Online Data-Driven-Based Optimal Output Tracking Control Without Initial Stabilizing Policy.

Area of Science:

• Control Systems Engineering
• Adaptive Control Theory
• Machine Learning for Control

Background:

• Optimal output tracking is crucial for linear systems, but challenging with unknown system dynamics.
• Existing adaptive control methods often require specific conditions like initial stabilizing policies or data storage.

Purpose of the Study:

• To develop a novel adaptive dynamic programming algorithm for optimal output tracking control.
• To address the challenge of unknown system models in continuous-time linear systems.
• To overcome limitations of existing adaptive control techniques.

Main Methods:

• Integration of adaptive dynamic programming and optimal control theory.
• Proposal of a dual policy iteration (PI) learning algorithm with two PI schemes.
• Adaptive learning of the optimal tracking controller without prior stabilizing policies.

Main Results:

• The proposed dual PI algorithm successfully learns the optimal tracking controller.
• The algorithm converges without requiring an initial stabilizing policy, persistence of excitation, or historical data.
• Simulation results confirm the algorithm's effectiveness and superiority over existing methods.

Conclusions:

• The developed algorithm offers a robust solution for optimal tracking control in systems with unknown models.
• It provides a significant advancement by removing restrictive conditions common in prior adaptive control approaches.
• The method demonstrates practical applicability and improved performance in simulation studies.