WBCD 2026 Competition Tracks

WBCD 2026 Competition Track

Authoritative source. This page is a readable summary. The wbcd_tasks GitHub repository holds the authoritative, most up-to-date task specifications, scoring rules, bills of materials, and CAD/STL files. If anything here disagrees with the repo, the repo wins.

Per-track READMEs: Track 1 · Track 2 · Track 3 · Track 4

Autonomy Levels & Run-Based Scoring

Every track in this competition supports all three autonomy levels — teams choose whichever best fits their system. Teleoperation may be performed with any controller of the team's choice (e.g., VR headset, inertial motion capture suit); fully autonomous runs use a learned or programmed policy end-to-end with no human control during execution.

Teams declare one autonomy level per run (before the run starts). The sum of base points earned in that run is multiplied by the corresponding multiplier. The same level applies to every item/step inside the run — teams do not switch mid-run. Different runs may use different autonomy levels.

Level	Description	Multiplier
On-site Teleoperation	Human operator is physically present and controls the robot with direct line-of-sight	×1
Remote Teleoperation	Human operator controls the robot remotely via video feed only (no direct view of the scene; network latency applies)	×2
Fully Autonomous	No human control during execution; the robot runs a learned or programmed policy end-to-end	×4

What counts as a “run” is defined per track:

Track 1 — Logistics Picking: one 10-minute block (each team gets 2 runs, 20 min total)
Track 2 — Logistics Packing: one item through pick → scan → place
Track 3 — WBCD XtalPi Lab Experiments Track: one round (Steps 1 → 2 → 3)
Track 4 — Deformable Manipulation: one complete T-shirt cycle (Steps 1 → 2 → 3)

Note: The multipliers above are competition-wide defaults. The optimal multiplier for a given task ultimately depends on its difficulty and content — these presets serve as a balanced starting point and may be adjusted per track in future editions.

Track 1: Logistics Picking

Robot from Unitree

Autonomy & scoring: Supports on-site teleop (×1), remote teleop (×2), and fully autonomous (×4) — teams declare one level per run. Here, one run = a 10-minute block (each team gets 2 runs, 20 min total). See Autonomy Levels & Run-Based Scoring above for details.

Canonical spec: track-1-logistics-picking/README.md on GitHub is the authoritative, most up-to-date reference for this track.

Robot Demonstration

Teleop real robot video from Unitree -- bring your best and do even better!

This task focuses on whole-body control and end-effector coordination utilizing the Unitree G1 humanoid robot. The objective is to bridge the gap between research and practical application by simulating a logistics scenario in which items of varying difficulty must be transferred from shelves of different heights to a transport vehicle. The system relies solely on the robot's onboard perception to execute complex maneuvers including upright, bent, and crouched postures.

Robot Configuration: Unitree G1

End effectors: A pair of end effectors (clamps, three-finger or five-finger dexterous hands)

Allowed modes: On-site teleoperation, remote teleoperation, or fully autonomous. For teleop runs, any controller of the team's choice is allowed (e.g., VR headset, inertial motion capture suit). For fully autonomous runs, no human control is permitted during execution.

Task Rules

The competition time budget is 20 minutes, split into two 10-minute runs. The goal is to complete as many item transfers as possible across the two runs.

Runs: Each team gets exactly 2 runs of 10 minutes each. Teams declare one autonomy level per run (on-site teleop, remote teleop, or fully autonomous) before the run starts. The same level applies to every item in that run; the two runs may use different levels.
Control: Teleop runs may use any controller of the team's choice (e.g., VR headset, inertial motion capture suit). Fully autonomous runs permit no human control during execution.
Perception: Participants must acquire external information solely through the robot's onboard perception system.
Capacity: Within a run, there is no limit on the number of items the team attempts to transfer, provided they are not dropped.

Step 1: Shelf Picking

Navigate the robot to the shelving unit and extract items. The difficulty varies by shelf height, requiring specific body postures.

Human demo

Robot demo

Action	Description	Posture
1a	Pick item from Top Shelf	Upright Position
1b	Pick item from Middle Shelf	Bent Position
1c	Pick item from Bottom Shelf	Crouched Position

Success Criteria: Item securely grasped from the shelf without knocking over other items.

Step 2: Transportation

Transport the grasped items from the shelving area to the designated unloading area.

Human demo

Robot demo

Action	Description
2a	Stabilize item(s) during locomotion
2b	Navigate to the transport vehicle/table

Success Criteria: Maintain grasp on items throughout the movement. Drops result in penalties.

Step 3: Placement

Place the items onto the transport vehicle or unloading table.

Human demo

Robot demo

Action	Description
3a	Position item over target area
3b	Release item securely

Success Criteria: Item rests stably on the unloading surface.

Scoring

Scoring is weighted based on the difficulty of the whole-body motion required (posture).

Source	Posture	Points	Criteria
Top Shelf	Upright	+5	Successful transfer from high shelf
Middle Shelf	Bent	+8	Successful transfer from middle shelf
Bottom Shelf	Crouched	+10	Successful transfer from bottom shelf
Item Drop	—	-3	Per item dropped during transportation

Autonomy multiplier. Base points from all successful transfers in a 10-minute run are summed, then multiplied by that run's autonomy multiplier (×1 / ×2 / ×4). The same level applies to every item in the run; the two runs may use different levels. See Autonomy Levels & Run-Based Scoring at the top of the page.

Example

A team's two 10-minute runs, each with its own declared autonomy level:

Run	Autonomy	Transfers	Base total	Multiplier	Run score
Run 1	Remote teleop	4 top (+20) + 2 mid (+16) + 1 bot (+10) − 2 drops (−6)	40	× 2	80
Run 2	Fully autonomous	2 top (+10) + 1 mid (+8) − 1 drop (−3)	15	× 4	60
Total					140

Items Used in This Task

Full Bill of Materials with quantities, prices, and supplier links: track-1-logistics-picking README → Bill of Materials.

Human Demonstration

Sample task execution by a human operator for demonstration purposes.

Open authoritative README on GitHub (full rules, scoring, BoM, CAD)

Track 2: Logistics Packing

Robot from Galaxea

Autonomy & scoring: Supports on-site teleop (×1), remote teleop (×2), and fully autonomous (×4) — teams declare one level per run. Here, one run = one item through pick → scan → place. See Autonomy Levels & Run-Based Scoring above for details.

Canonical spec: track-2-logistics-packing/README.md on GitHub is the authoritative, most up-to-date reference for this track.

The objective of this challenge is to demonstrate robotic capabilities in logistics fulfillment. The system must pick, scan, and pack items from a primary storage container into three designated shipping parcels. The workflow involves handling diverse items (rigid boxes, irregular shapes, and soft apparel), identifying barcodes via wrist-mounted cameras, and securing the parcels by closing their lids.