Collin Phillips, Virginia Tech National Security Institute; Gavin Saul, Virginia Tech National Security Institute; Kevin Schroeder, Virginia Tech National Security Institute; Jonathan Kadan, Space Systems Command
Keywords: Space Domain Awareness, network, tasking, scheduling, dynamics model, search, maneuver detection, optimization
Abstract:
As the number of resident space objects (RSOs) grows at an unprecedented rate, the strain placed on a limited number
of ground-based sensing systems has increased dramatically. Due to combinatorial growth in the tasking assignment
problem, maximizing sensor time efficiency has never been more critical. Common techniques focus on optimizing
a tasking matrix that represents a bipartite graph between sensors and observation targets. When viewed as a joint
action space, the bipartite matching solution restricts the available actions and does not generally account for potential
coincidental information gain. Fortunately, tasking ultimately reduces not to a target label, but the boresight vector of
the selected sensor. The reality of incidental or serendipitous observations allows a single sensor to provide multiple
observations for a set of RSOs for a given task vector. We propose a framework and algorithm for optimizing sensor
efficiency by performing joint optimization of the sensor boresight, i.e. pointing direction, for an entire heterogeneous
sensor network. We show that our method outperforms more traditional matrix-based tasking and improves the number
of unique observations over the test period. Additionally, we present and explore a unique method for incorporating
negative information through the novel use of a genetic particle filter.
Date of Conference: September 16-19, 2025
Track: Space Domain Awareness