ICP-Based Mapping and Localization System for AGV with 2D LiDAR

Area of Science:

• Robotics
• Computer Vision
• Sensor Fusion

Background:

• Existing SLAM systems often require multiple sensors or significant computational resources.
• Automated Guided Vehicles (AGVs) need reliable perception for autonomous operation.
• Low-complexity, sensor-independent solutions are crucial for embedded robotic platforms.

Purpose of the Study:

• To develop a functional, real-time SLAM system for AGVs using solely a 2D LiDAR sensor.
• To create a low-complexity system adaptable to embedded platforms with limited computational power.
• To address limitations in current literature regarding sensor independence and resource efficiency.

Main Methods:

• Integration of scan alignment using the Iterative Closest Point (ICP) algorithm.
• Utilized Gauss-Newton optimization and the point-to-plane metric for pose estimation.
• Developed a lightweight graphical interface for real-time visualization of sensor data, pose, and map.

Main Results:

• Achieved high pose estimation accuracy: 99.42% (x), 99.6% (y), and 99.99% (θ).
• Demonstrated effective localization and progressive environment mapping in a controlled setting.
• System proved functional for robotic applications despite a moderate update rate.

Conclusions:

• The developed SLAM system offers a robust and adaptable foundation for mobile platforms.
• The modular architecture supports future extensions like trajectory planning and control.
• Potential applications include industrial automation, academic research, and robotics education.