Paper Info
Title
Neurodynamic programming and tracking control scheme of constrained-input systems via a novel event-triggered PI algorithm.
Abstract
In this paper, a new event-triggered approach for neurodynamic programming and optimal tracking control problem of constrained-input systems is proposed, where the desired trajectory can be generated as a large class of useful command trajectories. Firstly, the complex tracking problem is converted to a stabilizing control optimization problem by reconstructing a novel augmented tracking system with discounted performance index. Secondly, instead of the conventional time-driven control, an event-triggered policy iteration (PI) algorithm is designed to drive the system dynamic to track the reference trajectory, which requires less computation and fewer transmissions during the solving. Thirdly, the novel tracking control can be bounded as desired, which overcomes the unconstrained steady-state control from the general adaptive dynamic programming based tracking solution. Moreover, only critic neural network is used in the implementation of iteration learning, which simplifies the actor-critic architecture and reduces computational load. And by the means of Lyapunov method, the ultimately boundedness of the tracking system under the event-triggered PI algorithm is proved. Finally, the developed approach is applied to track a sinusoidal waveform and a periodic rectangular step signal in the simulations, where the effectiveness is also demonstrated.
Year
DOI
Venue
2019
10.1016/j.asoc.2019.105629
Applied Soft Computing
Keywords
DocType
Volume
Optimal control,Policy iteration,Event triggered control,Neurodynamic programming,Tracking control
Journal
83
ISSN
Citations 
PageRank 
1568-4946
1
0.36
References 
Authors
0
4
Name
Order
Citations
PageRank
Kun Zhang120.71
H Zhang27027358.18
Xiaodong Liang310.36
xie410636.98