On Stabilization of Quantized Sampled-Data Neural-Network-Based Control Systems.

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Systems Science

Background:

• Stabilization of complex systems is crucial for reliable operation.
• Previous research often overlooks communication constraints like state quantization.
• Neural network-based controllers offer advanced capabilities but require robust stabilization.

Purpose of the Study:

• To investigate the stabilization of sampled-data neural-network-based systems with state quantization.
• To introduce a novel approach for handling communication limitations in control systems.
• To derive less conservative conditions for system stabilization and performance guarantees.

Main Methods:

• Development of a novel piecewise Lyapunov-Krasovskii functional (LKF).
• Incorporation of a line-integral type Lyapunov function and sampling pattern information.
• Design of quantized sampled-data controllers using linear matrix inequality (LMI) approach.
• Utilization of a search algorithm for tuning parameter optimization.

Main Results:

• Reduced conservativeness in stabilization conditions compared to existing methods.
• Derivation of conditions for maximizing the sampling period.
• Establishment of minimal guaranteed cost control performance.
• Successful demonstration on an inverted pendulum system.

Conclusions:

• The proposed method effectively stabilizes sampled-data neural-network-based systems with state quantization.
• The novel LKF and LMI approach significantly reduce conservativeness.
• The method allows for maximizing sampling periods and ensuring control performance.