SceneSmith: Agentic Generation of Simulation-Ready Indoor Scenes

TL;DR

SceneSmith is a hierarchical framework that generates diverse, physically realistic indoor scenes from natural language prompts, significantly improving scene complexity and realism for robotic simulation.

Contribution

It introduces a novel agentic, multi-stage approach combining text-to-3D synthesis, dataset retrieval, and physical estimation to create detailed, simulation-ready indoor environments.

Findings

Generates 3-6x more objects than prior methods

Achieves less than 2% inter-object collisions

96% of objects remain stable under physics simulation

Abstract

Simulation has become a key tool for training and evaluating home robots at scale, yet existing environments fail to capture the diversity and physical complexity of real indoor spaces. Current scene synthesis methods produce sparsely furnished rooms that lack the dense clutter, articulated furniture, and physical properties essential for robotic manipulation. We introduce SceneSmith, a hierarchical agentic framework that generates simulation-ready indoor environments from natural language prompts. SceneSmith constructs scenes through successive stages$\unicode{x2013}$from architectural layout to furniture placement to small object population$\unicode{x2013}$each implemented as an interaction among VLM agents: designer, critic, and orchestrator. The framework tightly integrates asset generation through text-to-3D synthesis for static objects, dataset retrieval for articulated objects, and physical property estimation. SceneSmith generates 3-6x more objects than prior methods, with <2% inter-object collisions and 96% of objects remaining stable under physics simulation. In a user study with 205 participants, it achieves 92% average realism and 91% average prompt faithfulness win rates against baselines. We further demonstrate that these environments can be used in an end-to-end pipeline for automatic robot policy evaluation.