Your browser is not supported. Please update it.

12 October 2020

Beyond low latency – the unsung industrial use cases

The rise of ultra-low latency up the 5G agenda has almost obscured the importance of other applications that are not so delay-dependent, especially around the industrial IoT (IIoT). It is true that ultra-low latency is crucial for real time process control, particularly robotics and UAVs (unattended automated vehicles) that are dependent on very fast interactive wireless communications. But these are just part of a constellation of applications under the IIoT banner that will be served by 5G infrastructures.

Furthermore, the ultra-low latency use applications will have to dovetail with other more delay tolerant higher level processes for optimization of supply chains and distributed manufacturing operations, which cannot be practically contained under one roof or even campus.

We have been reminded of this by several recent announcements, including launch of a factory IoT platform by Japanese automotive components maker Denso, formerly part of Toyota. Whether or not the company is correct to claim this to be the automotive industry’s first cloud-native platform developed using open source software is hard to verify, but either way the project has some points of interest.

First implemented in October 2019, the platform now interconnects 130 production factories under the concept “as if under one roof”, which we will expand on in a few paragraphs time.

Being widely distributed, ultra-low latency is not achievable across this factory group, but is fortunately not necessary for the processes involved, which include analysis of workers’ movements and operation status of different facilities in near but not absolute real time. It then enables factories to respond quickly enough to production changes according to variations in local demand, which must interact with ultra-low latency processes inside each plant. Underlying this is a cloud-based distributed database fed by data from equipment and monitoring systems at the various production sites.

A key design goal was the ability for software engineers to update the system continually in the field in the light of experience, almost in a perpetual state of testing, which relied on agile methods under the heading of DevOps. This is where software is assembled in components as small as feasible, which makes subsequent testing and upgrading less costly and complex because changes can be confined to single or just a few objects.

As the name suggests, development and operations were brought together. They had traditionally often kept separate in the context of software or system development to the extent they were almost ghettoes, which impeded testing, subsequent maintenance and upgrading when the two sides had to come together. Under DevOps, software components are crafted and tested under smaller teams comprising both operations and development personnel. Eventually they are integrated and subject to wider systems testing, which again is repeated as changes are made, but with the ability to zoom in on those components that have been altered, with less risk of wider dependencies being disrupted.

Denso claims to have gone further still by involving non-IT employees in development by contributing to specification and ongoing improvement of the IoT platform, in conjunction with machine learning algorithms to optimize processes after changes have been made.  Employees can also contribute unstructured data in the form of feedback or observations, which is fed back via an AI related algorithm into the system with the goal of improving operational practices, although this part of the platform is work in progress. An ultimate goal is then to improve the human/machine interface.

That concept of “as under one roof” is significant because it can be seen as an admission that an earlier drive towards fully integrated manufacturing or fabrication was misplaced, certainly for larger operations involving complex and diverse supply chains.

The objective was to cut transportation costs, streamline Just In Time operation and achieve scale economies by having all steps from raw materials processing to finished goods if not under one roof, then at least across a single campus. This was promoted particularly for some emerging industries such as solar panel production which involved various stages culminating in the finished panels or modules. Yet despite emergence of dedicated suppliers of the integrated turnkey production line equipment, adoption was low.

There were various reasons for this, one being that the logistical challenge of ensuring all raw materials required for the different stages were delivered on time to a given location proved insurmountable for many industries.

It actually reduced rather than increased reliability. As a result, the IIoT will evolve around local production plants within each of which ultra-low latency processes are employed, interconnected within a larger umbrella embracing the wider supply chain or ecosystem. Parts of this wider distributed infrastructure, as in Denso’s case, will be very much “as if under one roof”, all under the control of a single company, but there will also be interaction among distinct suppliers and participants in the chain.