From one internal state to many interacting ones

Everything covered so far applies naturally to a twin of a single robot: one physical unit, one stream of sensor data, one virtual model tracking one internal state. But the same underlying idea, a live virtual replica synchronized with physical reality, scales up to represent an entire production line, warehouse, or fleet of robots operating together. A facility-scale twin doesn't just track many robots' individual states in parallel, it also has to represent how those robots interact: shared floor space, task handoffs between units, contention for charging stations or shared tooling, and the knock-on effects when one robot's slowdown or failure changes what every other unit in the system should do next.

This interaction layer is what makes facility- or fleet-scale twins genuinely harder than just running many single-robot twins side by side. A single-robot twin's questions are largely self-contained: is this joint wearing unusually, would this speed change stress this specific unit. A fleet twin has to answer questions that only make sense at the system level: if robot A slows down for maintenance, does robot B's queue back up, does a scheduled change to one unit's task sequence create a bottleneck two stations downstream. Those questions require the twin to model relationships between units, not just each unit's own condition.