scholarly journals Event-Triggered Optimal Neuro-Controller Design With Reinforcement Learning for Unknown Nonlinear Systems

2019 ◽  
Vol 49 (9) ◽  
pp. 1866-1878 ◽  
Author(s):  
Xiong Yang ◽  
Haibo He ◽  
Derong Liu
Author(s):  
A. Ghafoor ◽  
J. Yao ◽  
S. N. Balakrishnan ◽  
J. Sarangapani ◽  
T. Yucelen

In this paper, a novel event triggered neural network (NN) adaptive controller is presented for uncertain affine nonlinear systems. Controller design is based on an observer, called as Modified State Observer (MSO), which is used to approximate uncertainties online. State is sensed continuously yet sent on feedback network only when required, in aperiodic fashion. Lyapunov analysis is used to derive this condition which is dynamic in nature since it is based on tracking error. In this way ETNAC helps to not only saves communication cost but also computational efforts. MSO formulations have two tunable gains which let you do fast estimation without inducing high frequency oscillations in the system. A benchmark example of 2-link robotic manipulator is used to show the efficacy of the proposed controller.


Sign in / Sign up

Export Citation Format

Share Document