Quantization-Based Adaptive Actor-Critic Tracking Control With Tracking Error Constraints.

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Nonlinear Dynamics

Background:

• Investigates adaptive actor-critic (AC) tracking control for continuous-time nonlinear systems.
• Addresses challenges of unknown nonlinearities and quantized inputs, common in real-world applications.

Purpose of the Study:

• To develop an AC control strategy that enforces tracking error constraints within time-varying boundaries.
• To enhance control performance by introducing novel critic functions and a specialized controller structure.
• To guarantee system stability and performance despite uncertainties and disturbances.

Main Methods:

• Employs a tracking error transformation technique to create an augmented error system.
• Designs specific critic functions to supervise tracking performance and tune neural network (NN) weights.
• Develops a novel adaptive controller to mitigate the impact of NN reconstruction errors, input quantization, and disturbances.
• Utilizes Lyapunov stability theory to ensure the boundedness of closed-loop signals.

Main Results:

• Demonstrates the ability to maintain tracking errors within predefined time-varying boundaries.
• Achieves improved tracking performance compared to existing reinforcement learning-based methods.
• Confirms the boundedness of all closed-loop signals through rigorous stability analysis.
• Validates the effectiveness of the proposed adaptive control method via simulations on a two-pendulum system.

Conclusions:

• The proposed adaptive actor-critic control strategy effectively handles nonlinear systems with unknown dynamics and quantized inputs.
• The novel approach ensures tracking error constraints are met and enhances overall system performance and stability.
• Simulation results confirm the practical applicability and robustness of the developed control technique.