Full-state tracking control of a mobile robot using neural networks

Area of Science:

• Robotics
• Control Systems
• Artificial Intelligence

Background:

• Nonholonomic mobile robots present significant control challenges due to their complex and often unknown dynamics.
• Existing tracking controllers may suffer from issues like speed jumps and require extensive prior knowledge or training.
• Ensuring robust and stable trajectory following is crucial for autonomous navigation.

Purpose of the Study:

• To develop an efficient neural network-based tracking controller for nonholonomic mobile robots with completely unknown dynamics.
• To guarantee stable system performance and achieve precise trajectory tracking.
• To address limitations of previous controllers, such as the speed jump problem.

Main Methods:

• A mathematical model was developed for the nonholonomic mobile robot.
• A single-layer neural network was designed, leveraging robot regressor dynamics to linearize unknown nonlinear dynamics.
• Lyapunov theory was employed to rigorously prove system stability.
• The controller generates real-time smooth velocity control signals.

Main Results:

• The proposed neural network controller ensures guaranteed tracking performance and system stability.
• The approach effectively handles unknown robot dynamics and unmodeled disturbances without boundedness assumptions.
• The controller resolves the speed jump problem inherent in some prior tracking controllers.
• No offline training is required for the neural network.

Conclusions:

• The developed neural network controller offers an effective and efficient solution for trajectory tracking in nonholonomic mobile robots with unknown dynamics.
• The method provides guaranteed stability and robust performance, outperforming previous approaches.
• Simulation and comparison results validate the practicality and effectiveness of the proposed control strategy.