Recent research has highlighted the requirements for edge orchestration and processing in Beyond-5G and 6G networks. The nature of the edge networks seems to fall into two camps. One of these is familiar to traditional cellular networks: a hierarchical, dense network of stationary cells with additional compute capability at the edge. The other is ad-hoc, distributed, and, to a great extent, self-organising.  

Examples of the former research include:  

• A Novel Genetic Service Function Deployment Management Platform for Edge Computing 
• 5G network-oriented hierarchical distributed cloud computing system resource optimization scheduling and allocation 
• An Enhanced Reinforcement Learning Approach for Dynamic Placement of Virtual Network Functions 
• Elastic resource management and network slicing for IoT over edge clouds 

Meanwhile, ad-hoc proponents have produced research papers such as these:  

• Three-Tier Architecture Supporting QoS Multimedia Routing in Cloud-Assisted MANET with 5G Communication (TCM5G) 
• Assessment of Available Edge Computing Resources in SDN-VANETs by a Fuzzy-Based System Considering Trustworthiness as a New Parameter 
• Blockchain-Empowered Framework for Decentralized Network Management in 6G 
• Federated learning meets blockchain at 6G edge: a drone-assisted networking for disaster response 

The question for the future of 6G networks is how to reconcile these two visions of the edge. It is clear that the evolution of vehicles as compute and storage platforms as well as the forecast growth in sensor networks has produced a requirement for some forms of coordination and cooperation between ad-hoc networks and a traditional cellular network.  

The tension is very real according to Alex Reznik, Chair of ETSI MEC, who said, “Personally and intellectually I have always been a big fan of decentralised approaches. When done right they lead to much more robust systems that can withstand attacks, multiple failures and adapt to changing conditions much better. The main challenge is to do it right.”  

On the other hand, Reznik said, “There is… a trend to centralise because for a lot of this you need somebody who’s willing to make a very substantial investment – we’re talking billions of dollars today – plus potentially purchase spectrum. They’re going to want to have control.”  

In practise both models – the “fixed” cellular network and the ad-hoc network – will need to co-exist and interrelate. The key question is how to make this happen efficiently in a way that delivers what all parties need. 

Sensor Networks 

IoT sensor networks are expected to create one pattern of ad-hoc or mesh networks at the edge. In this case Reznik believes that IoT gateways create an effective solution as they effectively mask the complexity of the IoT network from the “fixed” cellular network. The on-device intelligence in a sensor network is typically expected to be minimal, both as a way to drive down the cost of sensors and also to extend their lifespan, which means that they will likely depend on intelligence in the edge and cloud.  

Upkar Dhaliwal, Analyst and System Architect at Future Wireless Systems, agrees but points out that, “I was doing IoT gateways in 2008-2009… It works well for a handful of nodes that won’t change,” – a far cry from either emerging public safety networks such as the U.S.A’s FirstNet, Vehicle-to-Everything (V2X), or Tactical networks.  

Vehicular Ad-Hoc Challenges 

With more complex ad-hoc networks such as those created by connected vehicles, whether road vehicles or unmanned aerial vehicles, there is a question over the role of the wider cellular network. As Dhaliwal points out, “The real challenge you have in moving to a wireless world is that we’re only just getting to deterministic resilience in the wired world where you can manage and control your nodes and node density; So, just think how difficult it will be to do five-nines reliability wirelessly where the nodes are moving at velocity in three dimensions”.  

As a result, Reznik explains, “What seems to be emerging is a [Vehicle-to-Vehicle] communications system which is meshy but it’s actually mostly sensing…in the same way that you have an active radar for airplanes. Then you have access to a system which looks like a cellular system which can help with some coordination, but mostly it collects a lot of the sensing data and creates a more global picture and feeds that back to the cars. Ultimately the decision-making is largely autonomous within each vehicle.” 

Even without being used for mission-critical applications, Dhaliwal points out the challenges facing current IEEE-backed V2X standards such as DSRC, “Imagine being at a busy junction like just outside Los Angeles airport, where they have 12 lanes. Just imagine 24 lanes of traffic with cars that are moving or stationary. Do you think that V2X can keep up with that? It’s going to take a few more years to crack that one.”  

As a result, IEEE’s Working Group 802.11bd NGV is working closely with 5GNR, the radio technology developed for 5G networks. Meanwhile, ETSI is also working towards solving the questions presented by the automotive industry.  

“In today’s state of the industry, [automotive] is probably one of the more engaged verticals that ETSI MEC is working with,” Reznik commented. “We have an ongoing collaboration with 5GAA [the 5G Automotive Alliance], and so we’re taking into account what they want to do, digesting it and extracting what we need to do.  

“The reality is there are two things; there’s a service information API that we’ve standardised specifically serving V2X, and it’s just about exposing information that exists in telco networks that would be specifically useful for these situations. The other aspect is the mobility. As the car goes from one point, there’s an application instance that’s supporting it; People always ask ‘how do you move that application instance as the car moves?’ So, we actually have standardised an API that allows you to move an application instance from a MEC platform to a MEC platform. 

“Having said all that…there has to be a superbly good reason why you’d ever want to move context. If you think about everything involved in that, it’s a big deal. A well-designed cloud native application provides you a much easier answer. The client in the car has the state, has the context, and perhaps the centralised application over in the central public cloud has some of it as well, but the services that you place at the edge are completely oblivious to it.” 

Dhaliwal commented that “the AECC [Automotive Edge Computing Consortium] are also working to address where cellular, automotive, and far edge computing converge.” 

Beyond 5G Integration 

Bearing these constraints in mind, there are arguments for wanting more advanced ad-hoc networks to be either essentially divorced from the “fixed” cellular model or considering them an added demand on the system. However, can vehicular ad-hoc networks provide a resource to support the development of future networks beyond 5G? A collaboration by the UK’s University of Essex and Taiwan’s National Cheng Kung University argues in favour of exactly this concept. 

“Roads and mobile vehicles form an important part of our base infrastructure. A Connected Autonomous Vehicle (CAV) system that is composed of smart roads, roadside units (RSUs), and vehicles can provide considerable resources, physical space and services for communication, computing and intelligence,” it said. “The unique CAV features such as controlled mobility and ease of deployment can strongly support the functions of 6G systems through infrastructure extension, monitoring and maintaining 6G networks, to achieve the 6G goals of ubiquitous wireless intelligence and minimize network operation costs. With expected tremendous investments on the 6G and CAV infrastructures, the 6G communications and CAV systems could be jointly designed, planned and operated with a much better reuse of system resources and services.” 

There are practical limitations to this in the short term. Zahid Ghadialy, Parallel Wireless’ Senior Director of Innovation & Technology, pointed out in an interview with 6GWorld that the lifetime of most vehicles is around 15 years, meaning that it will be many years before a significant proportion of in-vehicle infrastructure is fully capable of contributing.  

Meanwhile, Dhaliwal is concerned about the lack of global coordination within the automotive industry. “Do you think Stuttgart, Nagoya, and Detroit, three thiefdoms of the automotive industry, will have the vision to let other people get involved? It’s not a technical decision, it’s a business decision. So, perhaps AECC and 5GAA will merge. Some of them are more policy and regulatory driven… the Japanese guys may have the vision, but they tend to be quite slow to move. German industry is leading at present”.