Mapping ad hoc communications network of a large number fixed-wing UAV swarm

Abstract

In 2015, a group of Naval Postgraduate School (NPS) professors and students set a record when they flew 50 fixed-wing unmanned aerial vehicles (UAVs) simultaneously as a self-organizing swarm. These vehicles were able to execute behaviors based on message notification from a single ground station, and then decide within their swarm group how to order themselves. They were able to accomplish this by communicating over their 802.11n wifi connections. Understanding the strengths and weaknesses of this network will be essential to scaling the swarm to larger sizes or even creating partitioned sub-swarms. The work covered in this thesis is to build a model of the NPS swarm's communication network in ns-3 simulation software and use popular network metrics to illustrate the performance of the network as swarm size increases. It also applied four routing protocols to the swarm and compares their performance to the broadcast protocol. The swarm's communication network was not very tolerant of overhead. This thesis concludes that any routing protocol applied to the (NPS) swarm in the future should consider protocols that reduce or strictly manage overhead generated by either routing tables or multiple message copies. Goodput and packet delivery ratio were the quantitative metrics used. While they illustrate reliability, they do not give a good picture of latency. It would be useful to add latency as a quantitative metric to future work because some swarm messages are more time-sensitive than others. It may be that more than one routing protocol or a protocol with variable settings would be best for this swarm and its various message priorities.