AI-driven 5G Autonomous Networks: To Infinity and Beyond!

Touchdown confirmed

About a month ago, when all around the world billions of people hushed on Feb.18, at about 3:55 p.m. EST, to hear NASA aerospace engineer Swati Mohan calmly called the play-by-play of the Mars 2020 Perseverance rover landing that slowed to a gentle 1.7 mph at Jezero Crater and came to rest. “I remember watching my first episode of ‘Star Trek’ at the age of 9 and seeing the beautiful depictions of the new regions of the universe that they were exploring,” she said. “I remember thinking, ‘I want to do that. I want to find new and beautiful places in the universe. The vastness of space holds so much knowledge, that we have only begun to learn.”

The human race is in constant pursuit of exploratory and scientific advancements. Each technological revolution throughout human history has been an iteration in productivity and driven society to a new era. Over the past 120 years, the Industrial Revolution, Electricity Revolution, and Information Technology Revolution have been the three major breakthroughs in human civilization. Now, a new revolution is emerging, known as the Fourth Industrial Revolution. This revolution is driven by Artificial Intelligence, 5G, and Cloud Computing, and will create a world where all things are sensing, connected, and intelligent. It's coming faster than anyone could imagine. 

Today technological advancement in the Telecom Industry is at the turning point of network automation with the Autonomous Networks initiative propelling comprehensive upgrades driven by the large-scale deployment of SDN, NFV, and 5G. The concept of Autonomous Networks is of a network that provides the service lifecycle on demand with minimal or no human intervention. The network can configure, monitor, maintain and repair itself independently. By making use of emerging technologies such as cloud computing i.e. essentially unlimited parallel computing processor resources and effectively infinite storage, together with on-demand bandwidth that provides dynamic connectivity, utilizing big data solutions to gather and analyse data from the ubiquitous array of network sensors for better pattern matching of potential service interrupting instances, and rules, model and intent driven software decision engines that learn (via analytics and rules, machine learning or eventually the intent by artificial intelligence) how to maintain service and repair resources in the event of some form of service or resource failure the concept of a fully autonomous network is becoming a reality.

Why are Autonomous Networks becoming so important?

According to IDC statistics, 63.5% of telecom organizations are investing in AI to improve their infrastructure construction. Ovum predicts that, by 2025, the telecom industry world- wide will invest US$36.7 billion in AI software, hardware, and services. The overall market value of AI use case software in the telecom industry will increase from US$315.7 million to US$11.3 billion in 2025, with a compound annual growth rate of 48.8%, quickly propelling the industry to become the world's largest AI application industry.

Traditionally, tasks in areas such as network monitoring, anomaly detection, optimisation and problem solving have been carried out by human engineers, but with the rollout of 5G and IoT, and the massive increase in the number of connected devices and applications they are expected to bring, this will no longer be practical. On the supply side, Autonomous Networks would simplify the industry chain structure to enable a built-in flexible production mode for creating on-demand services. On the demand side, it would embrace a digital ecosystem with innovation agility, enabling operators to effectively handle high-value yet complicated requirements from customers and markets.

Figure-1 below, presents a closer look at use cases where Autonomous Networks add value across different disciplines:

Figure-1: Business Capability Model

Autonomous Networks – Digital Transformation Framework

As COVID-19 continues to cause a profound impact on digital transformation, by promoting simplification, automation, intelligence, and collaboration, the telecom industry requires an improved production architecture capable of providing on-demand network services to meet flexible and situational customer requirements. In other words, the telecom industry has entered an architectural innovation phase. As production equipment becomes autonomous and management layers are flattened, operators are developing a "products on-demand, services orchestrated" mode, and creating a new digital ecosystem, enabling industry transformation through open innovation and collaborative production. This ecosystem not only improves the economic efficiency of the telecom industry, but also acts as an autonomous digital infrastructure platform facilitating the smart society and digital economy.

Figure-2 below, presents an Autonomous Networks Framework describing different layers of Digital Transformation:

Figure-2: Digital Transformation Framework

According to TMForum (represented in Figure-3 below), the business vision of Autonomous Networks is to provide innovative ICT services and capabilities with “Zero X” (zero wait, zero touch, zero trouble) experience for the users of vertical industries and consumers, which makes them simpler to consume by the users, and leaves the implementation complexity with the providers. 

Figure-3: TMForum's Zero X Model

As per the report, Autonomous Networks are able to provide two types of innovative services:

• As a Service: One stop, real-time, on demand, automated, E2E full lifecycle network/ICT services 
• As a platform: Enablement of business collaboration & ecosystem between verticals and network/ICT service providers

They can also enable highly automated business and network operations of “zero x” experience for innovative services as well as existing services.

The Rise of the Machines

To achieve Autonomous Networks, the zero-touch system must be able to handle the complexity caused by continuous changes to the system at the same time that it delivers services to users and manages issues such as the cost of resources versus the budget available, the legal compliance of the service and the security of the setup. This is challenging for technical systems in real-world scenarios. Autonomously operated and self-adapting networks will make it possible to utilize the capabilities of 5G networks in new business models and achieve an unprecedented level of efficiency in service delivery. Intents will play a critical role in achieving this zero-touch vision, serving as the mechanism that formally specifies what the autonomous system is expected to do.

Everything an autonomous system needs to know about its goals and expected behaviour must be defined with intents. The system will not perform any operation unless it relates to the fulfilment and assurance of an intent, which means that all goals – including those that may have been considered “common sense” in human-operated systems – must be expressed as intents. An intent in an autonomous system is ideally expressed declaratively – that is, as a utility-level goal that describes the properties of a satisfactory outcome rather than prescribing a specific solution. This gives the system the flexibility to explore various solution options and find the optimal one. It also allows the system to optimise by choosing its own goals that maximise utility.

Figure-4 below, presents a Data Analytics life-cycle management overview – highlighting the role of AI with dynamic network operational and customer usage data event streaming:

Figure-4: Data Analytics Life-cycle Model

And the Figure-5 presents, an efficient one-stop model training and quick model verification as it integrates processing capability for telecom domain features (meaning that key features can be quickly identified), and has built-in AI algorithms for exception detection, root cause analysis, optimisation control, and service prediction:

Figure-5: Data Analytics AI Training Model

These training platforms are preconfigured with different classes of 30+ pre-integrated telecom model services including traffic prediction, KPI anomaly detection, intelligent control, and alarm correlation. Experience in the telecom domain is standardized as services, and users can obtain models by simply entering data. The wizard-based model development process provides telecom templates for data preparation, feature extraction, and model training, improving training efficiency. The platform integrates the telecom knowledge graph, with more than 50 built-in data analysis tools. Telecom experience is converted into tools to facilitate decision making by experts.

Conclusion

The Autonomous Networks initiative defines a new paradigm for network architectures and operation & business models. It optimizes the production cost structure and improves the E2E production flexibility and business agility of digital services. As such, it can aggregate the efforts of all parties within the telecom industry.

It can be said that to the telecom industry, the Autonomous Networks initiative embodies Industry 4.0 in the manufacturing field.

And finally, at PCCW, what’s compelling is that combination of large-scale software automation capability, from one of the largest networks in the world, combined with our users’ ability to provision connectivity from any data centre to any data centre in over 100 countries around the world; to orchestrate connectivity from one user to another on the platform seamlessly and fully automatically; and to provision connectivity to Amazon, Azure, Google, Alibaba, Tencent and IBM, fully automated in minutes. We think we’re the largest network in the world to provide that level of automation.

– Ashish Kar

Author is the CTO – Platforms at PCCW Solutions with 24+ years in the ICT industry and an innovation gameplanning coach. He has been helping global operators to digitally transform their businesses and monetise and grow their subscriber base using cloud native architectures. He can be reached on email at [email protected].