PlayData pairs glove signals with video for a robot-data prototype

In an [October 29 report](https://wro‑association.org/seoul‑students‑win‑international‑award‑for‑turning‑human‑play‑into‑robot‑training‑data/), World Robot Olympiad describes PlayData, a Seoul student team’s prototype using a data glove and body camera. WRO says it records first‑person video, finger movement, hand motion and contact force, with a system that synchronizes and structures the recordings. Those streams answer different questions about an interaction. A scene viewed by a camera and a movement recorded at the hand are observations to interpret together, rather than interchangeable versions of the same thing. For a prospective user of such data, a useful description would identify the task, the object and what each recorded channel represents. Consider a hypothetical demonstration of turning a knob. A reviewer might want to distinguish reaching for it, making contact, turning it and letting go. That example illustrates a way to inspect a collection plan; it is not a claim that PlayData supplies those labels or that a robot has reproduced the action.

The [organizer’s account](https://wro‑association.org/seoul‑students‑win‑international‑award‑for‑turning‑human‑play‑into‑robot‑training‑data/) places the international event in Hyderabad on September 28, 2026. The project explores gathering interaction data from activities people already perform, including escape‑room play. The practical question behind synchronization is which observations belong to the same part of an interaction. In the knob example, a force reading associated with a later turn would answer a different question from one associated with the initial touch. A reviewer can ask how the recording describes that relationship before treating a video frame and a sensor value as a single example. Other useful questions concern how the collection marks the beginning and end of a task, and how someone inspecting it can recognize an interrupted or incomplete record. These are prospective requests for a data description, not documented PlayData schema fields or evidence that its synchronization has been independently checked.

WRO’s [report](https://wro‑association.org/seoul‑students‑win‑international‑award‑for‑turning‑human‑play‑into‑robot‑training‑data/) describes a prototype and its collection concept; it does not establish a publicly usable dataset or a measured robot‑learning result. For a team considering reuse, the next evidence would be more specific than a successful capture demonstration. It could include an accessible sample with channel definitions and time alignment, a stated permission basis for reuse, and a description of the task a receiving system is intended to learn. Whether such materials are available is separate from whether the collection hardware exists. A later learning result would need its own account of the data used, the receiving robot and what the evaluation actually measured. The current report is useful as a description of an interaction‑capture approach. It is not a reason to assume a transferable dataset, general robot capability or permission to record people’s activities. OpenTools has not inspected the hardware, collected data or run a robot experiment. A tactile glove on display at the MIT Museum, photographed April 17, 2026. This is archival sensor‑glove context, not PlayData’s prototype, captured training data or a robot‑learning test. Photo: [Daderot / Wikimedia Commons](https://commons.wikimedia.org/wiki/File:Scalable_tactile_glove,_2019_-_MIT_Museum_-_Cambridge,_MA_-_DSC09084.jpg), [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/). Cropped for display.