Inverse Reinforcement Learning H _∞ Optimal Control for Takagi-Sugeno Fuzzy Systems

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Fuzzy Logic Systems

Background:

• Disturbances in control systems pose challenges for achieving optimal performance.
• Imitating expert behavior is crucial for developing effective control strategies.
• Takagi-Sugeno (T-S) fuzzy systems offer a framework for modeling complex nonlinear systems.

Purpose of the Study:

• To develop an inverse reinforcement learning (RL) H-infinity optimal control approach for T-S fuzzy systems.
• To reconstruct the expert system's cost function and imitate its behavior.
• To address scenarios where the learner system's dynamics are known or unknown.

Main Methods:

• A learner-expert framework with two inverse RL algorithms was proposed.
• Algorithms involve optimal policy updates using game algebraic Riccati equations (GAREs).
• Gradient descent correction and inverse optimal control iteration refine the control strategy.

Main Results:

• The developed algorithms were proven to be convergent.
• The fuzzy inverse RL optimal control methodology ensures asymptotic stability.
• A Nash equilibrium solution was achieved, demonstrating effective control.

Conclusions:

• The proposed fuzzy H-infinity optimal control method is effective for systems with disturbances.
• Application to an autonomous surface vehicle (ASV) system validated the methodology.
• The approach successfully reconstructs expert cost functions and imitates expert behavior.