Cross-Docking Task Data

What makes cross-docking different

Most warehouse data is organized around inventory: goods arrive, rest in a location and are later picked. Cross-docking removes the rest. A trailer or container is unloaded, its contents are identified, sorted by outbound destination, staged near an outbound door and loaded, often within a short window. The warehouse acts as a transfer point instead of a store.

That changes the data a buyer needs. There are no storage locations to confirm and no pick faces to read. The central questions are about flow: which unit goes to which door, what happens when the plan and the load disagree, and what the worker does when an outbound vehicle is late. The physical work overlaps with trailer and container unloading and trailer loading and load building, but cross-docking adds the middle: reading, deciding and staging under time pressure.

The order-picking review used here only as general background does not cover this setting; it reviews layout, storage assignment, routing, batching and zoning for picking inside warehouses (de Koster et al., 2007). This page therefore reasons from how the process is organized and turns it into diligence questions.

Task anatomy

Operations differ by freight type, so confirm the following with the supplier:

1. Inbound arrival and door assignment, based on an appointment or an advance shipping notice.
2. Unload, by hand, conveyor or powered equipment.
3. Identify each unit from its label, as with the barcode scanning workflows.
4. Decide the destination from a system instruction, a sort sheet or a worker's knowledge.
5. Move and stage at or near the outbound door, by lane, pallet position or cart.
6. Consolidate units for the same destination, sometimes building mixed pallets.
7. Load outbound against the plan; see trailer loading and load building.
8. Handle exceptions (see warehouse exception handling): damaged units, missing labels, overages, shortages and unplanned freight.

Where the freight is parcels or small packages, the sort step looks like parcel sortation by hand. For pallet freight, it looks more like mixed palletizing. Confirm which one the supplier's cross-dock is.

What the data can support

Time-stamped scans at receipt and loading give weak labels for when a unit entered and left the flow. The pick confirmation as labels page discusses the clock-alignment problem in a related setting. The receiving and inbound checks page covers the inbound side of the same process.

Suppliers and rights

Distributors, carriers, less-than-truckload terminals, parcel hubs, retailers' distribution centers and third-party logistics providers may hold dock video, door schedules and scan records. Where dock cameras exist, they are often installed for security and claims, so they may be mounted wide and high and show goods flows well and hands poorly. Many cross-docks run under tight time windows, so recording that disturbs the flow is unlikely to be approved.

Rights depend on who owns the freight and the shipment data. A carrier handles goods owned by shippers, and the shipment records identify parties, quantities and destinations. A license from the operator may not cover those parties' information. See 3PL providers as data suppliers and client-owned inventory rights. Workers appear on dock cameras, and a use for robot perception should be kept apart from any timing or productivity analysis of individuals. This is not legal advice.

Commissioned collection

A commissioned program might pick a cross-dock lane or a single sort area and record with chosen cameras: one wide view of the staging zone, one at the sorting point, and a worker-mounted view if allowed. Door assignments, load plans and scan logs should be captured with a shared clock. Because the flow is time-critical, a buyer should discuss with the operator whether the recording can be restricted to specific shifts or lanes, and whether a mock inbound with supplied freight is feasible for exception cases.

The access and safety rules at a live dock belong to the operator. This page does not state them, and a data program should follow whatever procedures the site applies.

Illustrative record

# Illustrative only - synthetic values, not an available dataset
event_id: xd_000512
site_id: site_021
inbound: {door: door_07, freight_type: palletized_mixed}
outbound_plan: {destinations: [dest_A, dest_B, dest_C]}
units:
  - {unit_id: unit_0001, scan_in: 14:10:05, scan_out: 14:42:19, staged_at: lane_B}
  - {unit_id: unit_0002, scan_in: 14:10:41, exception: label_unreadable}
video:
  views: [dock_wide_fixed, staging_overhead]
  sync: shared_clock
excluded: robot actions, handling forces, vehicle or equipment control

Any sizing sketch is only illustrative: for instance, two freight types, three destinations and two exception classes make twelve cells, and those inputs are arbitrary, not minimums. The right quantity depends on architecture, task and embodiment.

What this data cannot support

• Robot control. Unloading, staging and loading require robot or teleoperation data. Dock footage, shipment records and recognition labels alone do not provide the actions, balance or contact information a controller needs.
• Planning optimality. Records show how a site ran, not the best sort or door assignment.
• Time-pressure behavior. Human shortcuts under pressure may not be behaviors a robot should imitate.
• Equipment operation. Forklift and powered-equipment control is a different task with separate safety considerations.
• Cross-site generalization. Dock layouts, freight mix and systems differ.

Keep evaluation docks, sites or freight types apart from training data when the claim is about generalization.

Questions to ask suppliers

1. What freight types move through the cross-dock: parcels, cartons, palletized, mixed?
2. Which cameras cover doors and staging areas, and at what height and angle?
3. Are door assignments, appointments and load plans exportable and linkable to video?
4. What scan events occur at receipt, staging and loading, with what clock accuracy?
5. How are exceptions such as mislabels, shortages and damage recorded?
6. Who owns the freight and shipment data, and which parties must consent to use?
7. Can new recording be scheduled for specific lanes without disrupting flow?
8. What notice have workers had, and can they be identified?
9. What redaction is applied to labels, trailers and plates, and does it hide what you need? (Redaction does not settle who may license the footage.)

Discuss your requirements

If you are scoping cross-dock or flow-through handling data, describe the freight types, record linkage and permitted uses to SourceX. SourceX reviews the requirement, looks for US businesses that may hold suitable material, checks the data and the supplier's licensing permissions, and follows up about potential sources. If no existing asset fits, SourceX can scope a commissioned collection of new recordings. A request does not guarantee a match. See the humanoid training data guide and the warehouse data hub for the wider map; the catalog page on dispatch logs is related.

Describe the workplace demonstration data your robotics project needs

Sources

• Design and control of warehouse order picking: A literature review — warehouse design background; it does not address cross-docking.