Multi-sensor fusion for mobile robot perception

Abstract

This paper presents a multi-sensor fusion perception system to overcome key limitations of single-line LiDAR in mobile robots, including pose drift, vertical blind zones, and poor robustness against transparent surfaces or sensor failures. The proposed architecture integrates LiDAR-IMU tightly-coupled odometry, RGB-D camera-based 3D mapping using ORB-SLAM2, and ultrasonic near-field detection. The design is expected to improve positioning accuracy and environmental awareness in indoor scenarios. The work is currently at the design phase, with physical implementation planned for future research.