Neural-Network-Based Adaptive Funnel Control for Servo Mechanisms With Unknown Dead-Zone

Area of Science:

• Control Systems Engineering
• Robotics
• Nonlinear Control Theory

Background:

• Servo mechanisms often exhibit unknown dead-zone characteristics, degrading performance.
• Traditional funnel control (FC) has limitations in systems with higher relative degrees.
• Backstepping methods can suffer from 'explosion of complexity'.

Purpose of the Study:

• To propose an adaptive funnel control (FC) scheme for servo mechanisms with unknown dead-zones.
• To improve transient and steady-state performance of servo systems.
• To address limitations of existing FC and backstepping techniques.

Main Methods:

• Developed a modified funnel variable using tracking error to overcome FC limitations.
• Applied error transformation for controller design.
• Integrated an improved funnel function within dynamic surface control.
• Utilized a novel command filter with Levant differentiator to avoid complexity explosion.
• Employed neural networks for approximating unknown dead-zone and nonlinear functions.

Main Results:

• The proposed adaptive funnel controller ensures output error stays within a predefined funnel boundary.
• Effectiveness validated through comparative experiments on a turntable servo mechanism.
• Successfully approximated unknown dead-zone and nonlinear functions using neural networks.

Conclusions:

• The devised adaptive funnel control method effectively handles unknown dead-zones in servo mechanisms.
• The approach improves both transient and steady-state performance.
• The combination of techniques offers a robust solution for complex control problems.