How we sync twelve robots to a single BPM

Millisecond-level sync across a fleet isn't magic — it's a careful stack of network timing, motion file pre-loading, and a trigger protocol we've spent two years refining. Here's what's actually under the hood.

The hardest part isn't the robots. It's the network. A single packet drop at the wrong moment means one unit fires late, and late at scale reads as broken to an audience. So we don't rely on the network for the trigger itself — we use it only for pre-sync, then cut to a local timer on each unit.

Each robot runs a pre-loaded motion file. The trigger is a broadcast UDP packet carrying a shared epoch timestamp. Every unit calculates its own offset and fires from local memory. By the time the packet lands, the motion is already in RAM.

The result: ±1ms across twelve units. In practice, that's indistinguishable from zero to any human watching. But we keep measuring it — because the day we stop is the day it drifts.

We're currently moving to a hardware sync line for outdoor shows, where RF interference can jitter even a reliable local network. The principle is the same: pre-load everything, trigger once, trust the local clock.