Data-Driven Finite-Horizon H_∞ Tracking Control With Event-Triggered Mechanism for the Continuous-Time Nonlinear...

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Optimization Theory

Background:

• Model-free control problems with optimal tracking are challenging.
• Finite-horizon H∞ optimal tracking with constrained inputs requires advanced methods.
• Event-triggered control offers potential for efficiency but faces challenges like Zeno behavior.

Purpose of the Study:

• To develop a neural network (NN)-based adaptive dynamic programming (ADP) event-triggered control method.
• To address the model-free finite-horizon H∞ optimal tracking control problem with constrained control input.
• To achieve a near-optimal control policy efficiently.

Main Methods:

• A data-driven model is established using a recurrent neural network (RNN) from input-output data.
• An augmented system with an event-triggered mechanism is developed.
• A novel event-triggering condition is proposed to avoid Zeno behavior.
• Time-dependent activation functions for NNs are considered due to time-dependent Hamilton-Jacobi-Isaacs (HJI) equations.

Main Results:

• A relationship between event-triggered and time-triggered HJI equations is established.
• The method finds real-time approximations of the optimal value.
• Uniform ultimate boundedness of the closed-loop system is ensured.
• Effectiveness is verified through two simulation examples.

Conclusions:

• The proposed NN-based ADP event-triggered control method effectively achieves near-optimal tracking control for model-free systems.
• The approach successfully handles constrained inputs and avoids Zeno behavior.
• The method demonstrates robustness and efficiency in simulations.