Roman Titkov

Moscow Institute of Physics and Technology

Papers

Total Citations

H-Index

About

Roman Titkov is an emerging researcher working at the intersection of 3D scene understanding, natural language processing, and autonomous systems. His work focuses on open-vocabulary object grounding — the challenge of enabling intelligent agents to locate and identify objects based on complex natural language descriptions within three-dimensional environments. Titkov's most notable contribution is his development of a 3D scene graph-based approach that moves "Beyond Bare Queries," addressing a critical limitation in existing CLIP-based methods. While prior open-vocabulary techniques performed adequately with simple, unambiguous object queries, they struggled with contextual or relational descriptions requiring deeper spatial and semantic reasoning. By integrating structured 3D scene graphs, Titkov's framework enables autonomous agents to interpret nuanced, ambiguous language grounded in real-world spatial relationships — a meaningful step forward for robotics and embodied AI applications. This research has accumulated 13 citations across its versions, a promising indicator of early impact for work published as recently as 2024–2025. Titkov's contributions are particularly relevant for researchers developing next-generation perception systems in autonomous navigation, human-robot interaction, and scene understanding, where bridging language and 3D spatial comprehension remains an open and vital challenge.