Adaptive Critic Learning-Based Robust Control of Systems with Uncertain Dynamics.

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Machine Learning

Background:

• Model uncertainties are inherent in control systems due to factors like imperfect modeling, external disturbances, and system nonlinearities.
• Addressing these uncertainties is crucial for reliable and effective system performance.

Purpose of the Study:

• To develop a novel method for robust control of uncertain systems.
• To transform the robust control problem into an optimal control problem for a nominal system.
• To implement an adaptive critic learning algorithm for online optimal control solution discovery.

Main Methods:

• The robust control problem is reformulated as an optimal control problem for a nominal system by selecting a suitable cost function.
• An adaptive critic learning algorithm is developed, utilizing only a critic neural network (NN) and omitting the actor NN common in prior research.
• Feasibility analysis of the proposed control algorithm is conducted.

Main Results:

• The proposed method successfully transforms the robust control problem into an optimal control problem.
• The adaptive critic learning algorithm, using only a critic NN, effectively learns the optimal control solution online.
• Simulation results validate the availability and effectiveness of the presented control method for uncertain systems.

Conclusions:

• The novel method provides an effective approach to robust control for systems with uncertainties.
• The simplified adaptive critic learning algorithm (critic NN only) offers a viable alternative to existing methods.
• The study demonstrates the practical applicability of the proposed control strategy.