Core Concepts
The author proposes a CNN-based adaptive controller that learns control policies online and guarantees stability through Lyapunov analysis.
Abstract
The content introduces a CNN-based adaptive controller for uncertain nonlinear systems. It discusses the design, weight adaptation law, stability analysis, simulations, and comparisons with DNN controllers. The proposed method shows promising results in tracking control of uncertain systems.
Key points include:
- Introduction of a CNN-based adaptive controller.
- Derivation of weight adaptation laws using gradient descent optimization.
- Stability analysis based on Lyapunov functions.
- Comparative simulations with different variations of the proposed controller and DNN controllers.
- Discussion on the effects of design parameters and comparison between CNN and DNN controllers.
The study demonstrates the effectiveness of the proposed CNN-based controller in tracking control tasks for uncertain nonlinear systems.
Stats
"ks = 1, ρ = 105, Ac = -10I"
"Ts = 0.1, n0 = 10, α1 = 100, α2 = 0.01"
"CNN1 had 244 total weights"
"DNN had 250 total weights"
Quotes
"The proposed controller learned the desired control policy and provided asymptotic convergence without using pretrained data."
"A simulation study demonstrated that the proposed CNN-based end-to-end controller ensures asymptotic convergence of tracking error during online adaptation."
"The difference in performance between DNN and CNN1 was not significant in the simulation results."