Bayesian detection and classification for space-augmented space situational awareness under intermittent communications

This paper examines the problem of detecting and classifying objects in Earth orbit using a space-augmented space surveillance network (SA-SSN). A SA-SSN uses a combination of ground- and space-based sensors to monitor activities over a range of space orbits from low earth orbits up to an altitude higher than the geosynchronous orbit. We develop a cost-aware Bayesian risk analysis approach for object detection and classification, using range-angle sensors with intermittent information-sharing between the sensors. The problem is formulated in a simplified two-dimensional setting where the SA-SSN is composed of four ground-based sensor and a space-based orbiting sensor satellite. This is done in order to reduce computational complexity while retaining the basic nontrivial elements of the problem. We will demonstrate that objects in geosynchronous orbits can be detected and perfectly classified (under appropriate sensor models) if they intermittently cross the field of view of some sensor in the SA-SSN, and that performance degrades for objects located in non-geosynchronous orbits. We will conclude the paper with future research directions on how to address the detection and classification of objects in non-geosynchronous orbits.