Dtn routing protocols for drone swarm telemetry

Drone swarms pose a particular challenge to telemetry networks, due to the number of airborne nodes involved, and their potential to overwhelm the available bandwidth on the communications channel with simultaneous telemetry streams. Previously, we saw that mobile ad-hoc (MANET) routing protocols could exacerbate this issue by flooding the network with routing-control packets. In this work we model the Naval Postgraduate School fixed-wing drone swarm and compare the performance of several disruption-tolerant networking (DTN) routing protocols designed to address these challenges. INTRODUCTION This work models the network formed by a swarm of fixed-wing aerial vehicles developed at the Naval Postgraduate School (NPS) and compares the performance of several Disruption Tolerant Network (DTN) routing protocols, observing tradeoffs between the potential routing protocols for the current swarm configuration to select candidates for future larger-scale swarms. In previous work, we performed this comparison using Mobile AdHoc Network (MANET) routing protocols, and found that the density of nodes in the radio communication environment posed a significant challenge to those protocols [1]. In 2015, a group of NPS professors and students set the record for largest fixed-wing unmanned aerial vehicle (UAV) swarm flown at one time [2]. The swarm had 50 vehicles flying simultaneously and successfully demonstrated distributed decision-making with all processing occurring on swarm vehicles rather than a centralized control station. This swarm uses broadcast messaging, and no multi-hop routing is implemented. While this approach provides low-latency communications when all nodes are within range of each other, it introduces two problems. The first is that as the number of nodes increases, it will eventually saturate the communication channel, since no spacial-reuse is possible with this scheme. Early signs are that this will occur with swarm sizes not much larger than the current 50 nodes. Secondly, future mission tasking is likely to require distributing nodes geographically such that not all are in communication range of each other, as well as sub-swarm missions where the swarm splits temporarily and later regroups. BACKGROUND AND RELATED WORK In this section we examine the existing state-of-the-art in drone-swarm research. This paper is authored by employees of the United States Government and is in the public domain. Non-exclusive copying or redistribution is allowed, provided that the article citation is given and the authors and agency are clearly identified as its source. Approved for public release: distribution unlimited. A. Swarm Networking Issues Drones and drone swarms may move rapidly with respect to their radio range, causing them to frequently lose and regain connectivity to ground stations and other vehicles. In the case of the NPS drone swarm, the cruising speed is 18 m/s. Currently, the NPS swarm uses no multi-hop routing protocol and instead uses 802.11n infrastructure-mode wireless. Messages to drones are addressed to the IP broadcast address application-layer message header determines whether a specific node should receiver or drop each message. Depending on future mission planning, keeping all nodes within range of the ground station may not be desirable, leading to the need to multi-hop messages from more distant nodes back to the ground station. In addition to network and routing layer limitations, prior work has shown TCP to perform poorly in lossy wireless environments [3], and this effect is exacerbated with each additional wireless link in the path. The TCP routing protocol was designed for long-lasting connections along an established path. Attributes of TCP, like the three-way handshake, slow start algorithm, and congestion control algorithm were not designed for scenarios where packets are lost due to causes other than congestion, or where paths are frequently broken. UDP is not encumbered by the same restraints as TCP, but offers no acknowledgment that packets were delivered to the destination [4]. Channel bandwidth is also an ever-present constrain in telemetry systems, and can easily be overwhelmed with 10s or 100s of nodes attempting to communicate simultaneously. On one hand, multi-hop routing may help with this, by permitting more spacial reuse in the channel allocation plan, however overhead from routing protocol control messages can also consume significant amounts of bandwidth. B. DTN Routing Protocols The following routing protocols were employed for comparative analysis: • DSDV: a proactive distance-vector based MANET routing protocol [5]. • AODV: a reactive distance-vector based MANET routing protocol [6]. • OLSR: a proactive link-state MANET routing protocol [7]. • Epidemic: Simple flooding-based DTN routing protocol [8]. • Vector: Epidemic-based routing protocol that limits flooding based on direction of travel [9]. • Gapr: Geolocation-assisted predictive DTN routing protocol [10]. • Gapr2: Improvement on GAPR that reduces message replication [10]. For the sake of space, we do not go into the details of each protocol, but refer you to their respective source documents. METHODOLOGY This section provides a review of the initial simulation built to model the NPS swarm in ns-3, as well as modifications made by this study to improve the mobility model and test additional routing protocols. Ns-3 is a discrete-event network simulator intended to implement the entire