A neuro-fuzzy controller for mobile robot navigation and multirobot convoying.

Area of Science:

• Robotics
• Artificial Intelligence
• Control Systems

Background:

• Traditional control methods struggle with complex nonlinear dynamic systems.
• Integrating human expertise (fuzzy logic) with adaptive learning (neural networks) offers a promising approach.

Purpose of the Study:

• To develop and implement a Neural integrated Fuzzy conTroller (NiF-T) for nonlinear dynamic control problems.
• To demonstrate the real-time application of NiF-T in autonomous mobile robot navigation and multi-robot convoying.

Main Methods:

• The NiF-T architecture combines Fuzzy logic Membership Functions (FMF), a Rule Neural Network (RNN), and an Output-Refinement Neural Network (ORNN).
• FMF fuzzify sensory inputs, RNN interpolates fuzzy rules, and ORNN refines the output, with adjustable weights for online tuning.
• The system was tested on wall following, hall centering, and multi-robot convoying behaviors.

Main Results:

• NiF-T successfully enabled autonomous mobile robot navigation and multi-robot convoying behaviors.
• Complex behaviors like wall following and hall centering were achieved with a minimal number of fuzzy rules (5 and 9, respectively).
• Both RNN and ORNN components required limited training iterations (hundreds and less than one hundred, respectively) to learn the control rules.

Conclusions:

• The Neural integrated Fuzzy conTroller (NiF-T) provides an efficient and effective solution for nonlinear dynamic control problems.
• NiF-T demonstrates strong adaptability and learning capabilities, achieving complex robotic behaviors with minimal rules and training.
• This approach offers a powerful framework for developing intelligent control systems in robotics and other dynamic applications.