Thinking beyond AI – quantum computing in transport

Quantum computers can process large amounts of data simultaneously, and solve certain problems much, much faster. But can they help with transport modelling?

The quantum machine Courtesy Erik Lucero, Martinis Group, University of California, Santa Barbara. CC BY-SA 3.0

The quantum machine Courtesy Erik Lucero, Martinis Group, University of California, Santa Barbara. CC BY-SA 3.0

In early May I attended a seminar on Quantum Computing in Transport, organised by techUK. Always on the lookout for new ways to improve transport modelling, I could not refuse the invitation.

For those of you not familiar with the concept, quantum computing operates on the same principles as quantum mechanics. Rather than bits that have a value of 0 or 1, a quantum computer works with qubits (quantum bits), which can represent both 0 and 1 simultaneously. And they need to be super-cooled, to around zero degrees Kelvin (minus 273 degrees Celsius), to eliminate thermal noise and vibrations. 

Join the UK's modelling community at Modelling World, held in Birmingham on 19 June 2024. Sign up here

I even held a so-called wafer, which is the quantum processor found in a quantum computer. It’s not really that different from a conventional CPU or GPU processor (but David Burnett from D-Wave was VERY keen to get it back after his talk – perhaps not surprising given that a quantum computer can cost tens of millions of dollars or more). The reality is that for most of us, buying a quantum computer won’t be necessary - we will purchase quantum computing in the cloud. Currently Quantum Computing as a Service (QCaaS) costs between $1,000 and $5,000 per hour of processing time.

Compared with traditional computers, even supercomputers consisting of thousands of CPUs or GPUs, quantum computers can process large amounts of data simultaneously, and solve certain problems much, much faster. Certain problems, I hear you say?

Here it gets interesting. According to all of the speakers at the seminar, quantum computing is particularly promising for complex optimisations. And that’s exactly what many of our transport modelling problems are, and what our solution methods try to solve.

There is already a commitment and a National Quantum Strategy developed by the Department for Science, Innovation and Technology. Stuart Dick, Head of Quantum Technologies at the Department for Transport suggested four areas of specific opportunity in the field of transport:

·       Computing

·       Sensing

·       Imaging

·       Communications

Particular applications could be, for example, asset management, navigation and enforcement; but I also see real opportunity for speeding up traditional transport models, be they real-time system optimisation or longer term forecasting.

One of the most pertinent audience questions was whether the current fascination with artificial intelligence overshadows the potential of quantum computing. Perhaps not surprisingly, the speakers saw the two developments as complementary. Quantum computing is apparently well-placed to carry out the vast calculations behind many AI applications, faster and better.

And don’t forget the potential carbon savings. Despite the energy-intense need for super-cooling, the calculation speed improvements that quantum computing offers, can also lead to significantly reduced power consumption. According to an article in Nature Computational Science by Sophia Chen, the world’s fastest supercomputer, Frontier, draws 8 megawatts, whereas a Google quantum computer only requires 26 kilowatts, less than half a percent, and may solve the problem much faster.

Afterwards I spoke to Charlie Burnard and  Andre Carvalho from Q-CTRL. I asked them what should surely concern us as transport modellers: for quantum computing to be effective, can we just transfer our existing algorithms to a new computing environment, or do they need redeveloping into quantum algorithms– that will obviously be an obstacle to rapid implementation and practical acceptance.

What they told me is that a good description of the problem that needs solving is crucial – nothing new there. But it’s not advice that we always heed even in current transport modelling, using existing models for purposes for which they were not developed: if all you have is a hammer, everything looks like a nail.

A typical application often starts with a month’s exploratory period in which the quantum engineers sit down with the domain experts to really understand what the question is (and yes, there was a reference to the Hitchhiker’s Guide to the Galaxy), and which elements of the traditional solution algorithms can be reused, or need rewriting to make best use of qubits rather than bits.

Ashley Kerly from Fujitsi told me over a coffee that some of the transport applications to which they are applying their quantum technology include scheduling (of trains and buses), MaaS and traffic signal optimisation. I can see the attraction and would love to see a real-life evaluation of results vs our best current alternative approaches.

Which brings me to Modelling World. The programme for this year’s conference is now set. But you can be assured that in 2025 we will get a session together on quantum computing: it looks too promising not to spend a few hours listening to what the rest of the transport profession is doing. And who knows, what is already being explored by Modelling World delegates and their employers? I would be very pleased to hear from early quantum adopters. In the meantime, I will keep looking for relevant examples.

Tom van Vuren is Strategic Consulting Partner at Amey, Board Director at the Transport Planning Society and Visiting Professor at the Institute for Transport Studies at the University of Leeds