What follows are general concepts taken from the Big Data Science an Analytics community that we use as a foundation for Space Domain Awareness.
In general, one is interested in information acquisition, organization/management, analysis/exploitation, and decision-making. A key tenet of Big Data is to analyze all sources of information simultaneously so as to get the maximum mutual information on desired space domain awareness criteria and enable going from data to discovery. A “best practice” in organizing information is to do so in an ontology-based knowledge graph, leveraging a so-called Resource Description Framework or RDF. The basic structure of a RDF is a semantic connection in triples: Subject-Predicate-Object.
The power in organizing information in this semantically connected fashion is that it enables linking disparate sources of information, creating vast trees of descriptions of many different elements and encouraging “discovery” from data linking.
It should be straightforward to envision creating RDF triples for elements within the Space Domain, starting with known space objects, their characteristics (physical, functional, kinematic, etc.), who they belong to, who built them, etc. and by linking disparate sources of information in this RDF structure, being able to discover otherwise unknown relationships between space actors, the objects and their environments, etc. Imagine having a user bring an application that queries the RDF using Natural Language searching for any RSO behavior that is correlated with any measured space environment event. For instance, perhaps there are several hundred RSOs in LEO that have a change in their nominal states roughly 14 hours after a solar flare. One would then want to know which objects and what are common elements amongst them. Perhaps they all have a specific type of Kapton they’ve been built with. One might then attempt to find out what physical processes would cause this effect upon that specific Kapton. This is but one example of a significant discovery that could be made simply by organizing, managing, and exposing information in a certain way.
The RDF Triples are used in a Graph and are a middle layer in a holistic SDA framework. The information one would use get acquired via a variety of methods and sources into so-called “instances.” These can be structured (as what comes from sensors) or unstructured (as what could get reported by humans online, via tweets, etc.). The desire is to have automated processes that are always acquiring information and storing them as instances. The information acquired would have a so-called “landing zone” with metadata generated to provide a timestamp on this process. The information is stored as is and absolutely no changes or alterations are made to any instance data. The data are in their “raw” or acquired form.
There must be a separate process that takes the information from the instance data and maps these into the RDF Triples according to an agreed upon standard and semantically consistent vocabulary. The RDF knowledge graph should also have a dictionary that can be readily accessed by those who would interact with the RDFs. Given the international nature of Space Traffic, cultural competency is desired in defining the RDF vocabulary where the user is not asked to or forced to know this vocabulary but the RDF developers have established which country and sector specific terms map to which RDF entities, somewhat like a translation service.
The RDF Knowledge Graph (or Triple Store) is where the current knowledge of the Space Domain resides, with descriptions concerning all of these elements of information. The way in which the information within the RDF Triples can get updated or reconciled given evidence is via the top layer of applications that run or perform so called SPRQL queries on the RDFs. One can envision one application being Space Object Orbit Determination and Prediction, for instance. One application should be validating the RDF data for semantic, physical, and statistical consistency. Related content and relations can be discovered by navigating connected entities, reasoning across these.
This semantic web technology, RDFs, is still in its infancy and it remains to be seen if it can be made to be relevant to a Space Traffic monitoring need. Some salient research issues involve how to make the RDF able to handle dynamic information which in turn may be uncertain or random. There must be some competence in mapping Instance Data to the RDF framework. A vocabulary and dictionary must be developed. There is much yet to be done to make this into a near real time, robust, and resilient architecture that could underpin space safety and sustainability, but we have begun!
See ASTRIAGraph here