Towards efficient drone networking
Cellular networks and delay-tolerant networking will be key enablers for drone swarms to take off.
My first experience with drone research was quite early: I participated in an ONR-funded project on unmanned aerial vehicles (UAVs) as part of my Ph.D. studies at Stanford University from 1999 to 2003. The overall research environment was much different back then: drones were the size of small cars (and quite noisy), and experiments were very modest in scope; only military and search-and-rescue applications were envisioned by researchers at the time.
Things have definitely moved forward.
Improvements in communications technology — to a large extent driven by the smartphone revolution — and multi-rotor technology have led to an incredible range of drones that are very affordable compared to even the recent past. As a result, the variety and sheer number of applications is staggering, ranging from the outrageous (use of drones as clay pigeons), to the silly (drones walking dogs), to the mundane (use of drones for traffic monitoring), to the noble (delivering drugs to hard-to-reach third-world villages). In fact, it is this range of applications, and not the numerous impressive scientific advancements in the field, that I am most impressed with. In this regard, drones are similar to smartphones: both are versatile technologies whose applications are limited only by the imagination of a very large base of enthusiasts as well as the innovators and companies that cater to them. Therefore, I would safely bet that what we have seen of drones is only the tip of the iceberg.
Despite improvements in the technology, significant challenges remain. An important one is how to enable extensive, dependable communication between drones, i.e., how to make a communication network out of a drone swarm. Unfortunately, the wireless channel is notoriously challenging, and the situation is exacerbated by the fact that the dominant wireless technology available to experimenters is IEEE 802.11, which was designed with mostly stationary users sipping lattes in mind, and not aerial platforms tumbling through the sky. This is one of the two main reasons that the vast majority of drone applications that have been deployed so far do not involve drone-to-drone communication — the other main reason being that one can do an awful lot with a single drone, or at least a few drones that operate independently and in parallel.
However, the ability to form a network is crucial to numerous challenging applications, e.g., cooperative sensing and large-scale autonomous search-and-rescue. It is therefore not a surprise that solving the numerous technical challenges inhibiting wireless drone networks is very high on the research agenda.
How then can the drone networking problem be solved?
It is tempting, especially for a person of my background on multihop wireless networks (or packet radio networks, or ad hoc networks, depending on one’s preference for names) to proclaim that in 10 or 20 years’ time we will have managed to have efficient, large multihop networks in the sky, operating in complete autonomy and serving the communication needs of the swarm by offering low latencies and large data rates between arbitrary pairs of nodes. However, I believe that there will be two more, much more mundane, key enablers of drone networking, in addition to wireless multihop routing. I feel that both of them have not received the attention they deserve by drone researchers.
The first enabler will be cellular networks. IEEE 802.11 did put a large dent in the armor of 3G, and it is holding its own against 4G. It is also practically the only communication technology used in the consumer drone market for transporting data. However, due to huge investments in the cellular infrastructure, in part due to the smartphone evolution, the costs of using cellular networks are declining. More crucially, 5G focuses on low latency and machine-to-machine communication. Therefore, I bet that, in 10–20 years’ time, wherever a drone network is operating in inhabited areas, there will be a substantial 5G (or 6G, or whatever cellular technology comes after that) network infrastructure available, and the drones will be using it to satisfy much of their communication needs.
The second enabler will be delay-tolerant routing techniques. In other words, nodes will be trading off delay for bandwidth, so that traffic will be arriving at its destination with very large delays, comparable in fact to the times needed for the topology to change substantially. Humans do it all the time: we wait to come close to someone before initiating a discussion, unless we need to convey time-critical information (in the latter case, we might also go out of our way to convey the information faster). This will also be the modus operandi of drones in many applications whose traffic (or at least part of it) will be delay-tolerant, such as environmental monitoring, the creation of 3D models of construction sites, etc.
So, what lies ahead?
In 10 to 20 years drones will be routinely roaming our skies. They will be operating separately, or, quite often, in smaller or larger swarms. They will be operating autonomously, with a tight level of coordination, completing tasks that are already well known and anticipated, as well as tasks still unimagined. They will be using, for their communication needs, both cellular (whenever the infrastructure is available and the cost of communication is justified) and multihop (in all other cases) networking, trading off delay to improve other performance metrics whenever possible. These are my predictions; however, I could be wrong. As the saying goes, it is very difficult to make predictions, especially about the future.
Stavros Toumpis is Assistant Professor at the Athens University of Economics and Business. He obtained his PhD from Stanford University in 2003 for a thesis in wireless ad hoc networks.