Research Preview · Logistics
Teleoperated Parcel Sorting
A data pipeline that combines VR teleoperation, motion mapping, simulation, world-model-generated data and few-shot deployment.
Business Context
Parcel operations include long-tail object variation that is expensive to label frame by frame.
Operational Challenge
The system must collect useful demonstrations, transfer motion to the robot and generalize across parcel shapes and views.
System Configuration
VR teleoperation, motion retargeting, simulation and real-data fusion, BridgeV2W and VLA training.
Deployment Conditions
Selected parcel classes, configured sorting cells and trained exception handling.
Workflow
- VR Teleoperation
- Motion Mapping
- Simulation and Real Data Fusion
- World Model Generated Data
- VLA Training
- Few-Shot Deployment
Core technology
- Teleoperation Data
- BridgeV2W
- Cross-View Augmentation
- FAM
Measured value
Validation is reported by parcel set, view condition, cycle definition and recovery policy.
Data collection without frame-by-frame action labels
Operator motion is mapped into robot-compatible demonstrations. Simulation and generated cross-view data expand the training distribution before a smaller amount of real-robot adaptation.