Integrating Visual Perception and Control Strategies in Custom Omnidirectional Mobile Robots

Area of Science:

• Robotics
• Computer Vision
• Control Systems

Background:

• Autonomous mobile robots are crucial for warehouse logistics optimization.
• Precise positioning and autonomous planning for robot docking remain significant technical challenges.

Purpose of the Study:

• To develop and evaluate a custom vision-based control system for an autonomous omnidirectional wheeled robot.
• To compare the effectiveness of Classic Position-Based Visual Servoing with a Dual Lie Algebra approach for robot docking.

Main Methods:

• Utilized a stereo camera integrated with the Robot Operating System (ROS) for visual feedback.
• Formulated and experimentally validated two visual feedback control laws: Classic Position-Based Visual Servoing and a Dual Lie Algebra method.
• Employed a quaternion-based approach for pose error minimization in Classic Position-Based Visual Servoing.

Main Results:

• Both control methods successfully enabled robot docking.
• The dual pose-free approach demonstrated more robust and effortless robot platform movement compared to Classic Position-Based Visual Servoing.
• The dual pose-free method ensured convergence towards the desired point-feature configuration by computing 3D visual sensor velocities.

Conclusions:

• Integrating depth-based feature recovery with advanced algebraic strategies provides a stable control strategy for automated industrial scenarios.
• The dual pose-free approach offers superior performance for autonomous robot docking maneuvers.
• Vision-based control systems are key to overcoming positioning challenges in autonomous logistics.