6G may still be a decade or more away, but already exciting conversations have arisen about its potential capabilities. Among those include concepts surrounding wireless intelligence in network communications. It’s expected to be big – if not the defining feature of the next wireless generation.
It’s becoming increasingly clear that our current infrastructure paradigm – that is, the way we’ve organized and built our networks at the hardware level – won’t be able to deliver the features we envision for 6G sustainably. 5G may have solved the issue of speed and bandwidth sizes for transferring massive data rapidly, but the energy requirements for 6G render the current infrastructure unwieldy.
That’s led many experts to reconsider the way we organize and build our wireless networks. Here’s a closer look at the infrastructure limitations for 5G, and what future possibilities might entail.
What Is Wireless Network Intelligence?
Wireless network intelligence refers to the use of machine learning and artificial intelligence in wireless communications networks. It’s one of the biggest topics currently being discussed regarding 6G.
An intelligent wireless network on its own is not a new concept – we’ve been working towards it for some time. With data having become a key asset in many industries and sectors, most of our technological research around wireless networking in the recent past has related to the development of a network of devices capable of handling it. (You know this as the Internet of Things.)
5G solved a significant hurdle in this research by delivering the minimum speeds (and lane widths) necessary to make the IoT possible. 5G will augment the presence of the IoT in the world, but its limits have already proven glaringly obvious.
That’s because there’s another massive problem that even 5G cannot solve. It lies in the fact that our current approach to machine learning uses a centralized infrastructure relying on a remote data center.
Currently, “smart” networks and machine learning-based platforms leverage massive cloud storage facilities housing every piece of information related to the platform’s functions. When a user deploys machine learning to analyze data, the platform must first pull the information from the cloud and then sift through it all.
This produces extreme demands for energy, memory, and computing resources. These typically manifest for the user as latency and occasionally massive utility bills.
For some smaller applications of the IoT and machine learning, the centralized approach works. However, for high-stakes applications such as virtual reality and autonomous vehicles, even a millisecond latency presents unacceptable levels of risk. Unfortunately, combined with the resource consumption, intelligent wireless networks, as they’re currently realized, are unwieldy at best in these scenarios.
Some early research suggests that 6G may have an answer to this problem. It lies at the edge.
Rethinking Intelligence With Machine Learning at the Edge
A potential future full of things like immersive VR and fully autonomous vehicles has inspired much interest in developing an alternative network configuration. Currently, a major focus is a distributed, decentralized model that eliminates the cloud-based storage entirely.
In the “edge” model, data is stored not in the cloud, but across a massive constellation of devices operating at the “edges” of the network. These would include network base stations, mobile devices, drones, computers, and other smart devices. Every device develops its own locally trained model from its own data, and it’s these models that get shared across the network instead of the raw private data.
It’s fundamentally similar to the way crowdsourcing works.
The decentralized edge model has several obvious advantages. Notably, it eliminates the latency generated by sending data to the cloud for analysis by allowing devices to run machine learning inferencing directly. However, it also delivers critical advantages for data security as data is not moved, only inferences. Finally, it will improve reliability further because no single point of failure in the network exists.
Although highly experimental, machine learning on the edge is currently regarded as the only way to develop an advanced intelligent network in wireless communications. It also requires significant changes in our current infrastructure that might make the full realization of a decentralized network the goal of a generation past 6G.
The Advantages of Wireless Network Intelligence
If 5G was supposed to make the IoT a practical reality, then 6G is estimated to make artificial intelligence ubiquitous, with applications in a variety of industries. We’ve already covered some of those right here.
Beyond that, intelligent wireless networks will fundamentally change the way we conduct our communications by providing ready access to both wireless data and the inferences drawn from it using machine learning. We can expect:
- More insights received faster. One of the interesting things about the decentralized edge model is that it sidesteps many of the problems, which led to the development of 5G in the first place. We’ll still be moving massive amounts of data very fast, but the edge model makes it easier without gobbling up yet more parts of the spectrum.
- More accurate insights. By tapping edge devices directly, we’ll enjoy increased accuracy, particularly with positioning or geolocation services.
- Increased emphasis on strategic priorities. The ultra-reliability of the edge model will eliminate many of the infrastructure headaches that businesses currently face when deploying machine learning and AI. We expect the speed of business to both accelerate and become more innovative.
- Embedded security. Data security is a perennial concern, especially when data is stored in a third-party location. A decentralized intelligent network offers the potential to deliver hardware-level data security, solving some of the biggest challenges of cybersecurity once and for all.
A New Idea About Intelligence: Wireless Networks & 6G
Wireless networks have a lot to gain from novel applications of intelligence, but the current centralized infrastructure design won’t be able to handle the resource consumption. Machine learning currently requires significant resources, including electricity usage, powerful servers, and storage space to accommodate the data from which it draws inferences. That makes deploying it for wireless intelligence difficult – if not outright impossible.
We expect the centralized model of connectivity to change to a decentralized model in the future. However, 6G technology remains years away from development, and other solutions may surface.
We’re keeping an eye out on information and ideas that surface as 6G research ramps up. Sign up for our newsletter to keep informed.