Digital Twins: A Dynamic Model of the World

Posted by Iotics on January 7, 2018

Creating a model of the world
emperorThe first Emperor of China, Qin Shi Huang, (pictured right) ordered an elaborate burial chamber constructed on a scale almost unimaginable today. Around 700,000 men spent almost four decades building life-size models of warriors, farmers, officials, carts, horses, roads, farms, towers and palaces, as well as mercury simulations of the Yangtze and Yellow Rivers. These flowed mechanically to a mercury sea under a vaulted ceiling resplendent with depictions of heavenly constellations. The Emperor believed that if his model of the world were accurate enough, he would be able to continue to rule the empire in the afterlife. Understanding and seeing all as it happened in the real world, he would have omnipotence over every corner of the empire and realise his claim on universal and eternal rule as the “First August Thearch” (or god-ruler). His dream was one of total and enduring control through a complete understanding of everyone and everything within the empire.

Many Emperors shared the same dream of immortality through the construction of an enduring dynamic model of the world that was interchangeable with the real world: a facsimile so precise that to have mastery of the model was to have dominion over the world.

World building today
Transporting these emperors to the present, they would surely see progress in their dream of modelling the world, but have we managed to achieve the ultimate objective of creating a true facsimile of the world as it is? Some might say that Big Data has made available so much data that we must now be able to know the world, but the aim of Big Data is primarily to mine history, not to reveal the now. Big Data doesn’t provide dynamic behavioral responses. The Emperors’ desire to model the world as it is happening is proving more intractable.

model of a building

The contemporary analogues of Emperors are decision-makers in business. They want to know what assets their companies possess – where assets may well be digital as well as physical, now that sources of data are of at least equal value as real things. They want to be able to see the state of things as they are so they can make decisions. They want a single source of truth. They want to run “what-if…?” scenarios to see which decisions make the biggest positive or the smallest negative impact. In short, they need a dynamic model of the real world. The contemporary analogues of the Emperor’s 700,000 army of builders and scores of overseers and bureaucrats are the data scientists, analysts and apps developers who must build these dynamic models.

Could “Digital Twins” be the solution to ancient Emperors’ and their modern-day counterparts’ dreams? Digital Twin is a buzz-phrase that is currently climbing the hype cycle, but what actually is a true Digital Twin? Many people apply the phrase to CAD/CAM 3D models of buildings, engines, components, etc., but these are static models of things that don’t have any dynamic data and don’t exhibit the behavior of the thing they claim to model. The model of the building doesn’t show how the building interacts with people waiting to use the lifts; the model of the engine doesn’t show how it works when some components are worn. These models don’t meaningfully interact with the world or each other.

Abstraction and the rise of Digital Twins
levels of abstractionPeople say they have a model of something in a spreadsheet – but how does new data get into the spreadsheet to affect the model? In 1982, Michael Jackson proposed a method of describing software systems based around the concept of abstracted entities, their actions and their interactions. At the time, his method was only applied to embedded real-time systems as there wasn’t the ever-present connectivity to allow these entities to interact globally. His prescient thinking predicted the true “Digital Twin” with its concepts of Level-0 (the real world), Level-1 (the abstracted entities) and Level-2 (the interactions between the abstracted entities).

At Iotics, we initially set out to make a solution where a virtual version of a thing could meaningfully and securely interact with a virtual version of another thing. We called it “double abstraction”. At that time, we hadn’t heard the term “Digital Twin”. However, that didn’t prevent us building an infrastructure that included Digital Twins as a foundational element: only we were calling them “virtual devices”.

computer screenAs our knowledge and experience grew, we realized that the Digital Twins people talked about were only a subset of the four kinds of twins that we could model in our digital twin environment, Iotic Space.

The four types of twins that can be modeled are:

  • Asset Twins
  • Data Twins
  • Action Twins
  • Composite Twins

Additionally, you can make admixtures of the different kinds, if that suits your problem domain.

Types of Digital Twins
We’ll explore these types of twins in order of roughly increasing complexity:

  • An Asset Twin is a twin of a thing in the real world that only has metadata, i.e. what the thing is, what it does and, perhaps, where it is. This is useful up to a point, but Qin Si Huang would not be impressed as it can’t tell him about the now – e.g. how hot, how full, how much power is it consuming, etc. We have some Asset Twins in Iotic Space, but we don’t generally like them because of that lack of dynamic data.
  • A Data Twin is an asset twin that produces at least one “feed”, i.e. an event stream of useful data. It has all the metadata associated with an Asset Twin but has the huge advantage in that it can tell you how hot, how full and so on. A Data Twin is the first true Digital Twin. Examples of Data Twins include: virtual weather stations with feeds of temperature, pressure and wind speed; car parks with feeds of available spaces, current utilisation, filling or emptying trends.
  • An Action Twin is like a Data Twin except that it offers a control interface to accept actuation requests and their associated data, then performs some action based on the request. Action Twin examples include: HVAC system’s thermostat settings, lighting levels, reporting cycles.
  • A Composite Twin is a model of a real thing that simulates its activity. A Composite Twin might receive data from some Data Twins, do some processing, refer to a local data store and then decide to publish some feed data of its own, actuate some Action Twins and update its local model. Examples of behavior twins are: Simulations, Algorithms, Orchestrators, Models and Predictive Analytics Tools.

Beyond depiction – driving value
Iotic Space was built with a greater aspiration than merely to provide the environment for the creation of digital twins. As mentioned earlier, the design brief was to allow Digital Twins to “meaningfully and securely interact”. A lofty goal, but what does it mean? Let’s deal with those objectives in reverse order:

sensor digital twinWhat use is a system of Digital Twins if they don’t interact and communicate with each other? We broker four kinds of interactions between the Twins:

  1. I publish a feed for others to follow
  2. I follow another’s feed
  3. I advertise a control for others to actuate
  4. I actuate another’s control

By “broker” we mean that any follow of a feed, or attach to a control, is negotiated. The first Twin asks a central broker “Can I follow Twin x’s feed?”. The broker asks Twin x if that is ok and, if so, the connection is established. The owner of Twin x sets up policies to allow or forbid connection requests from those she does or doesn’t trust respectively. She can alter these at any time. The same applies to controls. All interactions are via event-based message streams and all APIs are push.

There would be little point in having Digital Twins that were spoofable or open to DDoS or man-in-the-middle attacks. We, at Iotics, have built in security by design. The abstraction of a virtual Twin provides one level in that you can only attack the Twin, not the real thing. The policy-based brokered interaction provides another level. All messages are encrypted with TLS 1.2, and contain sequence numbers to mitigate replay attacks. Each Twin has its own token, used in a challenge/response authentication protocol with the infrastructure. If a Twin is compromised, its token can be revoked and then all traffic to/from that Twin will be rejected.

iotic spaceIn 2017 Iotics partnered with (ISC) 2, the global membership organisation that certifies information security professionals, which recognized the value of Digital Twins within Iotic Space to help provide security and confidence to enterprises of all scales. Together we wrote a course on how to secure IoT networks using the Iotics approach.

“With the rapid adoption of consumer devices, IoT and sensors, organisations face a significant challenge to understand and secure their networks. Iotic Space provides the ability to model and then secure these networks.” Adrian Davies, (ISC)2

The last objective is that the Twins should provide meaning. We talked earlier about metadata and that a Twin should be able to tell the world about itself and its capabilities in a meaningful way.

A technology was needed to do this in a way that is:

  1. Machine readable
  2. Human understandable
  3. Extensible

Iotics chose the semantic web technologies of RDF and SPARQL to do this important descriptive piece. The semantic web opens up the possibility of autonomous interoperability between the Digital Twins in a true Internet of Twins.

Have we finally enabled a dynamic, responsive, up-to-date, digital model of the real world? Would the Emperors recognise this as their dream of an enduring dynamic model of their empire that was interchangeable with the real one?

Are the needs of decision-makers in business going to be met? It might be too early to say just yet how far Digital Twins will go in helping us in our collective world modelling and building, but there are promising signs that this is a fruitful approach.

Big Data, Machine Learning, Analytics and AI solve complementary problems, and will contribute to the worlds we wish to model, both real and imagined, but something needs to deal with the now, the transitory, and the immediate in just such a complementary way.

Emperors believed that if they could truly know the world then they could control it and could rule it most effectively. Today’s leaders of industry may not want to achieve godhood, but Digital Twins could provide the immediate visibility and understanding not only of assets, but their data, their actions and their behaviours.

They could move us beyond CAD-rendered depictions of individual elements to dynamic interoperable parallels of their view of the world as it is now, today, this minute. These Digital Twins of entire worlds and data estates can deliver the insights that will transform services and solutions, streamline processes and support new behaviours. The objectives are massive, but how do we start to explore, abstract, model and interact?

An interconnected, brokered, event-driven, dynamic Iotic Twin environment might just be the thing.

iotic approach