State observer-based Physics-Informed Machine Learning for leader-following tracking control of mobile robot

Area of Science:

• Robotics
• Control Systems
• Machine Learning

Background:

• Mobile robot coordination requires precise tracking control.
• Estimating leader robot (LR) dynamics is crucial for follower robot (FR) performance.
• Existing methods face challenges with uncertain models and real-time estimation.

Purpose of the Study:

• To propose a novel leader-following tracking control method for mobile robots.
• To develop an accurate speed estimation technique for the leader robot.
• To design a robust and stable controller for the follower robot.

Main Methods:

• Physics Informed Machine Learning (PIML) is employed for state observer dynamics learning and LR speed estimation.
• An error state model is utilized within PIML for stable learning.
• A parameter-dependent controller gain is determined using convex combination and polytopic model-based robust control.

Main Results:

• The proposed PIML-based state observer effectively estimates the leader robot's speed.
• The parameter-dependent controller ensures stable and accurate tracking performance.
• Simulations and experiments validate the effectiveness of the leader-following control strategy.

Conclusions:

• The novel leader-following control method enhances mobile robot tracking capabilities.
• PIML offers a robust approach for state estimation in uncertain robotic systems.
• The integrated control strategy demonstrates practical applicability through experimental validation.