Barrage Relay Networks Thomas R. Halford * * TrellisWare Technologies, Inc. 16516 Via Esprillo, Suite 300 San Diego, CA 92127–1708 Email: {thalford,kchugg}@trellisware.com Keith M. Chugg †* † Communication Sciences Institute Ming Hsieh Dept. of Electrical Engineering University of Southern California Los Angeles, CA 90089–2565 Email: chugg@usc.edu Abstract—A receiver-oriented perspective on capacity scaling in mobile ad hoc networks (MANETs) suggests that broadcast and multicast may be more natural traffic models for these systems than the random unicast pairs typically considered. Furthermore, traffic loads for the most promising near-term application for MANET technology – namely, networking at the tactical edge – are largely broadcast. The development of novel MANET approaches targeting broadcast first and foremost, however, has not been reported. Instead, existing system designs largely rely on fundamentally link-based, layered architectures, which are best suited to unicast traffic. In response to the demands of tactical edge communications, TrellisWare Technologies, Inc. developed a MANET system based on Barrage Relay Networks (BRNs). BRNs utilize an autonomous cooperative communication scheme that eliminates the need for link-level collision avoidance. The fundamental physical layer resource in BRNs is not a link, but a portion in space and time of a cooperative, multihop transport fabric. While initial hardware prototypes of BRNs were being refined into products by TrellisWare, a number of concepts similar to those that un- derlie BRNs were reported independently in the literature. That TrellisWare’s tactical edge MANET system design and academic research reconsidering the standard networking approach for MANETs arrived at similar design concepts lends credence to the value of these emerging wireless network approaches. I. I NTRODUCTION The past decade has seen a significant research focus on both the theoretical capabilities of and protocol designs for mobile ad hoc networks (MANETs). Efficient unicast data transport has been the primary focus of this effort. The most common formulation for capacity scaling analysis, for example, assumes sufficient node density so as to ensure link-level connectivity and considers the largest rate that can be guaranteed between randomly selected source-destination pairs. Gupta and Kumar’s seminal result [1] shows that this rate falls quickly as the number of network nodes increases. One consequence of this result is that protocols seeking to maximize efficiency for randomly paired unicast traffic must be carefully designed owing to the effective reduction in available bandwidth with network growth. Indeed, MANET protocol designers have sought to do just this by carefully designing link-level medium access control (MAC) and multi- hop routing algorithms that seek to minimize the number of link transmissions while avoiding collisions. Another potential consequence of Gupta and Kumar’s result is that networks without infrastructure are inherently ill-suited to the randomly- paired unicast traffic model. Although a simplification, this interpretation should not come as a surprise since interference associated with concurrent wireless transmissions must be avoided and routing of data packets utilizes intermediate nodes that are themselves sources and destinations. Much of the impetus for the research activity described above has been provided by the potential applications for MANETs, one of the most compelling of which is communi- cations at the tactical edge. A squadron of soldiers seeking to maintain connectivity in a challenging RF propagation environment (e.g., urban canyons, ships, subterranean struc- tures, etc.) is an example of edge networking; first-responder communications, such as remote search and rescue, is another. This tactical MANET application assumes no supporting in- frastructure and often no means for communicating outside of the squadron. Since nodes are both mobile and typically in rich scattering environments, link-level connectivity is unreliable and the network topology is highly dynamic. The traffic patterns and key performance metrics for tactical MANETs also differ substantially from those typically considered for ad hoc or sensor networks. Specifically, low-latency, network- wide broadcast and robust connectivity are the primary re- quirements of tactical MANET systems, while typical traffic includes interactive push-to-talk (PTT) voice and real-time video streaming from a small set of source nodes. In light of the design goals for tactical MANET, it is useful to consider the throughput and latency scaling properties of broadcast traffic. It is described in Section II of this paper how reinterpreting existing scaling results suggests that ad hoc networks may be better suited for broadcast and multi- cast than for the more commonly considered unificast traffic models. In particular, the aggregate useful data rate received in the network scales linearly with the number of nodes for broadcast, which is not unexpected since all relaying nodes are also intended recipients. Despite this apparent match to broad- cast traffic and the important tactical MANET application, there has been relatively little effort committed to designing MANET systems that target broadcast first and foremost. Section III of this paper describes the basic concepts of Barrage Relay Networks (BRNs), which are a type of MANET designed from the ground up for the demands of tactical edge communications. BRNs utilize autonomous cooperative communications to enable packets to ripple out from source nodes rapidly and reliably through the network. Each node, upon receiving a packet, repeats the data as part of one or