Intelligent connectivity will be a vital enabler of applications from artificial intelligence (AI) to augmented and virtual reality. It will be created from new 5G techniques such as network slicing and connect assets with intelligence at the network edge. Here, Padraig Scully, the chief research officer of IoT Analytics, discusses with Daniel Quant, the vice president of strategic development at MultiTech, how the concept of intelligent connectivity will move from testbed to reality.
Padraig Scully: Intelligent connectivity is a term that is being increasingly used to describe the combination of 5G, AI and IoT technologies. For the readers who haven’t heard of it, how would you define it?
Daniel Quant: Intelligent connectivity is the evolution of dumb, no smarts connectivity to a more resilient, resourceefficient, data communication architecture where decision making is distributed to the edge or assets making things quicker and easier to customise actions the way you want.
The reduced price point of sensors, tamper-resistant processors and memory have made this concept economically viable today. As have the maturity of real-time operating systems (RTOS) and device management services, particularly those developed for constrained devices.
PS: According to its promoters, intelligent connectivity will enable new capabilities and change the way we work and live. What are in your opinion the best examples to show how transformational this concept would be once it becomes reality?
DQ: Industrial edge intelligence use cases have begun with cost, efficiency and resiliency driving adoption. Looking ahead, factory automation, for example, will be improved, building on visual and sensory automated quality assurance, worker safety and improved outcomes based on ever more complex learning models and differing data points.
Autonomous cars is the most obvious future certainty based on intelligent wireless connectivity. Do we really expect every car to be sending 100GBs of data per day? Is that viable even on ‘6G’? Surely more intelligence and decision making locally, based on complex AI models that are periodically updated across fleets or regions, is the technology to mainstream scalability of autonomous vehicles.
And we’ve all heard about issues with virtual assistant devices in our homes. The ability to manage exceptions verbally to instruct how machines and systems need to perform complex operations will help reign in potentially sinister applications like face recognition in public places, using systems that have highly accurate updated models running locally, with target identities to look for, sent from a central platform and reported back to when and where identified.
PS: What is MultiTech doing to prepare for the arrival of this intelligent connectivity era?
DQ: For MultiTech, intelligent connectivity has been shipping to our customers for many years now in our programmable MultiConnect Conduit industrial gateways and cellular MultiConnect Dragonfly system-on-modules (SoMs) and LoRa technology-based MultiConnect xDot SoMs. All bring together secure processing and memory resources as well as operator-certified native wireless integration and RTOS for high-level programmability, based on well documented application layer application programme interfaces (APIs).
In 2019, MultiTech will bring to market more powerful programmable gateways and industrial routers capable of virtualising cloud/edge micro-services, enabling faster decision making, greater resiliency to network outages, and more efficient use of resources, thereby improving cost of ownership and opening up new applications.
PS: If we look at what is needed to enable intelligent connectivity, the main missing element today is the ultrafast and ultra-low latency connectivity promised by 5G networks. While some vendors claim that networks exploiting these capabilities may be available as early as 2020, others are more sceptical and believe it will take three to five years before this could happen. What are your expectations, realistically?
DQ: Absolutely! It seems each 5G press release or article is making more impressive claims than the last, whilst each release date appears to be getting closer.
The fact is, 4G-LTE Advanced Pro provides high broadband and improved latency, the time between making a request and receiving information, suitable for most applications in deployment today, with system latency averaging 50 or so milliseconds.
5G New Radio (5G-NR) has now been standardised by 3GPP as Release 15 in Sept 2018, targeting consumer-centric deployments later in 2019 with most operators not deploying until at least 2020.
The 3GPP 5G-NR system latency target is to be lower than one millisecond. Although significant progress was made on improved scheduling in Release 15, average system latency is at best sub ten milliseconds.
5G-NR Release 16, currently scheduled for a March 2020 release, has a Work Item specifically focussed on ultra-reliable low-latency connectivity (URLLC). Applications requiring time-sensitive networking, such as critical infrastructure – think oil and gas or smart-grid management – with submillisecond latency on 5G cellular networks will not be commercially available until at the very least 2022.
In short, 5G is initially targeted at the consumer and fixedwireless access markets with industrial applications becoming viable from 2022 onwards.
PS: One of the most discussed aspects of 5G is network slicing. What does network slicing mean in an intelligent connectivity context? How do you think the connectivity landscape will look like with network slicing?
DQ: Network slicing enables mobile network operators to provide a full turnkey service orchestration from edge to data center, if you like, a network-as-a-service, specifically dimensioned as many virtual networks built up from resource blocks tailored for each use case over a common physical network infrastructure. Clearly, connecting decision making from edge to cloud results in many complexities in between to reduce a physical separation to a virtual one.
Imagine slicing as enabling an Amazon style network model of just adding resources, in this case radios, servers and backhaul capacity on demand within a customers desired/agreed service level agreement (SLA).
PS: In order to provide the low-latency required to enable many of these use cases, such as autonomous cars, tactile internet or mobile VR/AR applications, computing resources will have to be distributed closer to the users. Can you tell us more about the relationship between intelligent connectivity and edge computing?
DQ: One relationship that perhaps does not hit immediately for users is that, as more intelligence is placed at the edge, more decisions are being made without your direct interaction, which does not always work out quite like you imagined. For example, my Nest is constantly thinking it’s smarter than me and adjusting my heating against me. Therefore, it’s imperative that solutions come with secure lifecycle management, to take ownership of, provision and customise behaviour from activation throughout the life of the asset in the field. Attention to usability and business models can make or break an application.
PS: Sticking to the 5G topic, many believe that the vendors are focusing too much on pushing the technology without having determined the actual needs of the industry and the society. And that as a result, once the technology is out, the audience ready to embrace it may not be as big as expected. What are your thoughts on this?
DQ: 5G for industry is still a few years off. That said, there is enough overlap between consumer wants and the needs of commercial business. So yes, 5G, even in the early days, will be used for improved satellite/home office connectivity, retail and fixed wireless media and Internet services.
PS: What are the main challenges that need to be addressed before the intelligent connectivity vision becomes a reality?
DQ: The reality is intelligent connectivity is still getting up onto its feet today, and has been for a while now. However, closing the gap between the device and the cloud continues to become ever more complicated. The industry is working hard to make this simple and to scale at pace, which requires efficient and seamless integration with AI platforms located in the cloud or the data centre utilising standardised interfaces and protocols and security architectures. It feels and probably looks very chaotic right now, but it will align, just as the Internet and media industries have managed to achieve.
PS: Looking more ahead, which industry sectors do you believe will be transformed the most by the intelligent connectivity era?
DQ: Energy, utilities, transportation, manufacturing, warehousing, mines and more. Look to industries that are underserved by AR and AI, and real-time decision making today which need to transform the most to remain competitive and relevant in tomorrow’s world.
PS: What role do you see MultiTech playing in the intelligent connectivity era?
DQ: Providing Industrial markets, often depending upon critical infrastructure, the assets and toolsets to cost effectively transform their operations into a data-driven digital business.
MultiTech will continue to provide its customers with the agility to utilise a combination of licensed and unlicensed wireless technologies and to deploy privately or use public networks, as well as the versatility to customise asset and network behaviour over the life of the application to meet the needs of the business, while understanding the real world trade-offs around capital and operating costs, security, speed of execution, and resiliency to backhaul outages and cyber security threats that differ for every application.