playrolllabsGet in touch ↗

Give physical AI
a human starting point.

We’re building a way for robotics teams to collect human demonstrations through playable simulation tasks.

Inside the approach ↘
CONCEPT ILLUSTRATION
WHY THIS MATTERS

More ways to collect
human experience.

Teaching robot behavior needs examples. Collecting them on physical hardware requires equipment and operator time. Playroll’s proposed approach moves part of that collection into simulation, making it possible to invite more people to contribute.

OUR APPROACH

Play is the interface.
Learning is the goal.

A robotics team defines a behavior it wants to teach. We aim to turn that task into a browser simulation people can play, capturing their actions as demonstrations for training.

Human play
↓
Demonstrations
↓
Robot learning
WHAT WE’RE BUILDING TOWARD

Task-specific demonstrations

Human attempts at the behaviors a team wants to teach, rather than unrelated gameplay.

Different human approaches

A collection model designed to invite multiple ways of solving the same task.

A repeatable collection loop

Define a task, collect examples, then target the next gap in a robot’s training.

Early development. Data quality, cost, and transfer to physical robots still need validation.

What should robots
learn next?