Topology Control in Aerial Multi-Beam Directional Networks

Comparing the performance of a centralized algorithm that retains robust connectivity and reduced throughput to a distributed algorithm that offers higher throughput but fewer network nodes.

In multi-beam directional networks, nodes are able to simultaneously transmit to all neighbors or receive from all neighbors. This spatial reuse allows for high throughputs, but in dense networks can cause significant interference. Topology control (i.e., selecting a subset of neighbors to communicate with) is vital to reduce the interference. Good topology control balances the number of links utilized to achieve fewer collisions while maintaining robust network connectivity.

Directional communication systems offer many benefits over omnidirectional systems, such as increased spatial reuse, longer ranges, and in military networks, lower probability of detection and more resistance to jamming. New approaches to directional communication are increasingly becoming a reality due to recent advances in fully digital phased arrays. In a fully digital array, each antenna element has an analog-to-digital converter behind it, allowing for precise control of the input and output beam patterns. This allows for simultaneously receiving (or transmitting) independent data streams in different directions. This multi-beam capability (i.e., simultaneous transmissions or receptions) allows for a dramatic increase in network capacity.

In addition to this new capability, emerging airborne networks are being developed as mobile ad-hoc networks (MANETs) in which mobile nodes self-organize without infrastructure. As these nodes move through space, the topology of the network changes as new nodes become reachable, and previous connections are broken. This changing topology creates a challenge for directional networks. Although spatial reuse can be high, there is still interference in the system. For instance, though beams may be relatively narrow, multiple nodes can be located within the same main beam. In this case, simultaneous transmission to all is impossible; only one of these neighbors can be transmitted to at a time. As an example, at a range of 50 nmi, a 10-degree main beam covers over 60 nmi2. This is the key challenge: how to control the topology of a directional network by selecting neighbors in order to reduce the interference while still keeping as many links as possible.

This problem has been well studied in omnidirectional networks, but many open directions of research exist for the directional case. Topology control with a degree constraint and the difficulty in finding optimal solutions has also been examined. Others have focused on the effect of topology control on the number of hops between nodes, called “hop stretch,” and developed both near-optimal and more implementable solutions.

A similar approach has been taken, but with a different goal in mind. For example, beam directions can be chosen to cover as many nodes as possible, assuming that there is a single flow to transmit to many neighbors. Conversely, the problem of independent flows to each neighbor is studied, with an eye toward having as few nodes in the same beam as possible.

While high throughput is an important aspect in communications, robustness and a high node degree are also vital. A network with high throughput due to many links with low link utilization leads to long delays before packets are correctly transmitted and high interference. However, the high number of links results in fewer hops in a multi-hop flow, which may be desirable from a robustness standpoint. The tradeoff between throughput and connectivity is a function of the network traffic demand, which in general cannot be predicted, and the density of the nodes in the network. A highly dense network with a very low traffic load may not be well served by an aggressive topology control algorithm that removes many links.

This work was done by Brian Proulx, Nathaniel M. Jones, Jennifer Madiedo, and Greg Kuperman for MIT Lincoln Laboratory. MIT-0005