Decentralized Belief Propagation for Satellite Constellation Resource Management

Increased constellation sizes have driven the need for the development of decentralized in-orbit autonomy. In the context of multi-target tracking (MTT) applications, distributed data fusion (DDF) techniques can enable constellations to autonomously perform estimation tasks. These techniques can strain the computational and bandwidth resources of the constellation, leading to the development of additional in-orbit autonomy to manage the constellation resources. In this work, we present a method of efficient, decentralized constellation resource management where satellites infer the available capacity of their neighbors. This inference is used to inform resource management planning where satellites exchange computational tasks and optimize bandwidth utilization. We show that this planning significantly reduces the resource burden on a MTT constellation, advancing the feasibility of decentralized in-orbit autonomy.