After years of research, we can finally say that 5G has been born. It might be a premature birth since it’s a non-standalone version of 5G that 3GPP finalizes this December. On the other hand, which newborn doesn’t need support in the beginning? (In this case by relying on an LTE core network and an LTE anchor for the control plane.)

Being an academic researcher, I’m proud to say that academia has played an instrumental role in developing key concepts for 5G, such as algorithms for mmWave beamforming and Massive MIMO spatial multiplexing, and defining emerging use cases; in particular, massive machine-type communications (mMTC) and ultra-reliable low latency communications (URLLC). However, when I visited IEEE GLOBECOM in Singapore earlier this month, it was clear that the industry is now completely in charge of what 5G will be. For example, there were two panels dedicated to Massive MIMO. One would imagine the panelists being academic professors arguing about the number of ADC/DAC bits needed per antenna and how to mitigate pilot contamination. Instead, eight out of ten panelists were from the industry. The discussions in the panel “Massive MIMO – Challenges on the Path to Deployment”, which I describe in detail in another blog post [1], revolved around the disparity between the academic view and what the industry has already standardized. The industry will call any system with at least 32 antennas Massive MIMO, irrespective of whether it is implemented using codebooks for channel feedback, beam-steering, or by exploiting TDD channel reciprocity. A narrower definition is used in academia since our role is to strive for “optimality”. From that perspective, Massive MIMO must be a TDD system that utilizes channel reciprocity to obtain high channel estimation quality [2].

The point is that 5G has been designed to keep many options on the table and, therefore, only vaguely resembles what people have predicted [3]. For example, despite all the fuss about 5G being mainly about mmWave communications, the first deployment appears to be fixed wireless access at 15 or 28 GHz [4], where the wavelength is 1-2 cm…

Five years ago, the grand vision was that 5G would be a highly disruptive technology [5]. We now know that 5G is born as an evolution of LTE. The envisioned technology components are in place but in an elementary form. This is, of course, rather natural; there is no need to deliver a new network generation providing 1000x higher area capacity from one day to another. If the first release of 5G delivers 3x higher area capacity and basic support for mMTC and URLLC, that would be enough for the time being and 3GPP can continue evolving the technology as the data traffic grows and new business opportunities emerge. Maybe ten years from now, 5G will resemble what visionary researchers have analyzed and simulated – or maybe the technology will evolve in some other unexpected direction.

While the industry program at GLOBECOM focused on the path towards 5G deployment, a main discussion item around the coffee tables was: What to do next? Since the main components of the 5G physical layer have been set, there is a risk that further physical layer studies from academia are practically irrelevant. One answer that I picked up from the GLOBECOM panels that I attended was that we should shift focus to the higher layers, which have received much less attention from academia and been less standardized so far. From a more long-term perspective, the key question is: which new research directions in communications engineering deserve the attention from the thousands of academic researchers that once worked on 5G?

While in Singapore, I heard many people talking about the potential application of machine learning in communications. Interestingly, while communication researchers attended GLOBECOM, 8 000 machine-learning researchers gathered at the annual conference on Neural Information Processing Systems (NIPS). Machine learning is a key component in face-recognition and self-driving cars [6], among many other futuristic applications. Inspired by the apparent success of such applications, it is tempting to apply similar tools in communications. However, we must keep in mind that machine learning is a method to train computers to solve a given task as well as possible, without having an explicit system model or proof of optimality. In contrast, in communications, we are used to having concrete and reliable system models that allow for finding the optimal solutions. Maybe machine learning will play an important role also in communications engineering, but so far, I haven’t seen any convincing applications.

Another exciting prospect is the use of drones in communications. One option is to use Massive MIMO technology to communicate with drone swarms [7] that, for example, could be filming a sports event from many different directions. Another option is to use drones as flying base stations [8], which can densify the network infrastructure on demand. While these applications sound like science fiction, the technology for building them is rather mature.

The Research Vision Panel, organized by Prof. Lajos Hanzo at GLOBECOM, provided further suggestions of new topics in communications that deserve the community’s attention. Visual light communication seems to be necessary if we want to deliver data rates at the order of 1 Tbit/s in the future. Molecular communication is an unconventional new research direction that uses chemical signal rather than electromagnetic signals. Moreover, I believe that security and robustness towards jamming will be important as the society becomes increasingly dependent on wireless communications.

Clearly, there are new research topics in communications engineering that call for our attention now when 5G has materialized. The future will continue to be exciting.

References

1. Emil Björnson “Challenges on the Path to Deployment” http://ma-mimo.ellintech.se/2017/12/07/challenges-on-the-path-to-deployment/
2. Emil Björnson, Jakob Hoydis, Luca Sanguinetti (2017), “Massive MIMO Networks: Spectral, Energy, and Hardware Efficiency”, Foundations and Trends® in Signal Processing: Vol. 11: No. 3-4, pp 154-655. http://dx.doi.org/10.1561/2000000093
3. Jeffrey G. Andrews, Stefano Buzzi, Wan Choi, Stephen Hanly, Angel Lozano, Anthony C.K. Soong, Jianzhong Charlie Zhang, “What Will 5G Be?,” IEEE Journal on Selected Areas in Communications, vol. 32, no. 6, pp. 1065-1082, June 2014.
4. AT&T, Verizon want to continue testing 5G in 28 GHz band, https://www.rcrwireless.com/20170816/5g/att-verizon-5g-testing-28-ghz-tag17
5. Federico Boccardi, Robert W. Heath, Angel Lozano, Thomas L. Marzetta, Petar Popovski, “Five disruptive technology directions for 5G,” IEEE Communications Magazine, vol. 52, no. 2, Feb. 2014.
6. Apple Executive Reveals More of its Self-driving technology, https://www.wired.com/story/apple-executive-reveals-more-of-its-self-driving-technology/
7. Prabhu Chandhar, Danyo Danev, Erik G. Larsson, “Massive MIMO for Communications with Drone Swarms,” IEEE Transactions on Wireless Communications, to appear.
8. Zdenek Becvar, Michal Vondra, Pavel Mach, Jan Plachy, David Gesbert, “Performance of Mobile Networks with UAVs: Can Flying Base Stations Substitute Ultra-Dense Small Cells?” European Wireless, 2017.