Adaptive sampling artificial-actual control for non-zero-sum games of constrained systems

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Nonlinear Dynamics

Background:

• Physical constraints on actuators pose challenges for controlling continuous nonlinear systems.
• Existing control methods struggle with symmetric and asymmetric input constraints in non-zero-sum games.
• Predicting system states and approximating system dynamics are crucial for effective control.

Purpose of the Study:

• To develop an artificial-actual control strategy for continuous nonlinear systems with input constraints.
• To address non-zero-sum games with symmetric and asymmetric input constraints.
• To enhance control efficiency and data communication through aperiodic sampling mechanisms.

Main Methods:

• Utilizing improved Elman dynamic neural networks (EDNNs) for artificial system modeling and state prediction.
• Designing a non-quadratic value function to handle diverse input constraints.
• Employing polynomial parameterized adaptive dynamic programming (ADP) to solve coupled Hamilton-Jacobi equations (HJEs).
• Introducing event-triggered mechanism (ETM), dynamic ETM (DETM), and self-triggered mechanism (STM) for adaptive sampling.

Main Results:

• Optimal control laws for two players were derived using ADP.
• The proposed sampling mechanisms (ETM, DETM, STM) ensure system stability and avoid the Zeno phenomenon.
• Simulations validated the effectiveness of the artificial-actual control algorithm.
• Unique characteristics of each sampling trigger mode were highlighted.

Conclusions:

• The artificial-actual control approach effectively manages input constraints in nonlinear systems.
• Adaptive dynamic programming and novel sampling strategies provide robust and efficient control solutions.
• The developed methods offer a promising direction for advanced control system design.