Speaking Engagements & Private Workshops - Get Dean Bubley to present or chair your event

Need an experienced, provocative & influential telecoms keynote speaker, moderator/chair or workshop facilitator?
To discuss Dean Bubley's appearance at a specific event, contact information AT disruptive-analysis DOT com

Monday, December 05, 2016

Industry 4.0 & IIoT needs shared spectrum to succeed, either for 5G or WiFi

[Warning, Long Post! 
TL;DR summary: Industrial automation increasingly needs good wireless connectivity for robots, sensors and other systems. Mobile operators and telcos are poorly-suited to providing it, especially in-building or site-wide. Regulators need to focus on more shared spectrum, to allow other stakeholders to build/own/operate IoT wireless networks, either for 5G, WiFi or other technologies]


A topic I've been hearing repeatedly about recently is Industry 4.0. This is the reinvention of many sectors with IoT, analytics/AI, sensors and assorted other technologies. Sometimes called the IIoT (Industrial IoT), the 4th Industrial Revolution, or Industrie 4.0 (for Francophones), it follows on from mechanisation, mass production and computing. 

You'll often hear the clunky term "cyber-physical systems", or perhaps derivatives like Smart Factories, Smart Agriculture, Smart Utilities and so on. You'll also often hear Industry 4.0 linked with a discussion of 5G, and perhaps governments' broadband or spectrum policies. (Obviously my esteemed readers wouldn't use the useless & obsolete word "digital").

But what you won't hear much about is Industry 4.0's dependency on what *type* of spectrum, and by association, what types of networks are deployed, and by whom.

Credit: Christoph Roser at AllAboutLean.com."

My view is that for Industry 4.0 to succeed, there need to be privately-owned and -run wireless networks for IIoT. Traditional mobile operators cannot provide all the connectivity requirements, and industry will not pay for all wireless connectivity "as a service", especially inside buildings or across their plant/campus facilities. (Just as fixed telecom operators do not provide most in-building ethernet or other wired systems today).

This needs to be reflected in spectrum policy - and, especially, various models of spectrum-sharing. Strategic over-reach by MNOs, or bodies like the GSMA, will otherwise risk delaying the adoption of Industry 4.0 altogether, or cause enterprises to seek sub-optimal approaches to implementing IIoT, with consequent damage to the broader economy. 

It doesn't matter whether specific IIoT implementations use 4G, 5G, WiFi, short-range or LPWAN technologies like SigFox or LoRa, there needs to be the option to have (locally) self-controlled spectrum for owned networks for critical or secure IoT.  

As I've written before, Telecoms is too Important to Leave Up to Telcos (link). 

It's also sometimes too important to leave to today's WiFi, LPWAN or ZigBee. Enterprises in the IIoT era cannot not just purely rely on unlicensed wireless for their future, mission-critical connected systems.

I want to drill into this more, as it's critical that the cellular industry - and policymakers - do not over-reach, or ignore secondary/tertiary issues, as has often been the case in the past. My recent experiences with the 5G / mobile industry discussing "verticals" suggests that it is doing so again. 

There is also a "spectrum land-grab" going on, with the mobile industry lobbying governments to suggest regimes that favour traditional MNOs, rather than domain specialists and innovators. This needs a bright light shined upon it, as it is often duplicitous and ignores inconvenient facts about particular sectors, different models, and competing constituencies.


Smart Industries & Factories: thinking realistically about the connectivity needs 

The 5G-PPP group, for example, has written a very good document on the rise of the “Factories of the Future” (link), detailing numerous ways technology will enhance manufacturing and related industries. It talks persuasively not just about realtime connectivity for automation (robots, production systems etc), but also the growth in data collection and analytics, improved logistics and energy-management and so forth.

The report also covers the way manufacturing companies are themselves becoming services companies, with various in-field business models related to their products (eg airline engines provided as “power by the hour”). It’s a fascinating summary of the changes that lie ahead for industry, and the on-site communications needs that extra use of new integrated IT systems will bring. It highlights the need for low-latency, high-speed, super-reliable wireless connectivity – in some cases with requirements as stringent as 1ms cycle times, jitter below 1µs and other parameters like timing sync in the nanosecond range.

But while the 5G-PPP report – and assorted other documents like it – suggest that 5G might be able to deliver the various QoS characteristics, it doesn’t make a case for why that connectivity should be provided onsite as a service by an MNO, in that operator’s licensed spectrum, and paid for out of OPEX on an ongoing basis. 

Indeed, it points out that many other network technologies will coexist and need to integrate with each other – Industrial Ethernet, WiFi, ZigBee, LPWAN and various industry-specific standards such as WirelessHart and ISA100.11a [no, I hadn’t heard of them either]. Most mobile operators are ill-suited to perform that integration, as most of those networks are owned, not service-based.

A similar set of critical-wireless issues could apply in many different industrial sectors:
  • “Closed loop” industrial systems in a factory or plant
  • Equipment in hospitals, from care-robots to portable bedside sensors to surgical systems in operator theatres
  • Trains and operational systems on railway networks
  • Automated vehicles at an airport (eg baggage handling systems)
  • Construction machinery and sensors on a building site
  • IoT systems in remote regions not covered by normal cellular networks (eg mines, offshore energy facilities)
  • High-voltage grid sensing and operations by utility companies
  • Potentially, various use-cases for smart cities, and smart farms, such as automated transport systems or agricultural machinery

In a nutshell, there are going to be many instances where the IoT systems will demand extremely high wireless network performance, reliability and availability on a particular company’s site. That connectivity system will need to comply with that industry and location’s specific regulatory requirements (eg safety), and integrate with both the other network types and the IT/industrial systems in use.

In terms of business models, while some aspects of IT are outsourced (eg to cloud services), industrial companies do not want to obtain every capability "as a service", with ongoing payments. In many cases, they may want owned assets, for security, IT integration or control/regulatory reasons - or even because their cost of capital is even lower than telcos'. Where they do want to acquire IoT-related managed services, they are likely to want a central provider, rather than obtaining connectivity through a different contract.

It is naïve to imagine that AT&T or Vodafone is going to be providing “robot anti-collision-as-a-service”, via an edge-computing node built into a 5G small cell, running in the MNO’s 42GHz mmWave spectrum with the MNO’s SIMs/eSIMs in the robots. Neither are patients going to be too happy being operated on by a surgical robot, which is sending pictures to the doctor via an IMS core. (Does RCS messaging have a spleen emoji, I wonder?).

It is also naïve to imagine that safety-critical systems will be trusted to WiFi running in totally unlicensed and uncontrollable spectrum.

In other words, closed-loop industrial systems are unlikely to benefit from a telecom operator’s involvement given that it is, quite literally, out of the loop. Even if it is possible to give control of a “network slice” to the company, how far does that extend in reality? Is a manufacturer going to be happy just being a glorified MVNO, without “administrator” access all the way to the radio? If a mining company wants to completely shut down a RAN temporarily (eg while detonating explosives), how does it do that? 

In other words, there needs to be a good way for large industrial facilities to operate wireless networks, in some form of licensed spectrum, without the involvement of MNOs.
 

The regulatory answer: spectrum-sharing. For cellular or other technologies

There are three broad types of spectrum management:
  • Dedicated licensed - the familar style of license that is used for mobile networks today, where specific operators (MNOs) gain rights, usually through auctions, to particular frequencies exclusively. (This also applies to broadcasters, satellite operators, government bodies etc). This allows full management and therefore guarantees of QoS.
  • Unlicensed / license-exempt - used for WiFi and other "ISM" (industrial, scientific, medical) applications. No specific user license is needed, which lowers costs but also risks interference and congestion.
  • Shared spectrum, or Dynamic spectrum access - multiple users get access to a given band, but it is not a "free for all". There are some mechanisms to ensure separation, fairness, pre-emption, manageability and so forth - but not nationwide exclusivity.
In my view, the last option needs to be utilised much more, if Industry 4.0 has a chance of success. We need to avoid strategic over-reach by the cellular sector, which is lobbying hard for more dedicated spectrum in an (understandable but doomed) approach to insert MNOs into many new IIoT value chains. It also hopes to disenfranchise other stakeholders in what is likely to be a "land grab" in coming years - there are various new bands like 600MHz, 3.5GHz and whole swathes above 6GHz that are being lined up for "harmonisation" in 2019 at the next World Radio Congress.

At this point it is worth noting that there are various approaches to using shared spectrum emerging. We have had "TV white spaces" for a while, although it has had relatively little use. Some legacy wireless networks such as PMR (public mobile radio) have used shared frequencies. Fixed-wireless links for point-to-point connections, are commonplace in industry already. Scientific uses (eg radioastronomy) coexist with other applications. 

Looking forward, there are various technical approaches such as listen-before-talk, or geographic (or other) databases that allow geo-fencing of spectrum rights. There can also be forms of spectrum sub-leasing, with the approval of regulators - or occasionally just illegal or corrupt use of someone else's bands, in remote areas.


There are also differences between sharing existing "primary" users' bands, or dividing up new bands on a basis other than dedicated national licenses. For instance, the former could include exploiting military or coastguard frequencies in peacetime, or inland. The latter might use a registry or database, and employ a spectrum manager function to optimise fees for the government, or perhaps look to maximise benefit to the wider economy.

This post isn't the right place to discuss all the differences between models such as LSA (licensed shared access) mostly examined in Europe, or the more-ambitious SAS (spectrum access system) discussed in the US.

There are plenty of analyses and white papers around, such as from Intel (link), GSMA/Deloitte (link), Plum Consulting (link) and many more. They all vary in their in-built biases, and do not always think through the typical structures and business-models of verticals that might need connectivity in future. I'll delve more into the options another time.

What is critical is that whatever system is used makes it easier for new wireless stakeholders and innovators to emerge, especially around IIoT applications for in-building and on-campus networks. While there are some opportunities for today's MNOs, there are many other plausible use-cases for 5G especially, that can only be viable if others are involved.

Network equipment vendors need to play a careful and nuanced game here - they obviously don't want to alienate their traditional MNO customers, but they also need to build their direct-to-enterprise solutions and sales, to rescue flattening revenue lines.


What companies would deploy networks in shared spectrum?

So if traditional telcos & MNOs aren't always the right organisations to provide onsite IIoT / Industry 4.0 connectivity, then who is?

This is a complex and sector-specific question, and will likely vary over time, and by country.  

Some industrial companies will be big enough and skilled enough to build and run their own internal wireless networks, just as they run their internal fixed networks. They can hire 5G or advanced WiFi engineers as easily as telcos. A car manufacturing plant might deem it strategically important to run its own network. So might a major airport. They would ideally have a geographically-specific shared spectrum license.

It could also be the manufacturers or integrators of IIoT systems - perhaps GE, IBM, ABB, Siemens, Schlumberger, Honeywell, Philips or similar firms. As they operate in multiple locations, they may have to deal with multiple spectrum regimes and approaches, depending on local rules. A power station in Japan might have a different network ownership model to a chemical factory in Canada, or a hospital complex in Souh Africa. There may also be different roles for private (or MNO/MVNO) use of 4G/5G, WiFi running in reliable spectrum, or LPWAN technologies.

Then it is likely that we will see the emergence of vertical-specialist services companies. They could be airport-management firms, oilfield-services providers, or simply dedicated IoT and industrial connectivity players. Some will evolve from today's niche MVNOs and platform providers. Perhaps Cisco Jasper could be one, network vendors' managed-services arms, or startups and spin-offs from existing industrial suppliers. Further, many of the previous set of IoT manufacturers are moving to become forms of service providers themselves, as the 5G-PPP paper discusses, for which wireless connectivity could be an embedded enabler, in the same fashion as electricity from embedded generators.

Lastly, we will probably see WISPs, fixed-broadband and fixed/cable operators move into this space - especially where they have existing big enterprise customer base and integration/management skillsets and resources.

All of these groups could benefit from easier availability of "controllable" spectrum, beyond the current limited scope for fixed point-to-point links. In most cases, I'd expect geographically-limited rights to be the best approach. This would likely need some efficient mechanism for creating the issuance, registration and resale of such sharing rights. Perhaps a blockchain-based registry (and perhaps even smart spectrum contracts) may emerge as a suitable underlying platform, although the timelines might not be aligned perfectly here.

Companies that cover a very broad area for IIoT, and have both fixed assets and mobile users will have a very different set of requirements, and may turn out to need dedicated spectrum to themselves, or operate in partnership with MNOs somehow. Railways, utility companies and public safety agencies are examples. For instance, control systems for the high-voltage electricity grid cannot rely on MNO connectivity - as the MNOs' cell sites rely on the utility firm's power. If there's a power outage, you risk ending up with a chicken/egg dilemma, unless there's robust diesel backup at every cell site

Again, however, there will be a need to look at spectrum policy, in the light of the needs - technical, operational AND commercial - of different end-use sectors.


Conclusion & Recommendations for Regulators

Regulators need to continue with a push for shared spectrum models, especially around 5G. While MNOs definitely need a lot more exclusive spectrum for mobile broadbrand, it is becoming much less-clear that IoT and IIoT use-cases are optimal for the traditional services/subscription model preferred by the mobile industry. There are separate arguments for allocations of shared spectrum below 1GHz, between 1-6GHz and for 6GHz and above.

WiFi has demonstrated the huge value to economy and society of unlicensed wireless. It has unleashed immense consumer welfare and innovation in usage models, including in industry. Critically, it can be offered as a service, owned outright, provided as an amenity, sponsored, bundled, integrated and numerous other commercial forms. 

WiFi's "generativity" needs to be enhanced with "controllability" for mission-critical IoT, and also perhaps ultra low-cost IoT for wide areas.

This suggests that shared spectrum should be made available, with the expectation that it will be used for private 5G (or 4G) networks, to cover use-cases and business models that MNOs are poorly-equipped to address. Governments and regulators need to be wary of representations from the cellular industry that push a "monoculture" of national, subscription-based connectivity. If they go that route, they need to think about coverage rules applying to the inside of enterprise facilities, not just broad national statistics.

Supra-national bodies like the ITU and the various European regulatory bodies such as the EU Commission, CEPT and BEREC also need to grasp the limits of the incumbent cellular model in enabling Industry 4.0. It has its role, but so do other paths and options. However, the industry "noise" from trade associations, academia and elsewhere is overwhelming in its narrow view of the traditional MNO model. The coming of Industry 4.0 demands a reset - not a fallback to historical approaches to wireless.

It's also worth pondering if the concept of running LTE in unlicensed spectrum mostly used by WiFi (MuLTEFire, LAA, LTE-U etc) should be considered in reverse. Is it possible to use WiFi in mostly-empty, dedicated frequency bands notionally designated for 3G, 4G, 5G? Would that need change to ITU's regulatory definitions of IMT (International Mobile Telecommunications)? Future wireless won't always be international, moving about, or even "telecoms" in the traditional sense, anyway, so it is debatable whether all 5G should really be deemed IMT-2020 by ITU and national regulators. 

Such a move would be hugely controversial, but since lots of studies show that LTE and WiFi can coexist happily, it should be technically feasible in notional licensed bands as well as unlicensed. There are already pilot examples of WiFi running in licensed spectrum - bodies such as WiFi Alliance and IEEE 802 should consider it more fully.

The bottom line is this: if industry and society is to benefit from IIOT, automation, smart systems using AI, and a whole host of other innovations, it needs reliable wireless connectivity that can be owned outright, not just paid for on a service basis. The best way to deliver this is via shared-spectrum models that allow new entrants to build private, or system-integrated wireless networks for Industry 4.0.


Dean Bubley is an analyst and futurist for the telecoms industry. His company, Disruptive Analysis, advises vendors, operators, regulators and investors on technology trends, business models and policy. This spans network innovation, communications apps/services, and advanced technologies. Please get in touch if you are looking for consulting or public-speaking assistance - information AT disruptive-analysis DOT com.

2 comments:

James said...

This is a well constructed and highly relevant piece of work - if we are to reap the benefits of new 5G technologies we are going to have to open up ways of running infrastructure well beyond the existing MNO monopolies that currently dominate the usable market (and spectrum).

Our Governmental aim should to be to get EVERY piece of spectrum in EVERY location used by somebody ALL of the time - this currently is NOT the case, where certain operators (e.g. BT) make very little use of large blocks of spectrum licensed exclusively from Government. Consequently I am a big fan of Licenced Shared Access (LSA), where users can dynamically bid for spectrum leases in real time.

The concept of assigning huge swathes of spectrum on a national basis for indefinite periods of time on the basis of a single monolithic auction is fatally flawed and is hugely inefficient. Let us focus on ways of getting better efficiency out of our valuable spectrum - whilst keeping the door open for new and innovative radio uses. I remain committed to the concept of LICENCED spectrum through which we can provide mission critical systems with reliable, predictable and interference free radio access - but this does NOT mean just awarding all the spectrum to the operators with the largest budgets!

The next twelve months are going to be interesting!

Hamish said...

Such a move would be hugely controversial, but since lots of studies show that LTE and WiFi can coexist happily, it should be technically feasible in notional licensed bands as well as unlicensed.

Precisely. While some were concerned at the news:

"Qualcomm said it was pleased with the FCC’s move as well as a letter from the Wi-Fi Alliance that indicated the group did not oppose Qualcomm’s application. “We are collaborating with the Wi-Fi Alliance to develop a coexistence test plan, and we anticipate using that plan for joint lab and field tests to validate that LTE-U will not have any adverse impact on Wi-Fi. We would like to thank the FCC and the Wi-Fi Alliance for working with us to reach this important result..”

http://recode.net/2016/01/31/fcc-allows-further-testing-of-lte-service-in-unlicensed-spectrum-used-by-wi-fi/


I can't understand anyone opposing Verizon et al dispelling the myth
of interference as David Reed has been attempting to do all these
years, http://www.salon.com/2003/03/12/spectrum/

If cellcos can prove their services work without exclusive access to
spectrum, and within Part 15, surely then Google et al can begin the
lobbying for the opening up of the cellcos spectrum to Wi-Fi.

It doesn't seem likely that "LTE-U will not have any adverse impact on
Wi-Fi" would be sufficient without the reciprocal.

Sounds like a win.

But, tbh, it's likely the cellular beneficiaries have some
excuse^H^H^H^H^H^H reason LTE-U cannot be LTE-universal...