Carlos Roberto Cueto Zumaya

3D Scene Graphs (3DSGs) have emerged as a powerful representation for robotic perception and reasoning, unifying geometry, semantics, and relationships into a structured, multi-resolution model of the environment. While recent frameworks have demonstrated their utility for navigation, planning, and interaction, they remain constrained by limitations in scalability, real-time adaptability, and long-term support.

This research aims to advance 3DSG-based situational awareness by addressing one of the primary barriers to real-world adoption: scalability and usability. We aim to develop a management system that leverages mature, large-scale storage technologies—such as graph and vector databases—for efficient storage, querying, and updates. Designed with deployment in mind, the system will make 3DSGs directly applicable to downstream tasks enabling real-time decision-making.

By improving how 3DSGs are maintained and exploited, this work seeks to enhance autonomous systems’ ability to adapt to evolving environments, reason over spatial–semantic structures, and operate reliably over extended periods. Expected outcomes include robust situational awareness, improved decision-making, and greater operational resilience, with applications spanning autonomous vehicles, field robotics, and smart infrastructure.