DAZAI CHEN
2026

Smart Warehouse CAD-to-SimReady Digital Twin

An industrial OpenUSD / Omniverse warehouse scene that connects CAD-derived assets, SimReady warehouse components, operational event overlays, camera intent, and validation artifacts.

[Role] Industrial Digital Twin / CAD-to-OpenUSD Pipeline Developer

From CAD files to an inspection-ready warehouse twin

Industrial digital-twin work usually starts with fragments: CAD files, equipment models, warehouse layouts, camera locations, safety zones, and operational events. This project focuses on turning those fragments into scene structure that can be inspected.

This project builds a smart warehouse / inspection-cell digital twin in OpenUSD. The stage combines CAD-derived assets, NVIDIA warehouse / factory assets, camera intent, operational event overlays, and validation reports into one Omniverse-readable scene.

Smart warehouse OpenUSD digital twin overview with racks, receiving area, event overlays, and camera coverage

Warehouse overview. A composed OpenUSD stage with storage racks, receiving congestion, pallet-truck context, event beacons, and camera coverage markers.

Integration problem

Before

CAD files, warehouse asset packs, inspection targets, camera plans, and operations events live in separate tools. A finished render hides where those inputs came from.

After

Converted CAD, referenced warehouse assets, event JSON, camera coverage, safety zones, and validation checks become one OpenUSD stage that can be reviewed visually and programmatically.

Pipeline

CAD to SimReady OpenUSD pipeline diagram

Repeatable pipeline. CAD, warehouse assets, operational events, and validation reports are composed into a SimReady-oriented OpenUSD scene.

01 CAD / assets

NIST STEP parts, Kuka STEP source, warehouse asset packs, and SimReady components.

02 OpenUSD stage

A structured /SmartWarehouse stage with environment, storage, receiving, operations, inspection cell, sensors, and event overlays.

03 Event binding

Blocked aisle, near-miss, and defect-detected events are mapped into 3D positions with source IDs.

04 Validation

The stage is checked for required prims, authored references, up axis, scale, bounds, and event coverage.

Scene material

Receiving lane event overlay in the smart warehouse digital twin

Receiving event. Pallet congestion is mapped to a blocked-lane event while a pallet-truck path intersects the safety zone.

Kuka inspection cell and defect target inside the smart warehouse digital twin

Inspection cell. CAD-derived Kuka workcell, defect-detection target, converted part, and inspection overlay.

CAD to USD preview for warehouse inspection cell assets

CAD / USD check. The inspection cell uses converted or referenced engineering assets as part of the scene record.

Scene structure check

Passed

stage validation

1,298

validated prims in the current report

486

authored asset references

77m × 73m × 18m

normalized scene bounds

The validation script checks the default prim, units, up axis, required warehouse structure, authored references, nontrivial scene bounds, and event coverage. It also verifies three operational event types: blocked receiving lane, pallet-truck near miss, and surface defect detected.

BIM/VDC coordination layer

The warehouse scene already shows how CAD-derived assets, operational events, and validation checks become an OpenUSD stage. I added a BIM/VDC source path for the same reason: facility review also needs source element identity, not just a composed 3D view.

This extension uses a public buildingSMART IFC fixture as the BIM input, extracts entity/category metadata, assigns deterministic proxy bounds, checks representative fab-style equipment keepout envelopes, and writes the results as JSON, a Markdown issue report, and an OpenUSD-style USDA coordination layer.

BIM/VDC extension pipeline from public IFC source to metadata, proxy bounds, fab-style keepout checks, issue report, and OpenUSD-style coordination layer

IFC to coordination layer. The BIM layer keeps source identity, clearance checks, issue markers, and a USDA overlay readable beside the warehouse scene.

buildingSMART

public IFC fixture, CC BY 4.0

65

IFC entity types counted

14

product-like elements across 4 IFC categories

12

clash / clearance records from 3 equipment mockups

Generated materialWhat it preserves
element summary JSONsource URL, license, parser mode, IFC entity counts, GlobalId/name fields, proxy bounds
clash / clearance JSONequipment envelopes, issue IDs, severity, affected BIM element, distance, validation basis
USDA coordination layerBIM proxy elements, fab-style keepout boxes, and per-issue markers for OpenUSD-style composition

Source note. The IFC fixture is generic public architecture, not a semiconductor fab. The EUV scanner, wet bench, and AMHS stocker are representative keepout mockups. The boxes are deterministic proxy validation so the workflow stays reproducible; native IFC geometry and BCF-style issue export would be the next production upgrades.

Scope

Built / inspectable
  • OpenUSD stage composition script
  • CAD / STEP conversion checks for factory components
  • referenced warehouse and factory asset structure
  • event JSON mapped into scene coordinates
  • validation report and screenshot manifest
Simulated / scoped
  • public asset packs stand in for a customer facility
  • operational events are authored JSON
  • physics and collision proxies are a next-stage extension
  • deployment integration and live warehouse control remain outside this prototype

Where this workflow applies

This industrial companion to the smart-city traffic case connects most directly to Omniverse, OpenUSD, CAD/BIM/Revit import, SimReady assets, factory or warehouse simulation, synthetic data, robotics testing, and Physical AI workflows.

The practical value is straightforward: messy engineering / operations inputs become a structured 3D scene that can be inspected as data and reviewed visually.

Artifacts

NVIDIA Omniverse OpenUSD Python CAD / STEP USD Composer SimReady Assets NVIDIA Warehouse Assets