Abstract

This paper presents a framework for semantic mapping and autonomous navigation using low-cost LiDAR and RGB cameras. Our approach integrates real-time object detection from camera images with LiDAR-based mapping to generate a semantically enriched map, where detected objects are represented as landmarks with labels and images. The semantic map is then leveraged by the ROS2 Navigation2 (Nav2) stack to enable goal-directed autonomous navigation, allowing the robot to plan paths not only around obstacles but also toward specific objects of interest. The proposed method provides a low-cost, scalable, and practical solution for intelligent robot navigation in environments with visually identifiable objects.