Event-Triggered Guarantee Cost Control for Partially Unknown Stochastic Systems via Explorized Integral Reinforcement...

Area of Science:

• Control Theory
• Machine Learning
• Stochastic Systems

Background:

• Stochastic systems with randomly time-varying parameters pose control challenges.
• Existing methods may require full system knowledge and incur high computational costs.

Purpose of the Study:

• To develop an integral reinforcement learning (IRL)-based event-triggered guarantee cost control (GCC) approach for stochastic systems.
• To reduce computational cost and resource waste by implementing an event-triggered mechanism.
• To relax the requirement of knowing system dynamics.

Main Methods:

• Integral Reinforcement Learning (IRL) and optimal zero-sum game formulation via Hamilton-Jacobi-Isaac (HJI) equation.
• Multivariate Probabilistic Collocation Method (MPCM) for predicting mean values of uncertain functions.
• Event-triggered GCC using explorized IRL, MPCM, and critic-actor-disturbance neural networks (NNs).

Main Results:

• A novel GCC method combining IRL and MPCM is proposed, relaxing system dynamics knowledge requirements.
• An event-triggered GCC approach using NNs reduces computation and resource usage.
• Ultimate boundedness of controlled systems is proven using Lyapunov theorem.

Conclusions:

• The developed event-triggered GCC approach effectively controls stochastic systems with randomly time-varying parameters.
• The method offers reduced computational load and resource efficiency compared to traditional approaches.
• The approach is validated through simulation examples.