How quantum computers are transforming travel
Image credit: Volkswagen
While still in an early stage of development quantum computers are already making an impact in the automotive and aerospace sectors.
As cars become increasingly connected and aircraft ever more sensor-laden, automotive and aerospace companies are making a transition from being ‘nuts and bolts’ manufacturers to so-called ‘mobility companies’ that collect and use unprecedented amounts of data.
Data, it is believed, will provide new insights to help shape the future of transport, via futuristic business models and technologies that will mitigate the many challenges presented by growing populations and increasing congestion amid the need for decarbonisation.
All information, however, requires processing at super-fast speeds if it’s to be useful. To do this, companies are exploring not only AI and machine learning but also the more experimental realm of quantum computing, to see if it can extract better solutions faster than the classical equivalent.
Leading the charge is Volkswagen AG. In November last year, the company announced it had, for the first time, used a quantum computer to develop a traffic management system that, it claims, “will replace forecasts of urban traffic volumes, transport demand and travel times by precise calculations”.
According to the company’s chief information officer Martin Hofmann, this means taxis and buses won’t have to wait for passengers or drive considerable distances empty, and public transport operators can add additional trips to their fixed timetables in line with demand.
The technology could help the company achieve its vision of developing “an air-traffic control-type system that can augment the entire mobility system [of a city] and control it with intelligent algorithms that constantly interact with moving objects – a car, a bike, people – to give predictive optimised routing information”, according to Hofmann. In other words, build “a supercomputer to rule the roads”.
Volkswagen is not alone in its quantum efforts. Nearly every automotive company is currently developing a quantum strategy, including Daimler, BMW, Ford and Toyota. In aerospace, Airbus launched its ‘Airbus Quantum Computing Challenge’ in January.
To develop the traffic management system, Volkswagen experts used, via the cloud, a D-Wave Systems quantum computer or ‘quantum annealer’, to be more precise.
The company’s data science partner, Swiss firm Teralytics, fed anonymised movement data from Orange-connected smartphones and transmitters in vehicles around the Barcelona ICC conference centre into the D-Wave machine and programmed it to predict the need for taxis in the surrounding area.
According to Hofmann, the machine calculated traffic accumulations and customer demand with a 95 per cent probability meaning supply and demand of taxis could be optimised up to one hour in advance. So, using the algorithm, it would be possible to know where a taxi will be one hour in advance or in five minutes.
However, it’s important to note that the solutions provided by the D-Wave machine are only ‘probably’ the best available.
Quantum annealing computers are different from machines being developed by Google, IBM and Rigetti [see box below], which are more commonly referred to as ‘universal’ quantum computers because, in theory, they can be programmed to run any quantum program, whereas quantum annealing, which was originally proposed by Kadowaki and Nishimori in 1998, is a ‘quantum-mechanical metaheuristic’ (generic and approximate method) that can solve combinatorial optimisation and sampling problems only.
Quantum annealers vs universal gate quantum computers
Though both are ‘quantum computers’, quantum annealers and universal quantum computers differ.
First, there is the obvious physical difference: the most advanced D-Wave machine has 2048 qubits, whereas IBM’s commercial computer has 20.
However, more qubits does not mean more coherence, says Dr Stefan Woerner, global leader for quantum finance and optimisation, IBM Research. “Coherence, connectivity and errors are more important than the number of qubits; you can have a 5000-qubit computer, but once entangled you can only exploit the quantumness of five,” he notes. Furthermore, ‘universal’ quantum computers can, in theory, be used to run any quantum problem, whereas annealers can only be used currently for specific types of problem.
However, Bo Ewald, president of D-Wave Systems US, stresses that, unlike universal quantum computers, D-Wave’s annealer architecture is more error-resistant, “so people can still use them by running problems numerous times”, he explains.
Moreover, he stresses, the term ‘universal’ is misleading, as universality is currently “years away”.
Woerner says universal quantum computers are at the “noisy intermediate” stage. This era has already shown potential, with IBM running “toy problems” for financial risk analysis. But truly fault-tolerant gate computers, which are said to be 5-10 years away, will “provide a real opportunity to solve previously unsolvable problems”.
D-Wave, founded in Canada in 1999, claims to have the most advanced available quantum annealing machine, with 2048 qubits [quantum bits]. The company is working with Google, Nasa and Lockheed Martin, among others, on a diverse range of applications.
According to D-Wave’s US director, Bo Ewald, the computer’s architecture works by finding a low-energy solution to a problem it has been posed. “Take the Alps or the Rocky Mountains, for example: without adding or subtracting, the computer will find what is probably the lowest valley in the energy landscape,” he explains.
Each qubit in the D-Wave computer is made from a superconducting loop and starts off in a superposition of two states (0 and 1). In order to exchange information, these qubits are connected by couplers to build up a ‘fabric’ of programmable quantum devices in a chip about the size of a thumbnail.
The computer is programmed by giving a value to each qubit and then programming the strength that qubit has on other qubits. The computer then uses the qubit values, and the coupling, to give what is ‘probably’ the best solution.
It’s possible to know that what is produced is ‘probably’ the best solution because the problem is run thousands of times, with the best result represented over and again. For example, a problem run a thousand times might result in one answer represented 896 times – the ‘best’ choice – and the others only 104 times.
For a traffic-optimisation problem looking at around 4,000-5,000 taxis, it’s possible to receive around 5,000 solutions every 1-2 seconds, according to Ewald.
For the Volkswagen algorithm to test that the solution represented the most was truly the best one, researchers ran the same problem on commercially available optimisation software and got the same answer, but in a timeframe of “between 30 minutes and one hour”, he adds.
As noted, Volkswagen could get the same answer on a classical computer, which raises the question, how much of an advantage does quantum annealing provide for these optimisation problems?
Daniel Lidar, professor of electrical engineering, chemistry, physics and astronomy at the University of Southern California, says none of the currently available quantum computers can do something that a classical computer cannot.
“I think what’s confusing here is the disconnect between the hype in the media and the reality of experimental quantum computing,” he says. “With D-wave computers, you can solve non-trivial problems, but you can’t yet do it with the quantum speed-up relative to every conceivable classical algorithm, and they can only solve certain problems faster than some state-of-the-art methods,” he elaborates.
However, Lidar adds, this in itself is interesting, as it “tells us we’re probably on the right track, but not there yet”.
Neither Volkswagen nor D-Wave particularly dispute this assessment. But for the type of optimisation algorithms being used, the speed-up, whether exponential or not, is important – enough, even.
“For an industrial company like ours, it is useful and interesting to solve problems that can be solved classically, but faster and more accurately with other systems,” say Florian Neukart, principal scientist at Volkswagen Group Americas region.
“Traffic optimisation problems need to be solved within 1-5 seconds; it doesn’t matter what machine does it, but it may not be possible to solve the problem in this time frame on classical computers. That is why we are looking at quantum annealer systems.”
Neukart says Volkswagen could already commercialise the transport optimisation algorithm it launched last year but lacks the necessary partnerships and only has shared access to D-Wave’s machines, which is not suitable for time-critical commercial use. However, it is trialling the Barcelona algorithm now and hopes to launch a quantum-powered taxi app at Web Summit this November.
‘With quantum computing, you can optimise air traffic routes, not just one route at a time, but all routes in one go’
Airbus also sees huge potential in quantum annealing and quantum computing more generally. Chief technology officer Grazia Vittadini says the company deems the technology as potentially the “perfect match” of “unparalleled speed to manage an unparalleled amount of data” that goes hand-in-hand with the aerospace industry’s gradual shift from a purely ‘nuts-and-bolts business’ to a more data-driven one.
The company currently has sensors on around 3,000 Airbus planes – each A350 has more than 250,000 – and wants to use the data collected to solve key issues for the aerospace industry, including optimising aircraft climb and aircraft loading, better computational fluid dynamics and wingbox design.
Among other projects, the company is working on D-Wave computers to try and “win time”, as Vittadini says, the potential for which she calls “mind-boggling”.
Using D-Wave computers, Airbus has seen positive results for investigating all possible routes of failure in very large and complex systems, including safety-critical failures in aircraft and spacecraft. “To do this, we must investigate every single possible anomaly to get to the problem root, and today these types of simulations typically run over several days,” explains Vittadini.
A classical computer simulates different scenarios one by one, until all possible combinations of faults are known. It will then calculate the probability of one of these scenarios leading to a critical failure. Quantum computers can calculate a combination of scenarios simultaneously, and in conjunction evaluate the probability of failure over time.
Vittadini says testing together with QC Ware, a quantum software company that Airbus has invested in, on classical and D-Wave computers combined for this analysis proved 400 per cent faster than classical computers alone. “This fault tree analysis is a very simple example, but it gave us true confidence in the future potential,” she says.
However, she notes that D-Wave machines can calculate a combination of failure routes in parallel, but not the probability of failures. Those are extracted afterwards, based on the known routes of failure.
Airbus also had success optimising air-traffic management, which is becoming a real bottleneck for industry growth.
“With quantum computing, you can optimise air traffic routes, not just one route at a time, but all routes in one go,” says Vittadini. “Potentially, we could coordinate 100 planes in parallel, each flying to five different locations a day, arriving in time, as scheduled, with minimum time on the ground; this is a typical quantum type of problem statement,” she explains.
Dr Markus Leder, from Daimler’s artificial intelligence and quantum computing arm, says the company is working with D-Wave, Google and IBM. He believes that although “fuzziness” is inherent in quantum physics and unavoidable, it is more important “to find a good solution quickly than the best solution after a long period of time. This is what a quantum computer can do much faster than a classical computer.”
And the next generations will be better. D-Wave hopes to release a 5,000-qubit computer in “a year, maybe less”, says Ewald. He expects commercial solutions supported by D-Wave computers in the next year, and in 10 years for them to be scheduling airport traffic and routing autonomous vehicles. Volkswagen has already completed research on managing the real-time speed of autonomous vehicles using D-Wave.
Yet Seth Lloyd, a professor of mechanical engineering and physics at MIT, who says he has been sceptical of D-Wave’s efforts for some time, says some of the expectation around quantum annealers is likely “hype”. But he adds: “Conversely, it does make sense that, if quantum computers can solve problems better than classical ones, which is quite possible, you want to be there already – I think by persisting we are going to find out if it does make sense.”
The major manufacturers certainly think so. However, it might not be technology that scuppers a ‘quantum-powered traffic optimisation app’ but acceptance and regulation. Either way, these firms seem determined to continue their journey.
Better battery design
Both annealing and universal quantum computers hold huge potential for the development of more efficient batteries – the holy grail for automotive manufacturers developing electric vehicles.
Batteries are inherently quantum. The process which powers them, the chemical reaction, is by nature quantum-mechanical. The chemicals are molecular components or even smaller – atomic ions – that slowly diffuse from one end of the battery to the other, resulting in a chemical reaction.
Classical computers are notoriously bad at simulating quantum-mechanical processes, whereas on quantum computers, experts say, it would be significantly easier and give a considerable commercial advantage to any company.
Volkswagen’s CIO Martin Hoffman has said the company has seen a “strong indication” and “first good results” that quantum computers could optimise batteries for 30-40 per cent more power.
Company experts have already shown on a D-Wave machine that it’s possible to simulate simple molecules, such as lithium hydride and beryllium oxide, which can already be modelled on a classical computer.
“To know we can do this is good, but the hope is the next generation of D-Wave chips will come with more interactions between the qubit nodes – from six connected currently up to 15 – so we can make more complex simulations on smaller chips,” says Florian Neukart, principal scientist at Volkswagen Group Americas region.
To model complex systems currently requires physically building systems and testing them – so digital simulation would create huge time savings.