Aerial VLN

VLN: Vision-Language Navigation

Aerial VLN studies how drones interpret language instructions, recognize relevant places or objects from onboard cameras, and navigate to them through open environments. Approaches range from modular vision-language models with spatial maps and memory to end-to-end vision-language-action policies, world models, and diffusion-based trajectory planners; many combine slower task reasoning with faster control. Benchmarks such as CityNav, TravelUAV, and OpenFly test instruction following across simulated scenes, unseen environments, and real flights. A central challenge is grounding long-horizon instructions reliably while maintaining safe, continuous control under viewpoint changes, limited onboard compute, and the gap between simulation and reality.

Latest papers

61 papers9 papers in the last four weeks, level with the four weeks before. 0.1% of all new papers.

OVerSeeC: Open-Vocabulary Costmap Generation from Satellite Images and Natural Language
Feb 20, 2026Rwik Rana, Jesse Quattrociocchi, Dongmyeong Lee +5Aerial VLN
The University of Texas at Austin, Austin, TX, USA. · DEVCOM Army Research Laboratory, United States.

Feb 20, 2026J/K move · Enter open · S save

OVerSeeC: Open-Vocabulary Costmap Generation from Satellite Images and Natural Language

Rwik Rana, Jesse Quattrociocchi, Dongmyeong Lee, Christian Ellis+4

The University of Texas at Austin, Austin, TX, USA. · DEVCOM Army Research Laboratory, United States.

Aerial VLN

Open paper AskPDF arXiv

Aerial imagery provides essential global context for autonomous navigation, enabling route planning at scales inaccessible to onboard sensing. We address the problem of generating global costmaps for long-range planning directly from satellite imagery when entities and mission-specific traversal rules are expressed in natural language at test time. This setting is challenging since mission requirements vary, terrain entities may be unknown at deployment, and user prompts often encode compositional traversal logic. Existing approaches relying on fixed ontologies and static cost mappings cannot accommodate such flexibility. While foundation models excel at language interpretation and open-vocabulary perception, no single model can simultaneously parse nuanced mission directives, locate arbitrary entities in large-scale imagery, and synthesize them into an executable cost function for planners. We therefore propose OVerSeeC, a zero-shot modular framework that decomposes the problem into Interpret-Locate-Synthesize: (i) an LLM extracts entities and ranked preferences, (ii) an open-vocabulary segmentation pipeline identifies these entities from high-resolution imagery, and (iii) the LLM uses the user's natural language preferences and masks to synthesize executable costmap code. Empirically, OVerSeeC handles novel entities, respects ranked and compositional preferences, and produces routes consistent with human-drawn trajectories across diverse regions, demonstrating robustness to distribution shifts. This shows that modular composition of foundation models enables open-vocabulary, preference-aligned costmap generation for scalable, mission-adaptive global planning.