Digital Twin Computing

Digital twins are not just about data, they are about behaviour and so much more.

Educational 21 Oct 2021 by Mark Wharton

Kindred spirits

I recently wrote an Insight “Digital Twins – Behave!” where I set out how digital twins should be considered as more than just data twins, they also exhibit behaviour. Well, it seems that I’m not alone. I attended the ISO IET digital twin workshop recently and part of their background reading was this paper from NTT, where they set out their vision for Digital Twin Computing (my emphasis).

I’m a kinesthetic learner: I learn by doing. At IOTICS, we’ve always adopted a “just start” approach, but it’s still nice to see when people of a more logical learning style agree with the conclusions you’ve made via a different path – that their theory matches your practice.

An example of this convergence of opinion is near the beginning of their whitepaper. NTT reviewed the current landscape of digital twins:

“Digital twins enable analysis of current states, future prediction and simulations of possibilities of objects in cyberspace. Based on the computation results and leveraging the versatility of ICT in cyberspace, digital twins also enables feedback to the real-world to enable intelligent control of real world objects…”

Look at all the words in that sentence that imply some kind of processing – computing, if you like: analysis, prediction, simulation, computation, intelligence… So we agree on the twin behaviour bit, but then they make another leap:

“The concept of “Digital Twin Computing” (DTC) we are aiming for is a significant advancement on the existing concept of digital twins – it is a new computing paradigm that will enable previously impossible new, large-scale high-precision, real-world reproductions by performing various operations to freely combine various digital twins and will enable new interactions in cyberspace … which will go beyond physical reproductions of the real world.”

This is visionary stuff. “We are aiming for” … “a significant advancement” … “A new computing paradigm”… That will “…go beyond physical reproductions of the real world.”. This is motherhood and apple pie to us at IOTICS – it’s why we started the company. NTT also say “…to freely combine various digital twins and will enable new interactions in cyberspace”. This is exactly what IOTICSpace was built to enable and is delivering right now.

I recently attended an online conference on the “Cyber Physical Fabric”. Mostly, this was very encouraging. People were talking about bridging the gap between the cyber and physical worlds and then implying a “fabric” (probably a decentralised semantic knowledge graph) to “weave” the whole thing together. But… where’s the compute in the fabric? The conference was missing an important trick.

Silos and silos

I’ve long thought that when people talk of data silos, they’re not considering the other side of the coin: these silos also have parallel, systemic, functionality silos. PLM systems have PLM data, but they also have PLM functionality. If you make a digital twin of a product in a PLM system you’re implying that it’s only a “reproduction” (NTT’s word) of the physical thing, but what about the “PLM functions” that apply to the twin? If you leave those functions in the PLM system, you’ve only solved one side of the problem and you haven’t broken down the silo(s) fully.

The incompleteness of the solution that only deals with data and does not include functionality is why, I believe, people are finding it difficult to dissolve the barriers between “operational” data and “analytical” data. There should be only data, with operational functions and analytical functions that operate on that data. Digital twins provide the accessible presentation of that data; the functional part is in what IOTICS calls “data interactions” and NTT call “digital twin operations”.

“Digital twin operations will offer standardised methods for large-scale and complex interactions between digital twins to enable the creation of virtual societies by freely combining various digital twins and using them for analysis and prediction”

Going further

Virtual societies of twins with complex interactions and, as NTT explicitly state, twins of humans as well as things. Now that’s a concept! But it’s not as far-fetched, science-fiction-y as it might appear at first glance. In fact at IOTICS this is very much the here and now.  IOTICS uses semantic web technologies such as RDF and SPARQL to link twins semantically. With RDF, a society is just a semantic link away: twins in a city, twins in a hospital, twins in transport, construction, and resource optimisation. Combine architecture, semantics and accessibility rules, and twins can be part of more than one society: a doctor is part of a university, a hospital and part of a city. We made digital twins of people living with dementia as long ago as 2014.

In their whitepaper NTT have, perhaps surprisingly, restricted the scope of their digital twins “using them for analysis and prediction”. I see this as an artificial limit to the application of digitals twins: the uses of Digital Twin Computing should include orchestration and operations as well.

Where design and build meet

Finally NTT propose these layers of architecture. Notice the separation of concerns. Digital twins are separate from the cyber/physical interaction layer. Digital world presentation comes from semantic overlays. Applications talk to the twins. Again, this thinking is congruent with the architecture of IOTICS, but we’ve always seen it as more circular, where the twins of the applications themselves become part of the digital twin layer and any twin can contribute back into the ecosystem.

Digital Twin Computing Architecture

Proposed architecture for Digital Twin Computing as envisaged by NTT.

Overall, chapeau! To the writers of the white paper. Theirs is a radical manifesto. Gradually replace your old systems with asset-focussed twins with rich metadata and data. Use the operations, the behaviour, the compute of the twins to model your world. The “applications” are there to visualise the things about the model that are interesting, to analyse the model and to interact with the model to orchestrate the operations you want. Then the cyber-physical interaction layer will keep the model up-to-date with the real world and, crucially, the converse: keep the real world up to date with the model.

There’s beautiful simplicity in this design and it gratifyingly mirrors what IOTICS have built.


Join Our Community

We enable the world’s data to interact safely and securely with other data, of all types, in all places, dynamically.