How should an agent search visual or geospatial space when observations are partial and costly? Our reinforcement learning research targets active search, exploration, and sequential decision-making in visual domains. Recent work includes diffusion-guided visual active search in partially observable environments (DiffVAS), goal modality agnostic active geo-localization (GOMAA-Geo), and partially-supervised frameworks for visual active search. We study interpretable policies, hindsight-observable POMDPs, and learning from limited supervision—often paired with geospatial embeddings and localization methods for real-world deployment.
Spotlight publications tagged for this research area.