Abstract | ||
---|---|---|
This paper presents a co-optimization scheme for an event-triggered control system to simultaneously optimize both the sampling instants and the control policy. A continuous time nonlinear affine system is considered and a novel performance index is defined to regulate the system states with minimum energy and optimal feedback frequency. To achieve this, a min-max optimization problem is formulated with the control policy and error due to event-triggered feedback as two non-cooperative policies. Using the two-player non-cooperative zero-sum game theory, solution to the min-max optimization problem is determined. The sampling instants are optimized by designing an event-triggering mechanism with worst-case sampling error policy as threshold while the control policy is designed to minimize the performance index. Solution to this min-max problem is obtained by approximating the solution of the Hamilton-Jacobi-Issac (HJI) equation. Artificial neural networks (NNs) are employed for the approximation in a forward-in-time and on-line manner. To neutralize the effect of the aperiodic availability of the state information on learning accuracy, a hybrid learning scheme is proposed.The local ultimate boundedness of the closed-loop event-triggered system is demonstrated using Lyapunov direct method and Zeno free behavior of the system is also guaranteed. Finally, simulation results are included to validate the proposed design. |
Year | Venue | Field |
---|---|---|
2018 | 2018 ANNUAL AMERICAN CONTROL CONFERENCE (ACC) | Affine transformation,Nonlinear system,Optimal control,Control theory,Computer science,Sampling (statistics),Control system,Aperiodic graph,Artificial neural network,Optimization problem |
DocType | ISSN | Citations |
Conference | 0743-1619 | 0 |
PageRank | References | Authors |
0.34 | 0 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Vignesh Narayanan | 1 | 10 | 5.19 |
Avimanyu Sahoo | 2 | 155 | 10.66 |
Sarangapani Jagannathan | 3 | 1136 | 94.89 |