Bryan Little, Purdue University; Carolin Frueh, Purdue University
Keywords: Sensor-tasking, machine learning, SSA
Abstract:
The object population in the space around the earth is subject to constant increase. With the increased sensor capabilities, more and more objects will be detected. Although this allows for significant advances in the understanding of all objects in the near-earth region and the expansion of object catalogs, it also puts significant stress on the sensor systems and makes sensor tasking a prime challenge. In order to solve sensor tasking as an optimization problem, various methods exist. Classical methods rely on the problem being formulated in a convex representation. Computationally intensive methods like machine learning have gained a lot of attention and are suitable for problems even when no convex formulation can be found. In this paper, the performance of a greedy algorithm is compared with the performance of a swarm intelligence method, the so-called ant colony. Ant colony optimization is a pathfinding optimization methodology based on probabilistic principles and can be seen as a version of reinforcement learning. As an application case the observation of all known objects in the geosynchronous region with one ground-based and one space-based sensor is used. The performance is evaluated in terms of number of objects that have been successfully tracked and the computational run time in a given observation cycle.
Date of Conference: September 11-14, 2018
Track: Space Situational Awareness