This study enhances robot navigation by integrating Large Language Models (LLMs) with ROS move_base, using osmAG for semantic mapping. The approach combines human-like contextual understanding with traditional navigation, improving adaptability and robustness in complex environments.
@article{zhang2025llm,title={Intelligent LiDAR Navigation: Leveraging External Information and Semantic Maps with LLM as Copilot},author={Xie, Fujing and Zhang, Jiajie and Schwertfeger, Sören},journal={IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)},year={2025},publisher={IEEE},url={https://arxiv.org/abs/2409.08493},dimensions={true},}
Generation of Indoor Open Street Maps for Robot Navigation from CAD Files
Jiajie Zhang, Shenrui Wu, Xu Ma, and 1 more author
This paper presents a complete and automated system to convert architectural Computer-Aided Design (CAD) files into a hierarchical topometric OpenStreetMap (OSM) representation for robust, life-long robot navigation. The method overcomes the limitations of traditional SLAM by leveraging permanent structural information, employing a multi-stage pipeline that includes topological segmentation, semantic text association, and multi-floor fusion to generate semantically rich and topologically correct maps.
@article{zhang2025cad2osm,title={Generation of Indoor Open Street Maps for Robot Navigation from CAD Files},author={Zhang, Jiajie and Wu, Shenrui and Ma, Xu and Schwertfeger, Sören},journal={arXiv preprint arXiv:2507.00552},year={2025},month=jul,url={https://arxiv.org/abs/2507.00552},dimensions={true},}
2024
Neural Surfel Reconstruction: Addressing Loop Closure Challenges in Large-Scale 3D Neural Scene Mapping
This paper introduces a novel neural surfel-based method for large-scale 3D scene reconstruction, addressing challenges in loop closure and bundle adjustment. By integrating neural descriptors with surfels and optimizing surfel associations, the approach achieves improved reconstruction accuracy and significantly reduces file sizes compared to traditional methods.
@article{zhang2024neural,title={Neural Surfel Reconstruction: Addressing Loop Closure Challenges in Large-Scale 3D Neural Scene Mapping},author={Cui, Jiadi and Zhang, Jiajie and Schwertfeger, Sören},journal={MDPI Sensors},volume={24},number={21},pages={6919},year={2024},month=oct,publisher={MDPI},doi={10.3390/s24216919},url={https://www.mdpi.com/1424-8220/24/21/6919},dimensions={true},}
