Towards Embodied Cognition in Robots via Spatially Grounded Synthetic Worlds
This work addresses the need for spatial understanding in interactive human-robot scenarios, but it is incremental as it presents a conceptual framework and dataset as a first step toward broader embodied AI systems.
The paper tackles the problem of enabling robots to perform Visual Perspective Taking (VPT) for embodied cognition by introducing a synthetic dataset generated in NVIDIA Omniverse, which includes RGB images, natural language descriptions, and ground-truth transformation matrices for supervised learning of spatial reasoning tasks, with a focus on inferring Z-axis distance as a foundational skill.
We present a conceptual framework for training Vision-Language Models (VLMs) to perform Visual Perspective Taking (VPT), a core capability for embodied cognition essential for Human-Robot Interaction (HRI). As a first step toward this goal, we introduce a synthetic dataset, generated in NVIDIA Omniverse, that enables supervised learning for spatial reasoning tasks. Each instance includes an RGB image, a natural language description, and a ground-truth 4X4 transformation matrix representing object pose. We focus on inferring Z-axis distance as a foundational skill, with future extensions targeting full 6 Degrees Of Freedom (DOFs) reasoning. The dataset is publicly available to support further research. This work serves as a foundational step toward embodied AI systems capable of spatial understanding in interactive human-robot scenarios.