This month we bring you a view from the front line of communications theory education on the future of our profession. Petar gives us an insightful, personal view of the future as we head towards connecting everything. Is the job nearly done? Will our profession become a shell of its former self? And how do we rejuvenate our field of study? I’m sure everyone will have an opinion. Feel free to share.
Alan Gatherer EIC
The Role of Communication Engineering Research After Everything is Connected1
Petar Popovski, Ph.D., IEEE Fellow
Professor, Aalborg University, Denmark
There is an old joke, circulating around the Balkans, about a lawyer and a tree. The main case of the lawyer was the dispute between two neighbors about a tree placed at the common border of their properties. The case was going on for decades, the lawyer was mediating, but no solution came and eventually, the lawyer retired and his son took his office. A couple of days later the son came to his father and said “I have solved the problem – I have cut the tree down.” Then the father said “That tree was feeding our family for decades, now you need to find something else to do.”
This technology-agnostic joke has a lot to do with researchers working with technology. Being very successful in creating a certain technology (call it X) also means that the perception of the general public, as well as the research funding agencies, leans towards “there is not much to do about that X in terms of research, it works already.” The most extreme example of this is the widely-feared singularity moment of artificial intelligence, after which the humans get a secondary role and the artificial intelligence decides what to do with them (including whether it will keep funding agencies to support research by humans).
Taking several steps back from the gloomy singularity predictions, we can look at the state of research in communication engineering. Communication technology, and specifically wireless communication, is arguably one of the fastest growing technologies in history. In just two decades, wireless connection achieved the status of a commodity. In 2016 the World Bank published its report , where it is shown that in the developing countries more people have a mobile phone rather than reliable access to other commodities, such as water or electricity. And the R&D party of communication technologies is still going on. Research and standardization of 5G is in full steam, supported with large and competitive research programs by, for example, the European Commission . The Internet of Things is seen as the post-smartphone business ecosystem for all stakeholders that are involved with communication-related products and services. Some more bold views  state that the Internet of Things will bring an end to the capitalism as we know it, by facilitating zero marginal costs for products and services. Regardless of how profound the consequences of the Internet of Things will be, we are moving to an age where everything that can be connected, will become connected. But what then, do we still need a research in communication engineering, say ten years from now? Or we just need to have engineers to maintain the systems? Will the exciting things happen only in areas that assume that connectivity is already there?
This is not only a speculative question, as there are clear signs that the communication engineering, and especially research in communication engineering, is losing its traction and gives way to other engineering disciplines, such as e.g. machine learning. I have not documented a statistically correct study to confirm these statements, but being an academic in communication engineering, I have talked in the past years to hundreds of my peers and concluded that we share the same observation. And the observation is that the number of students and Ph.D. students in communication engineering is seriously declining. The Deans are not hiring as many communication-focused faculty members. The role and popularity of communication-related courses in the curricula is rather modest compared to ten years ago. The research funding for the pure subject matter of communication engineering is declining. The latter has motivated (forced?) many academics to migrate to other areas that deal with probability and statistics and/or window-dress their communication engineering ideas with popular spices (big data, smart <something>, machine learning, etc.) in order to increase the chances of getting research funding. To avoid any misunderstanding - this is quite fine, part of the natural evolution of curious minds and broadening of the research horizons.
Yet, I still think that there are important research tracks that have the potential to produce new fundamental results in communication theory over a longer term. Furthermore, the rich methodology of communication theory can be purposefully used to achieve novel system design and optimization within other disciplines. Sure, there will be still the standard research track in communication theory in terms of improving transmitters and receivers, optimize the latency, or increase the rate. But what I am after are the tracks that have the potential to drive a new and exciting research in communication theory over a longer period. Having no ambition to be comprehensive, I have picked three such tracks. None of them is new, as a diligent reader can immediately find the “this has already been done” references, but I have selected them due to their potential.
Communication under adversarial impact. A majority of the works in communication theory have been dedicated to optimize the systems with respect to the physical constraints, i.e. the “nature,” such as noise, unintended interference, etc. The communication theory community has paid much less attention to security and optimization of communication systems against impact of adversaries. In 1949, Shannon (yes, the same Claude E. Shannon) published a fundamental paper on the topic of security , but this one has received less attention from communication theorists compared to his 1948 paper , the holy script of information and communication theory. At the time of writing this, Shannon’s security paper has ten times less citations compared to his 1948 paper. On the other hand, the main problems related to communication today in the coming years, will be likely related to security. Cybersecurity can impact elections, economic outcomes, and war conflicts. The growing importance of bitcoin and other cryptocurrencies is heavily relying on a reliable communication infrastructure. There is a significant potential for communication theorists to build models and address the fundamental problems in cybersecurity, ranging from how to deal with fake news and up to protecting the communication infrastructure for smooth transactions of bitcoin and other cryptocurrencies. The advances in communication with respect to the impairments that come from the “nature” will happen at a much slower pace compared to the advances that need to happen in the race with the “bad guys.” In other words, when dealing with adversarial impact instead of dealing only with physical limits, it is much more difficult, if at all possible, to reach a saturation point for research, since the adversary is doing his/her research as well. This is why this area has a long-term perspective for research in communication engineering and this opportunity should be seized. Here I am not speaking about “physical layer security”-like research, which is fine by itself, but very much similar to the traditional problems in communication and information theory. I am instead referring to pursuing new models and theories that can capture the large-scale cybersecurity problems.
Interaction among vertical services. Communication and networking technologies are based on the layering paradigm. The layers can be thought of as a hierarchical tower of bricks, with the brick underneath representing a module that providing a certain service for the brick (module) above it, see Figure 1(a). An end-to-end connection is built by creating a vertical pipe that goes vertically through all the layers (bricks) and each layer needs to be configured in a way that supports the specific requirements of that connection in terms e.g. rate, latency, etc.
Figure 1. Simplified layered communication models. (a) Layered model for a single service. (b) Layered model in which services reuse the modules of the lower layers. (c) Layered model for virtualized nodes and functions.
While it is obvious that layering is the basis of networking and the internet, it is less obvious that layering is also a key idea discussed in Shannon’s 1948 paper, although not explicitly referred to by using the term "layering". One of the main insights of Shannon was that source coding and channel coding do not need to be done jointly, but each of them can be done in a separate module (layer). Shannon, and therefore information theory, has mainly been concerned with questions of a type: how to design a communication strategy when a given layer can be used many times? The naïve way of communication in a layered system would go as follows: whenever a layer communicates something (call it a symbol) to the lower layer, the lower layer should code it and send it further to the layer below or to the receiver on the other side. The central idea introduced by Shannon is the one of coding by using a given layer multiple times. That is, a layer can aggregate multiple symbols that come from the upper layer, code them jointly, and send them as a larger message to the lower layer or to the receiver. This way of communication is statistically more efficient.
Besides this, in a context of networking, the power of the idea of layering/modularization is in the fact that a layer (module) can be reused to support different needs and services for the upper layers, see Figure 1(b). For example, the lower modules in this example (Module 1-1 and Module 1-2) are the Wi-Fi chips on the phone, while the modules at the higher layer (Module 2-x) are different apps. A research question that has attracted much attention in the past two decades was the cross-layer optimization: for example, how to adapt the operation of a certain module at a lower layer in order to support a certain service from the higher layer in a better way.
In the recent years, there has been a growing trend of Network Function Virtualization (NFV) and Software Defined Networking (SDN). A simplified representation of a virtualized communication system is shown on Figure 1(c). Here physical modules are used to instantiate functionalities that are used in virtual modules and layers. Assuming that multiple connections are using the same set of modules that are used to build tower of layers, how to configure the layers in order to support multiple vertical pipes? This is a different question from the one tackled in communication theory, but a very important one, as it starts to be dominant in communication systems. And, of course, the main question related to the theme of this article: What are the fundamental communication-theoretic problems that arise from interaction among multiple vertical connections, each with different requirements? For example, one connection may require a low latency at a fixed rate, another one aims to maximize the rate while having loose requirements on the latency.
The layering paradigm is deeply rooted in communication systems, stretching to areas as spectrum management/regulation and net neutrality. In spectrum regulation, and especially for unlicensed spectrum, there is the tendency to make spectrum usage rules generic, as much as possible independent of the requirements of the vertical service. Net neutrality also implies that the communication layer should be oblivious to the vertical service that uses the layer. Building a communication theory that addresses the fundamentals of interaction among multiple vertical services, has the potential to change the way we configure communication systems, as well as utilize and regulate the spectrum and communication services.
Communication beyond dedicated electronic circuits. Communication engineering is a sub-discipline of electronic engineering, while communication theory leans towards a mathematical sub-discipline. The latter means that communication theory is not necessarily applied only to electronic circuits that are built for communication and has wider implications. As for example, in the recently established area of molecular communication and nano-networks. A mathematical theory of communication can be applied to any system in which a state or a variable should be transferred from one point in space-time to another point in space-time. If it is the same spatial point we are observing in both instances, then we are talking about writing and reading in a memory. To specify the system, one needs to identify:
The transmitter: capable to modulate or “write” that state or variable onto some carrier or actuator, such as radio wave, a water flow, or even a wall paint;
The receiver: capable to measure or sense the carrier modulated by the transmitter;
Statistically characterize the channel, i.e. what the receiver measures for a given input by the transmitter.
Using this rather abstract approach, in my group we have recently worked  with methods that send information by modulating the power electronic components, such that a distributed power system as e.g. microgrid can operate and exchange messages without relying on external communication systems. In other words, communication without using dedicated electronic circuits. This is only a small example of what could be a large opportunity in identifying, building and optimizing communication systems by relying on the existing actuators and sensors in a given system.
One important aspect I have not touched upon is - how the developments of Artificial Intelligence (AI) and machine learning will affect the design of future communication systems. But this aspect deserves a separate article, since AI will disrupt and change a large set of disciplines, such that communication engineering will be part of the whole changed context. Instead, I have tried to focus to the core area of communication theory and engineering and outline potential areas in which the present methodology can be reused and evolved further.
- The World Bank, “World Development Report 2016: Digital Dividends”, Washington, DC: The World Bank, 2016.
- The 5G Infrastructure Public Private Partnership (5G PPP), https://5g-ppp.eu
- J. Rifkin. The Zero Marginal Cost Society: The Internet of Things, the Collaborative Commons, and the Eclipse of Capitalism, Palgrave Macmillan, 2014.
- C. E. Shannon, “Communication theory of secrecy systems,” Bell Sys. Tech. Journ., vol. 28, pp. 656-715, 1949.
- C. E. Shannon. A mathematical theory of communication. Bell System Tech. J., vol. 27, pp. 379–423, 1948.
- M. Angjelichinoski, Č. Stefanović, P. Popovski, H. Liu, P. C. Loh and F. Blaabjerg, "Multiuser Communication Through Power Talk in DC MicroGrids," in IEEE Journal on Selected Areas in Communications, vol. 34, no. 7, pp. 2006-2021, July 2016.
1The title of this article was inspired by the discussion with Prof. Bertrand Hochwald at Globecom 2016.