[wenc]

A "digital twin" is a online mathematical model/function/simulation that is continuously fed with sensor input data.

It's a new term for an old idea that has been around for decades in manufacturing; but what's old is new again in the context of IoT devices.

And it's only the first step -- manufacturing has gone a few steps farther.

Digital twins are primarily concerned with constructing a digital representation of a physical process/asset.

Model-based predictive control [1] (developed in 1980) has been used commercially to not only automatically control said assets, but do so optimally (by iteratively solving an optimization problem based on a mathematical model at every time period, and then reading new measurements off the feedback loop to update itself).

This type of optimal control technology is the reason why oil refineries are run so efficiently today.

[1] https://en.wikipedia.org/wiki/Model_predictive_control