Zum Inhalt springen
Technische Universität München

3D Vision and Digital Twining

Enabling robots to understand and interact with the 3D world is at the core of modern autonomous systems. Our research focuses on building reliable and scalable 3D vision solutions for localization, scene reconstruction, and environmental understanding. We aim to enable robots to accurately localize themselves while maintaining consistent and up-to-date representations of their surroundings, even in complex and dynamic environments. By combining information from multiple sensing modalities, our approaches provide robust and reliable perception in real-world conditions. Beyond localization, we work towards the efficient creation of digital twins—continuously updated virtual representations of the physical world that provide a consistent and up-to-date understanding of the environment over time. These capabilities enable a wide range of applications, including autonomous navigation, scene understanding and inspection, and robot collaboration. 

Key Topics
  • Visual and Visual-inertial Simultaneous Localization and Mapping (SLAM)
  • Learning-based and Data-driven Approaches for Localization and Mapping
  • Efficient Digital Twin Generation and High-quality 3D Scene Reconstruction
  • Foundation Models for Semantic Scene Understanding
  • Geometric, Semantic and Temporal Scene Modeling (including scene graphs)
  • Computationally Efficient SLAM for Resource-constrained Platforms
  • Communication-efficient SLAM for Remote and Multi-agent Systems
Our Work

We are currently working on the following projects:

Key Publications
  1. Su, Xin; Eger, Sebastian; Misik, Adam; Yang, Dong; Pries, Rastin; Steinbach, Eckehard: HPF-SLAM: An Efficient Visual SLAM System Leveraging Hybrid Point Features. 2024 IEEE International Conference on Robotics and Automation (ICRA), 2024.
  2. Su, Xin; Zhang, Xiaoang; Pries, Rastin; Steinbach, Eckehard: HPGS-SLAM: Hybrid Point-Guided Dense Visual SLAM with Online Mapping via Gaussian Splatting. IEEE Robotics and Automation Letters (RA-L), 2026.
  3. Su, Xin; Eger, Sebastian; Li, Jia; Misik, Adam; Yang, Dong; Pries, Rastin; Steinbach, Eckehard: AFDI-SLAM: Visual-Inertial SLAM with Adaptive Frame Control and Deep-Learning-based IMU Signal Denoising. IEEE Transactions on Multimedia, 2026.
  4. Wang, Zhenyu; Zhang, Yunzhou; Xu, Xiao; Xiong, Mengchen; Su, Xin; Meng, Fanle: CMIF-VIO: A Novel Cross Modal Interaction Framework for Visual Inertial Odometry. IEEE Robotics and Automation Letters 10 (2), 2025.
  5. Ni, Zhifan; Steinbach, Eckehard: REVNET: Rotation-Equivariant Point Cloud Completion via Vector Neuron Anchor Transformer. 28th International Conference on Pattern Recognition (ICPR 2026), 2026 
Contact

If you are interested in 3D Vision and Digital Twining or would like to learn more about our work, feel free to reach out to Xin Su, M.Sc. You are also welcome to get in touch directly with the researchers working in this field:

To top