Space Domain Awareness and Prediction via SpacePredict Machine Learning Platform

The population of visible orbital objects (active and passive) of around 29,000 is expected to grow to anywhere from 100,000 to 500,000 within 5-10 years. Given this significant increase in scale and congestion on orbit, and the competition which has been introduced between various nation-states, maintaining the chain of custody from launch, validation of intent, attribution, and resolution of anomalies or actions is of critical importance to both military and commercial operators.  The challenges encompass not just “traffic management” given the huge increase in physical objects in multiple orbits, but “prediction of traffic intent” that will engage military/civilian operators in this crowded and dynamically changing regime.  The ability to not just predict and avoid physical conjunctions with better fidelity will be needed, but the ability to assess an object’s risks (whether intentionally created or an un-intentional anomaly or failure) before it causes problems may now be required.  It is almost impossible for a human centric operations arena that the military or civilian sectors use today (e.g. CSpOC, etc) to respond fast enough to the expected numbers of conjunction alerts and unknown anomalies or maneuvers amongst the un-precedented growth expected in orbit. An artificial intelligence driven system that uses temporally based real-time information that can assess risk/problems and dynamically create courses of options may be the only way to enable human operators to predict then respond to this un-precedented environment. USC’s Information Sciences Institute (ISI) has created the first steps in a holistic space persistent orbital cognizance tool set. “SpaceAware” is a modular knowledge graph architecture with predictive analytics to address the first step of awareness. The current SpaceAware system has the ability to ingest daily updates of the physics based CSpOC catalog of orbital objects, scrape multiple online public sources, repositories, websites, blogs, mailing lists, etc. about every object in space in real-time. Kayhan Space is partnering with USC’s ISI to develop SpacePredict – a platform containing statistical and machine learning algorithms utilizing the realistic “capability attributes” generated by SpaceAware to perform predictive analyses on the orbital objects.