Industrial Data Loop

Data hub Channels Ring CORE DMP Hub Org · Version · ACL Distribute · Return CHANNEL BUS DL DATASET API SERVICE DDS STREAM Pull · Write-back · Stream 01 Annotation Annotation Platform 02 Algorithm Algorithm Platform 03 Simulation Simulation Platform Return META → DMP DMP → ring Return DMP distributes via Channel Bus; metadata closes the loop

DMP

Power the industrial data loop with an automated data engine

Training / fine-tuning / simulation / annotation — organized, distributed, and returned through DMP

Data Management Platform

The loop is centered on DMP

DMP drives annotation, training, fine-tuning, and simulation—it is the hub for organization, versioning, permissions, distribution, and return.

How DMP connects to other platforms

  • Download Versioned datasets, manifest validation, bulk download
  • API HTTP service APIs for pull and metadata write-back
  • DDS Real-time / near-real-time streaming (vehicle, robot, simulation)

Industries

Industry scenarios

Starting with autonomous driving and embodied AI: pain points enter DMP, ring modules consume, results return.

Autonomous driving

Multi-source capture is hard to align; versions drift across annotation, training, and simulation.

Trips and sensor packs are versioned in DMP; annotation and training bridges consume manifests; simulation results and hard cases return for traceability.

Related modules →

Embodied AI

Robot demo data is fragmented; simulation and real-world distributions diverge.

Demo trajectories and perception frames are organized in DMP; training and simulation bridges share one data contract; real-device feedback returns via DDS / API.

Related modules →

Start here

Continue from capability entries

The portal explains the loop; modules provide downloads and links; docs hold concepts, interfaces, and releases.