

Electrical and Computer Engineering

Research Center/Lab(s)

Intelligent Systems Center

Comments

Research supported by NSF ECCS #1128281, #1406533 and Intelligent Systems Center.

International Standard Book Number (ISBN)

978-1-4673-8682-1

International Standard Serial Number (ISSN)

0743-1619; 2378-5861

Document Type

Article - Conference proceedings

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 2016 Institute of Electrical and Electronics Engineers (IEEE), All rights reserved.

Publication Date

01 Jul 2016

Share

 
COinS