Tracking Control for Linear Discrete-Time Networked Control Systems With Unknown Dynamics and Dropout.

Area of Science:

• Control Systems Engineering
• Networked Control Systems
• Adaptive Control

Background:

• Networked control systems (NCSs) face challenges like data dropout due to network-induced issues.
• System dynamics are often unknown in practical NCS applications, complicating control design.

Purpose of the Study:

• To develop a robust optimal control tracking method for NCSs with unknown dynamics and network-induced dropout.
• To design a novel dropout Smith predictor for state estimation in NCSs.
• To present adaptive algorithms for online optimal control learning.

Main Methods:

• A novel dropout Smith predictor was developed for state prediction with communication delays and data loss.
• A dropout generalized algebraic Riccati equation was derived to find the optimal tracker solution.
• Off-line policy iteration (PI), online PI, and Q-learning PI algorithms were presented.
• Q-learning was employed for adaptive online optimal control using reinforcement learning.

Main Results:

• The quadratic form of the performance index is preserved despite data dropout.
• The proposed methods achieve proper optimal tracking performance for NCSs with unknown dynamics and dropout.
• Q-learning adaptively learns the optimal control online without prior system knowledge.

Conclusions:

• The developed methods effectively address optimal control tracking in NCSs with dropout and unknown dynamics.
• The Q-learning approach offers a data-driven, adaptive solution for real-world NCS applications.
• The study demonstrates the feasibility of achieving robust control performance under challenging network conditions.