More equipment, more advanced technology, more coverage and capabilities. There is a clear line of progress as each new generation of wireless technology comes into use, from 3G through LTE and advanced 4G, and now 5G, the next big thing in wireless.
The conversion from 3G to 4G over the past few years saw the deployment of more cells in various form factors, as well as new technologies, such as carrier aggregation and full-packet switching, making 4G networks far more complex than 3G. More complex but still manageable.
With the move to 5G, however, that manageability will really be put to the test. 5G networks will see the deployment of an order of magnitude with more cells (interlaced into the existing fabric of 4G LTE macro-cells) and antennas (both passive and active ones that utilize multiple frequencies), with more advanced technologies such as vRAN, which will allow for the partial or even full virtualization of networks, allowing faster and cheaper upgrades and innovations in service. 5G will also allow for the introduction of other technologies to handle a huge increase in capabilities, all in a multivendor environment — with an emphasis on enabling the ultra-low-latency data transmission that 5G is designed for.
With a more complex network comes more complex problems — in the sense that tracking down the cause of a service outage or other issue will be more difficult because of the system’s complexity. More importantly, the plethora of possibilities makes planning and managing a complex ecosystem more difficult, from determining how best to supply services based on current and future demand to where and how to deploy equipment and services in the most efficient manner possible.
Such a complex network requires a simplified solution, and deploying artificial intelligence (AI) and machine learning (ML) to solve these problems could be the solution carriers need to ensure they remain in control of their networks. AI/ML solutions would enable carriers to bring the benefits of 5G to customers, as well as aligning to a CTO’s KPIs, overcoming the current situation in which monolithic, centralized networks strain themselves to deliver the advanced services customers are increasingly demanding.
AI solutions, of course, use machine learning to discover patterns in large datasets, providing a clear picture of cause and effect in resource usage to enable data enrichment for better prediction and decision-making timing, demand and a thousand other factors that determine the quality of connections, whether voice or data. With such solutions, carriers will be able to get a full and accurate picture of the state of their network. And based on current conditions and coming conditions, AI/ML solutions can help carriers determine where and how to deploy resources to avoid demand crunches and potential service disruptions.
How can AI/ML improve network performance and unleash the full potential of 5G? One of the things artificial intelligence does well is efficiently deploy the technology, determining what conditions are required in order for equipment and algorithms to perform most effectively. For MNOs, that could mean analyzing overall demand, using that information to plan for 5G and MIMO site configuration, base stations and other hardware.
But AI/ML will do much more for 5G networks. Artificial intelligence and machine learning will unlock the power of software and algorithms that will allow for an efficient deployment of assets and resources. AI/ML solutions will be able to instantaneously analyze all relevant data and determine how different elements of the network — hardware and software — interact, and how they perform under specific circumstances, so that resources can be deployed to ensure quality of service and end-to-end latency control.
With that, the new technology won’t necessarily solve some of the old problems. Security, of course, will remain a major concern, with hackers developing new methods to attack 5G systems. According to a 2020 Swedish study, “Data security related to wireless networks faces different challenges in comparison to traditional systems since it involves over-the-air information transmission which increases the risks of eavesdropping, false base stations, and jamming attacks.”
Using these methods, hackers can generate false data that “trusting” AI/ML systems will accept as accurate and process as part of the database they work from, thus polluting the data and yielding incorrect information. Sophisticated hackers could thus completely compromise data on usage or demand, creating major gaps in service that could cause large losses for clients and the carrier itself.
Another issue for AI/ML systems in 5G networks is how the sheer volume of data will be analyzed. For machine learning to perform its magic, it needs lots of data. The more the merrier, and that data has to be stored somewhere. Assuming large amounts of data will be uploaded not only by phones but by IoT devices too, carriers will need far more storage space than they currently have in order to process that data.
Processing such large amounts of data will also require far faster computers (a major expense for carriers), which could lead to latency issues, perhaps even canceling out the gains in latency realized with faster 5G technology. One approach to dealing with this is the use of federated learning, in which the model developed by an AI/ML system for service improvement is tested against the data on edge servers, resulting in much faster data processing times.
The key to all this, of course, is automation. To realize the true benefits of AI/ML for 5G networks and all that they encompass, analysis should be automated, taking into account ever-changing conditions and shifting usage demands and resource availabilities, along with anticipating future needs and pressures. Automated AI/ML systems will be able to provide a clear and always-current picture of the state of the 5G network, ensuring that resources are deployed as needed, providing the best QoE possible and unleashing the power of 5G for the benefit of all users.