In transport nothing can be said to be certain, except delay and disruption. Be it road, rail, or air, it’s a problem our sector has never mastered. This is perhaps because the causes of disruption on any significant network are numerous, stochastic, and inter-connected. While our existing means of responding are heuristic, linear, and often over-fitted to past issues.
Four challenges have scuppered the development of game-changing operational digital twins for transport (e.g. traffic management and incident response decision support tools) to date, namely:
- A lack of sufficient and granular data,
- A reliance on simplistic hierarchical data models,
- Classical computing’s limitations in running and optimising statistical models,
- Challenges 1–3 in turn preventing us from taking advantage of sophisticated ML/AI.
Soon, we will have the tools at hand to meet each of these challenges:
- Rich, ubiquitous, real-time data from IoT,
- Complex multi-dimensional modelling using ontologies,
- Quantum computing!
- And AI that can interpret the outputs of our quantum stimulations and make them meaningful to normal humans.
When transport Digital Twins truly realise their potential, they will have Quantum Computers under the hood running optimisation algorithms of unprecedented depth... Quantum Digital Twins!
So how could it work? Well I’m not a physicist, but here’s my simplistic understanding… Quantum computing promises enormous advantages over classical computing, but only for distinct types of problems. It comes into its own in representing probabilistic processes. Simplistically, this is because where classical bits can only assume a binary state, quantum bits (qubits) can occupy multiple states simultaneously through a phenomenon known as superposition. Qubits can in turn interact with other qubits; probabilistic states (quantum parallelism). As a result, the number of states a quantum computer can represent increases exponentially with its number of qubits, allowing the modelling of systems with a vast range of possible states. It also means that quantum computers can scan through enormous quantities of information far faster than classical databases (the algorithm to do this was invented by Grover from Sesame Street, pictured below). As Andris Ambainis writes in his beautifully clear explainer, “we know that [quantum computers] will be faster for many computational tasks, from modelling nature to searching large amounts of data.”
How could we use the unique capabilities of quantum computing to build better Digital Twins of our transport networks? It’s fair to say that at this early stage not every use case is apparent. That said, the application of quantum techniques to anticipating and responding to delay feels obvious. This is because a big part of quickly identifying and responding to disruption comes down to:
a) scanning through large datasets for indicators, and
b) modelling complex inter-related scenarios.
This could mean processing terabytes of sensor data to look for the early indicators of disruption. We could then model the state of the network at that time, along with the potential potential states of traffic, weather, and (crucially) operational decisions as qubits. This would give us a range of possible outcomes with which to assess the impact of any reasonable course of action. Per McKinsey, "Today’s computers can handle only one set of inputs and one calculation at a time. Working with a certain number of qubits... a quantum computer can conduct calculations on up to 2^n inputs at once." The opportunity exists to model the logistics of identifying, assessing, and responding to transport delays with a level of verisimilitude that is impossible today.
I think this is why Quantum Computing will be a game changer for transport models and Digital Twins? Because it will finally allow us to rapidly run simulations that adequately express the complexity of our transport networks, and their interconnectivity with the built environment, national infrastructure, and external world. This will allow us to create Digital Twins that don’t just reflect reality, but which allow us to anticipate the consequences of our decisions. Quantum Digital Twins will allow transport operators to better understand how to interact with our networks, moving us away from heuristics, and allowing us to act to reduce both the likelihood and impact of disruption. As Flourishing Systems puts it, the aim is “not just to mirror, but also to help manage and get more from the system
of systems.” This is not Minority Report, we are not predicting the future (even quantum computing doesn’t let us do that), instead we are better anticipating the imminent causality of the present.
Of the financial sector’s early forays into quantum computing the Economist writes, "many financial calculations boil down to optimisation problems, a known strength of quantum computers." The same is true of many of the most tenacious problems associated with managing disurpted flows of traffic, be it the classic 'travelling salesman’ problem, or dynamic real-time routing of fleets of connected autonomous vehicles.
While there is undoubtedly a lot of hype around quantum computing, and real applications in transport are a few years away, industry heavy hitters are already finding uses. Volkswagen and DWave claim to have used quantum computing for route optimisation of bus fleets, and that they “designed the system so that it can generally be applied to any city and to vehicle fleets of any size.” DHL referenced Volkswagen’s pilot as evidence of quantum computing’s potential to deliver “adaptive re-planning and reallocation of assets in the event of unexpected shutdowns, late shipments or order cancellations,” presumably a boon to the entire logistics sector. Microsoft and Ford have piloted similar work- and claimed a 73% improvement in congestion- albeit using simulated quantum computing rather than the real deal.
The promotional nature of these pilots aside, the theme of route optimisation points to a wider industry need. Most network operators have control rooms where decisions are made in the heat of the moment. A points failure outside Victoria in the morning peak, a fire on an HGV in a tunnel on the M25, ash from a volcano drifting across Atlantic flight paths. Re-routing fleet, cancelling services, closing lanes, grounding planes, changing speed limits. The staff working in these centres are expected to react to an incredible variety of scenarios using a few standard operating procedures and some rudimentary visualisation software. However, there is rarely much time for quantitative feedback, and little opportunity to 'A/B' test different responses. The current state of the art isn’t so much 'optimisation' as it is 'damage control.’ What often remains unknown is the long tail of consequences of operational decisions, the outcomes of the paths not taken.
In my time at Network Rail it was clear that actions taken early in the day a key locations often had repercussions throughout the day, and on to other networks. That a network could be more, or less, ready to deal with disruption, meaning that a given event (for example slow boarding of a service) could have no impact, minor impact, or signficant self-reinforcing impact. And yet we struggled really quantify the consequences, identify the route causes, or meaningfully compare them against any counterfactuals. This was partly because of an ability to process the volume of data in question in a timely manner, but it was more a lack of the sophisticated models and computation required to process this information in near-real-time.
Quantum computing and Quantum Digital Twins will force us to think differently about how we operate our infrastructure. We will stop conceptualising the operation of our network as adhering to simple cause-and-effect, but rather to frame our actions as an informed bet on a probabilistic system that follows rules too complex to be fully 'knowable'. As McKinsey write, "Quantum computers won’t replace today’s systems. Instead, quantum computers will be used for distinct kinds of problems, incredibly complex ones in which eliminating an enormous range of possibilities will save an enormous amount of time." For any given disruption of a transport network there are a huge possible courses of action that operators can take. What we need our Quantum Digital Twins to do is to say “alright mate, 60% chance that if you do these three things, everything’s going to be OK, and failing that at least the long-tail of outcomes is manageable.”
Returning to the three challenges at the start of the article, what is truly exciting about the possibility of Quantum Digital Twins is that the technology is not emerging in isolation, but in parallel with the maturation of IoT, graph databases, and deep neural networks. What these technologies all have in context is that they force us to think of 'states' rather than linear causality. IoT sensors can describe a complex combination of states that our infrastructure passes through in its operation, where often the cause of a failure isn't associated with an asset being in a particular state, but rather in it having passed through a certain sequence of states. Similarly, graph databases and neural networks ask us to drop our simplistic 'hierarchical' structuring of data in favour of a far more nuanced and flexible graph of interconnected entities and relationships. These parallels may sound overly poetic, but in truth they point to the step change in conceptualising and modelling our physical world that the Digital Twin movement demands.
Quantum Digital Twins could be the secret to realising the vast potential of the Digital Twin movement. After all, what is a Digital Twin if not super symmetry.
