Inferring digital twins from captured data

A computing device in a communications network, the computing device comprising a digital twin that provides a model of behavior of a physical entity, the digital twin configured to:

• receive at least one stream of event data observed from an environment;
• compute, by at least one processor, at least one schema from the at least one stream of event data, the schema being a compressed representation of the at least one stream of event data, wherein:
• computing the schema comprises repeatedly inferring a structural type of the event data, as instances of the event data are received in the at least one stream of event data, by computing a literal type of an instance of the event data and computing a least upper bound between the literal type and a current value of the inferred structural type, and setting the inferred structural type to be a result of the least upper bound computation;
• and
• the inferred structural type of the event data is learned over time based on the at least one stream of event data observed from the environment;
• participate in a distributed inference process by sending information about the schema to another digital twin in the communications network and receiving information about another schema representing at least one other stream of event data received by the other digital twin;
• compute comparisons of the sent information and the received information;
• based on the comparisons, determine that the schema and the other schema are correlated;
• based on determining that the schema and the other schema are correlated, aggregate the digital twin and the other digital twin by redirecting the at least one other stream of event data to the digital twin instead of the other digital twin;
• and
• based on the aggregating, receive by the digital twin the at least one stream of event data and the at least one other stream of event data via the communications network.

A non-transitory computer storage media storing instructions that, when executed by at least one processor, cause a digital twin that provides a model of behavior of a physical entity to:

• receive at least one stream of event data observed from an environment;
• compute at least one schema from the at least one stream of event data, the schema being a compressed representation of the at least one stream of event data, wherein:
• computing the schema comprises repeatedly inferring a structural type of the event data, as instances of the event data are received in the at least one stream of event data, by computing a literal type of an instance of the event data and computing a least upper bound between the literal type and a current value of the inferred structural type, and setting the inferred structural type to be a result of the least upper bound computation;
• and
• the inferred structural type of the event data is learned over time based on the at least one stream of event data observed from the environment;
• participate in a distributed inference process by sending information about the schema to another digital twin in the communications network and receiving information about another schema representing at least one other stream of event data received by the other digital twin;
• compute comparisons of the sent information and the received information;
• based on the comparisons, determine that the schema and the other schema are correlated;
• based on determining that the schema and the other schema are correlated, aggregate the digital twin and the other digital twin by redirecting the at least one other stream of event data to the digital twin instead of the other digital twin;
• and
• based on the aggregating, receive by the digital twin the at least one stream of event data and the at least one other stream of event data via the communications network.