Latest Embodied AI Research

Explores the critical infrastructure requirements for deploying embodied AI systems in developing regions, addressing computational resources, connectivity challenges, and sustainable implementation strategies for scientific research applications.

Presents a novel framework combining vision, language, and action models to enable robots to reason about physical interactions, spatial relationships, and task planning in complex real-world environments.

Introduces a simulation-to-reality pipeline for training robotic hands to perform complex manipulation tasks, achieving human-level dexterity through physics-informed learning and domain randomization techniques.

Proposes a hierarchical navigation system that builds semantic memory representations of environments, enabling robots to efficiently navigate and locate objects in large-scale indoor spaces with improved generalization.

Investigates the social dynamics of human-robot collaboration, developing embodied agents that can interpret social cues, adapt to human preferences, and coordinate effectively in shared workspace scenarios.