Near Optimal Event-Triggered Control of Nonlinear Discrete-Time Systems Using Neurodynamic Programming

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Nonlinear Dynamics

Background:

• Controlling uncertain nonlinear discrete-time systems presents significant challenges.
• Traditional control methods often require frequent updates, leading to high computational load.
• Neurodynamic programming (NDP) offers a framework for optimal control but can be computationally intensive.

Purpose of the Study:

• To develop an event-triggered near-optimal control strategy for uncertain nonlinear discrete-time systems.
• To reduce computational burden by minimizing control updates.
• To ensure system stability and desired approximation accuracy.

Main Methods:

• Utilized event-driven neurodynamic programming (NDP) for control policy design.
• Employed a neural network (NN)-based identifier with event-based data to learn system dynamics.
• Implemented an actor-critic framework for learning cost functions and optimal control inputs.
• Derived an adaptive event-trigger condition to determine update instants.
• Applied Lyapunov techniques to guarantee closed-loop system boundedness.

Main Results:

• Achieved near-optimal control performance without traditional value or policy iterations.
• Demonstrated significant computational reduction through analysis of nontrivial inter-event times.
• Validated controller performance via simulation results.
• Successfully developed an event-driven NDP approach.

Conclusions:

• The proposed event-driven NDP method effectively controls uncertain nonlinear discrete-time systems.
• The adaptive event-triggering mechanism ensures computational efficiency and desired accuracy.
• The approach guarantees system stability, offering a practical solution for complex control problems.