MACHINE LEARNING IN MOBILE CUBESAT COMMAND AND CONTROL (MC3) GROUND STATIONS

Abstract

The Mobile CubeSat Command and Control (MC3) ground station network is a program designed to enable many organizations to command and control very small satellites (or CubeSats) in low-earth orbit. The MC3 network currently consists of ground stations that are geographically dispersed and utilize non-standard configurations of commercial off-the-shelf equipment. The non-standard configuration of each location poses a challenge for the small staff of MC3 network operators who monitor network and ground station health status. These operators rely on software and automation to ensure the MC3 network is healthy and can support any organization’s mission. However, the problem is that a normal state in one location can look different from the normal state at another location in terms of equipment and, therefore, health status. Determining the normal state using machine learning will facilitate further analysis of ground station health and the implementation of near-real-time health status monitoring to augment the MC3 network operators’ capabilities. The research focused on using the K-means++ unsupervised machine learning clustering algorithm to model the normal state. This research could not conclusively determine the normal state of the NPS MC3 ground station, but it does establish a launch point for further work