Semantic Scene Representations
Authors: Christian Rauch (TU Leoben) , Linus Nwankwo (TU Leoben) , Shail Jadav (TU Wien) , Jun Zhang (TU Graz)
Autonomous systems operating in real-world environments around humans face fundamental challenges in generalisability and explainability. As the complexity of these systems and their expected tasks grows, there is an increasing need for high-level interfaces grounded in semantic concepts from natural language. Language-grounded representations promise to facilitate scene understanding, enable generalisation to unseen environments and tasks, improve explainability of autonomous agent actions to human operators, and increase accessibility of complex systems for the general public without requiring domain-specific expertise. This workshop presents and discusses recent advances in semantic and naturallanguage- based 3D scene and environment representations, as well as localisation and mapping techniques, as foundational building blocks for interactive robotic tasks. The workshop brings together researchers working at the intersection of computer vision, robotics, and natural language processing to address open problems in open-set perception, multi-modal scene understanding, and semantic navigation.
Keywords:
How to Cite: Rauch, C. , Nwankwo, L. , Jadav, S. & Zhang, J. (2026) “Proceedings of the Semantic Scene Representations (SSR)Workshop at AIROV 2026”, Proceedings of the Austrian Symposium on AI, Robotics, and Vision. 3(1).