Adaptive Neuro-Fuzzy Control of a Spherical Rolling Robot Using Sliding-Mode-Control-Theory-Based Online Learning...

Area of Science:

• Robotics
• Control Systems Engineering
• Artificial Intelligence

Background:

• Conventional controllers struggle with unmodeled dynamics and parameter variations.
• Adaptive control is crucial for systems with uncertain or changing parameters.

Purpose of the Study:

• To develop an adaptive neuro-fuzzy controller (ANFC) for a spherical rolling robot.
• To enhance system stability and performance using a sliding-mode control (SMC)-based learning algorithm.

Main Methods:

• Implemented a hybrid control structure combining a neuro-fuzzy network and a conventional controller.
• Derived ANFC parameter updating rules using SMC theory.
• Proved learning stability via Lyapunov function analysis.

Main Results:

• The proposed ANFC with SMC learning eliminated steady-state error.
• Transient response performance of the spherical rolling robot was significantly improved.
• Effective control was achieved without prior knowledge of the robot's dynamic equations.

Conclusions:

• The adaptive neuro-fuzzy controller with SMC-based learning offers robust and efficient control for spherical rolling robots.
• This method overcomes limitations of conventional controllers in uncertain environments.
• It demonstrates a practical approach for adaptive control in complex robotic systems.