Authors: William Tärneberg, Gilles Callebaut and Liesbet Van der Perre

What’s in a name? ‘The spoken newspaper’ is the term grandma¹ used to call the news broadcasted on the radio. The curious name became a witness to how news media changed rapidly. In the same way, the term ‘Access Point (AP)’ literally becomes a somewhat crooked term in the context of 6G networks: nor will the infrastructure provide only access, nor will services be delivered via one point. This blog post clarifies why 6G applications and sustainability targets require, on the infrastructure side, a compute-connectivity platform, consisting of interconnected distributed resources. We introduce new essential terminology to enable a clear discussion in developing novel technologies, originating from lively academic-industrial interactions in the REINDEER project and illustrated in the figure below (spoiler alert!).

We, and the devices we use, rely more and more on information from other devices and sources, accessed wirelessly, and progressively dispersed around us and the globe. The vast population of devices generating all this data are generally referred to as Internet-of-Things (IoT) devices. They often inhabit the capillaries of the network, where they sense and report data, enabling us to make more granular and complex decisions. Therefore, in the past decade, the information centre-of-gravity has and is shifting from data centres at the heart of the network, to the edge of the network. Consequently, traditional cloud data centres are dispersing towards the edge, offering seamless computational capacity throughout the network, in the form of edge and fog computing. The potential number and types of devices that want to reliably connect, wirelessly, at low latency, and high throughput presents a huge challenge to wireless networks. Imagine, a throughput-hungry AR headset, that shares a space with a few thousand connected sensors, whilst being driven around in an autonomous vehicle. Satisfying the needs for that scenario, and many more like it is the precursor to the 6^th generation wireless networks (6G).

Researchers are in the midst of shaping the 6^th generation of wireless networks, capable of supporting divergent applications and especially services that go far beyond what the 5^th generation can offer. That effort includes addressing challenges in achieving remarkably high data rates, imperceptibly low latency, unrivalled dependability, and ultra-low power consumption [1,2]. Also, and with increasing urgency, the above should be accomplished with a negligible carbon footprint.  
Our vision for 6G is focused on realising large intelligent
surface (LIS) [3]. LIS is an extension of massive MIMO, where the number of antennas is increased several orders of magnitude and distributed throughout a three-dimensional space, rather than co-located on a plane. Accompanying the antennas are radio and computational resources. This spatial and computational diversity is the singular most significant distinction between the 6^th generation wireless networks and their predecessors.

With such diversity, the potential for 6G is monumental, but therein lies its greatest challenge. To seamlessly serve users with LIS, shared computational and radio resources will have to be dynamically allocated and orchestrated at a previously unseen scale and speed.
As there are no discrete base stations, all decisions will have to be distributed, which is referred to as a cell-free network.

In cell-free networking [4,5], all resources in the network can be used to offer a given service. It provides an interesting concept to fully use the available resource capacity. Although the theoretical potential of cell-free networking has been identified, many questions remain concerning how these systems can be designed to be practically feasible. For example, how can such a great pool of resources get coordinated and allocated efficiently? And how can both the infrastructure and the services provided in 6G be scalable?

Our incarnation of 6G is called RadioWeaves (RW) [6]. Cell-free networking and embedded edge computing are central components in RW. Further, as a means to begin to address the above challenges, in RW, we analogously group resources dynamically in both the temporal and spatial domains. In RW, we use the term federation(s) to denote such a group of resources, that jointly serve an application. In distinction to current wireless networks, the envisioned RW system, will also support precise positioning, wireless energy transfer and energy-neutral IoT devices [7]. Following this paradigm shift, a new terminology for the conventional AP is introduced, i.e., the contact service point (CSP). It provides a first contact point into the network from the perspective of the user equipment (UE) and supports more than just wireless communication, as mentioned above. To do this, CSPs can host a variety of hardware such as radio, charging, processing, data storage and other sensing elements.

The figure above illustrates the abstract notion of a massive number of spatially distributed CSPs, and their grouping in relation to users and applications, i.e., federations. Concretely, the figure shows an example deployment of RW in a smart factory, with four federations, shown in different colours. Before we proceed, we need to take note of the RW terminology used in the figure. RW CSPs are deployed throughout the production hall on the walls and ceiling and are dynamically assigned to federations to serve the devices and their running applications. The constellation of CSPs assigned to each federation is tailored to the particular application’s requirements. The video below goes into more detail about how we address this challenge.

Implementing RW is the next analogous step in bringing the benefits of 6G to society. We are therefore implementing two test beds, one located at KU Leuven, Belgium, and the other at Lund University, Sweden. The testbed at KU Leuven, Techtile [8], was inaugurated in October 2021 and already has the physical infrastructure in place, but work is ongoing on the software to manage and run the testbed. Techtile consists of a 4-by-8-by-2-meter room built with 140 modular panels, each of which contains software-defined radios, edge computing units, and sensors. The testbed is specifically designed to study scalable and low-cost RW systems. It, therefore, uses Ethernet for power, data, and synchronisation. Complementary, the Lund University testbed² is designed to assess high-throughput and latency-critical applications. The testbed consists of 16 panels, having 4-by-4 RF chains, equating to a total of 256 antenna elements.

More information regarding the concept and terminology can be found here: G. Callebaut, W. T ̈arneberg, L. Van der Perre, and E. Fitzgerald, “Dynamic federations for 6G cell-free networking: Concepts and terminology,” in 2022 IEEE 23rd International Workshop on Signal Processing Advances in Wireless Communication (SPAWC), 2022, pp. 1–5. doi: 10.1109/SPAWC51304.2022.9833918 [9].

The authors would like to thank the REINDEER team, and especially Emma Fitzgerald, Erik G. Larsson, Pål Frenger, Ove Edfors and Liang Liu, for the rich discussions that have strengthened the definition of the new terminology and realisation of the test beds.

¹Livine Cleemput, mormor of L. Van der Perre, original term ‘Gesproken dagblad’
² www.eit.lth.se/index.php?gpuid=325&L=1

The project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 101013425.