robolab-scenegen

Generate USD scene files from natural language descriptions for robot manipulation benchmarks.

423|59|Updated Apr 8, 2026
One-click install
npx skills add https://github.com/NVlabs/RoboLab --skill robolab-scenegen
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: robolab-scenegen
Source: https://github.com/NVlabs/RoboLab/tree/main/skills/robolab-scenegen
Command: npx skills add https://github.com/NVlabs/RoboLab --skill robolab-scenegen

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires numpy, scipy, isaacsim, and includes references (resource) and assets (resource) components.

What problem does it solve?

This skill removes the manual effort of building complex 3D simulation environments by translating natural language descriptions into valid USD scene files for robot manipulation tasks.

Core Features & Use Cases

  • Natural Language Scene Creation: Describe a tabletop arrangement, and the skill automatically selects objects, computes collision-free placements, and generates the corresponding USDA file.
  • Automated Physics Settling: Integrates with IsaacSim to perform physics-based settling, ensuring objects are in stable, realistic resting positions.
  • Use Case: Quickly generate a variety of cluttered tabletop scenes for testing a robot's pick-and-place policy without manually positioning every object in a 3D editor.

Quick Start

Invoke the robolab-scenegen skill and provide a description of the objects you want on the table to begin the automated scene generation process.

Frequently Asked Questions about robolab-scenegen

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate USD scene files for robot manipulation benchmarks from text?

You can generate USD scene files by parsing natural language descriptions into spatial and physical constraints. This process automates object placement, collision-free configuration, and physics-based scene settling within the RoboLab ecosystem.

Does automated scene generation work with IsaacSim for physics settling?

Yes, automated scene generation integrates with IsaacSim to perform physics-based settling. This ensures objects in the generated USD scenes are in stable, realistic resting positions for simulation validation.

Can I create cluttered tabletop scenes for pick-and-place testing without manual 3D editing?

Yes, you can create cluttered tabletop scenes by providing a text description of the desired objects. The skill automatically selects objects, computes collision-free placements, and generates the corresponding USDA file.

Do I need an object catalog to automate object placement in USD scenes?

Yes, you need access to an object catalog to automate object placement. A compatible IsaacSim environment is also required to render and validate the physics-based scene settling.

What's the best way to automate collision-free configuration for robot simulation scenes?

Automating collision-free configuration is best handled by parsing text descriptions into spatial constraints. This computes valid object placements automatically, removing the manual effort of building complex 3D simulation environments.