Learning Self-Triggered Controllers With Gaussian Processes.

Area of Science:

• Control Systems Engineering
• Machine Learning
• Networked Systems

Background:

• Networked Control Systems (NCSs) present challenges due to unknown plant dynamics.
• Designing effective controllers for NCSs requires addressing uncertainties in system behavior.

Purpose of the Study:

• To design self-triggered controllers for NCSs with a priori unknown plant dynamics.
• To jointly optimize control and communication policies within the NCS framework.

Main Methods:

• Employed Gaussian Process (GP) regression to learn unknown plant dynamics.
• Formulated an optimal control problem for joint policy design.
• Utilized a reinforcement learning framework for implementation and learning.

Main Results:

• Successfully learned plant dynamics using GP regression.
• Developed a method for jointly designing control and communication policies.
• Demonstrated the effectiveness of the proposed approach through numerical simulations.

Conclusions:

• The proposed self-triggered controller design is effective for NCSs with unknown dynamics.
• The integration of GP regression and reinforcement learning offers a robust solution.
• The approach enables adaptive and efficient control in uncertain networked environments.