About Motion Lab

Motion Lab is a collection of live browser simulations exploring humans and robots moving through shared physical spaces.

Each scenario exposes a different part of the same question: not simply whether a machine can detect an object, but whether it can understand how that object is moving and use that signal to act more naturally.

Motion Intelligence

Conventional autonomy perceives objects and plans routes around them. Motion Intelligence adds another signal between those layers: what observed movement suggests might happen next.

SpatioTemporal compresses movement across space and time into Motion Tokens, a learned vocabulary of movement. Sequences of these tokens allow a machine to distinguish between patterns such as approaching, slowing, hesitating, turning away or committing to a path.

It does not replace perception or planning. It supplies both with compact motion and emerging-intent information at the moment it matters.

A small model with a big job

Motion Intelligence in these scenarios comes from a model of around 30 MB, small enough to deploy directly onto a robot alongside its existing autonomy stack.

That compact size is deliberate. The model is designed as a fast, efficient layer concerned specifically with movement, intent and what might happen next.

Humans do not consciously calculate trajectories every time someone crosses in front of us. Much of our ability to move safely around other people is fast, instinctive and almost invisible. We notice a hesitation, anticipate a step, adjust our path and keep moving.

The result should be robots that move more naturally around us, anticipate our movements earlier and become easier for humans to understand and trust in shared spaces.

A small model helping robots read the room.

X-Ray Mode

X-Ray Mode exposes the observations, motion information, inferred intent and planner state made visible by each scenario.

It is a view into the missing middle layer between perception and planning: not simply where something is, but how it is moving and what it might do next.