## What is quantum computing at Volkswagen?

Quantum computing has gained serious mainstream attraction over the past couple of years. Decades of research from universities around the world have yielded industrial efforts in companies such as Google, IBM, D-Wave, Rigetti, and a host of start-ups rushing to create the next groundbreaking technology. The motivation behind quantum computing is promising, as Richard Feynman originally proposed: Nature is quantum (“dammit!”) and the natural power of quantum mechanics could be used to create a sophisticated computer. Since then, various algorithms have been written for quantum computers in the fields of combinatorial optimization, machine learning, quantum materials simulations, and more. Quantum computers are similar to regular computers in the sense that they have processors that perform operations and calculations to produce a result. However, instead of being composed of many bits of information that can be changed from 1 to 0 and 0 to 1 in a controllable way, quantum computers are made of quantum bits (qubits) that can be 0 and 1 at the same time. These qubits have a property called entanglement, which means that all the qubits in the system are connected with each other in a fundamental way. When we change the conditions of one qubit in our computer, it could possibly affect all other qubits in our computer without any additional work. This makes quantum computers extremely powerful, provided one can write an algorithm that exploits these quantum properties. Researchers around the world are constantly coming up with new ideas about how to use these novel devices.

But how is this relevant to Volkswagen? Volkswagen is much larger than just a car manufacturer, putting cars on the road. The challenges that need to be overcome to deliver our services span the entire technological spectrum: Physically building a car is an engineering problem. Scheduling deliveries of cars from factories to dealerships is a logistics problem. Designing better batteries in electric cars is a chemistry problem. Organizing warehouses for parts storage is an optimization problem. The list goes on endlessly, and the IT and R&D infrastructure in Volkswagen need to support all these different areas. And this is where quantum computing fits in Volkswagen, helping to solve these complicated problems with cutting edge technology. We aren’t simply interested in theoretical questions about algorithms and complexity. We need to know how these computers work on a fundamental level, but we also look into the future and see which specific areas of our business they can impact. When researchers talk about a new quantum algorithm, we look at our business and ask ourselves: where can this help us? Can we use this to organize our factories better? Can we now simulate new materials we weren’t able to before?

Historically, the focus has been on showing how quantum computers could use these algorithms to solve problems “faster” than traditional computers. A popular example for this is Shor’s algorithm, which factors a number into its prime factors, a process used often in cryptography and cybersecurity. Imagine a hacker trying to decrypt a secure file: the hacker knows the key is composed of some combination of numbers multiplied together, but not which numbers. The time it takes a computer (or a hacker) to go through these combinations grows rapidly with the number of digits in the key, which is what makes these methods of encryption (like RSA) secure. However, for a big enough quantum computer, something entirely different happens. Instead of going through individual combinations of numbers, the hacker needs to apply a sequence of instructions (called quantum gates) and the quantum computer can determine on its own what the correct combination of numbers produce the decryption key. Most importantly, the number of instructions for the quantum computer depends only on the number of digits in the key, making this algorithm extremely effective. This is the reason many people are now looking into something called “post-quantum cryptography”, the idea of creating new encryption algorithms that can withstand the processing power of a quantum computer. This example is a typical benchmark for quantum computing: what problem can we solve, or what algorithm can we write, where we know that quantum computers will *always* be better? Pursuing the answer to this question has led to the introduction of a term called “quantum supremacy”- a task performed by a quantum computer that beats every known classical approach. Over the last few years, researchers have been working hard to find cases of this, both in theory and in practice, with limited success. We now have a small collection of problems we think can be solved most efficiently with quantum computers (factoring and quantum chemistry simulations are most often used as examples), but building quantum computers that are big enough to run these algorithms is still a few years away. In the nearer future, a different benchmark is becoming popular, called “quantum advantage”. As opposed to supremacy, quantum advantage has the relaxed condition that the quantum computer simply needs to provide a tangible gain by using it as opposed to a classical computer. This could be in terms of cost, runtime, or some other beneficial metric. There is considerable debate about what the correct benchmark should be, and what the best path is to viable quantum computing.

At Volkswagen, we see these two ideas as separate, but not necessarily competing, concepts. On the one hand, quantum supremacy can help us solve problems that previously couldn’t be tackled. For example, simulating quantum properties of materials is notoriously difficult and requires significant computing power. This means that R&D in this area can involve painstaking hours of laboratory work developing a material, and taking careful measurements over and over again to get the necessary information. But quantum computers are inherently quantum materials themselves. If we had access to a large enough quantum computer, we could simply write a program that sets the conditions we want to test and let the quantum computer do the work. By measuring the results of this quantum program we could simulate the interactions and properties we are interested in before even producing the materials, allowing us to do research in a fundamentally new way. On the other hand, with quantum advantage, we can look to improve existing techniques to solve complicated problems. One field this is particularly suitable for is logistics. Every day at Volkswagen we ship things from one location to another. Whether we’re sending car parts from storage to factories, or products from factories to dealerships, a variety of conditions need to be taken into account when optimizing logistics. Weather or traffic conditions can change by the hour, and our systems in Volkswagen need to be able to deal with these unpredictable fluctuations, otherwise both our customers and the company suffer. With quantum computing, we could write complex programs that take such variables into account in real-time, allowing us to sort through endless combinations of solutions to these problems quickly, saving both time and money for the company and its workers. For this reason, we have worked hard over the past few years to build up our quantum computing expertise with people in San Francisco, Munich, and Wolfsburg, so we understand how to take cutting edge research and adapt it to help Volkswagen.

## How Volkswagen is investing in quantum computing now

Innovation is the key to survival in business. Being able to adapt to a changing market, and staying ahead of the technology curve is imperative to company growth. Stagnation will quickly lead to falling behind the rest of industry, making it hard to compete. We see our quantum computing team at Volkswagen as an investment in the future. Our goal is always to put the best products on the market, and the easiest way to accomplish this is by leading the industry through research and development. This means staying up-to-date with the latest trends in technology, and developing the skills necessary to use them. In this aspect, quantum computing is no different. We cannot wait for quantum computers to impact the market before we learn how to use them. That is why we are constantly learning and developing internally, connecting quantum computing to different branches in Volkswagen, to assess where it can be most effective. To this end, we have ongoing collaborations with quantum computing leaders Google and D-Wave Systems, and together work towards developing the tools and strategies needed to incorporate quantum computing into our existing infrastructure. The state of quantum computing now is similar to that of computers in the 1950’s. There are no standard operating systems, no programming languages or compilers. The work being done now by quantum computing researchers, both on the hardware and software side, is laying the groundwork for future users of these devices. The industry of quantum computing is ripe for developers at the moment, people who are eager to understand the technology at low levels, and help build the stack on which future applications will sit on. The main challenge for the coming years is to come up with techniques that intelligently leverage the quantum resource that’s provided by these devices. At Volkswagen we are contributing to this in many different aspects.

Volkswagen’s involvement in quantum computing started in late 2016 with a proof-of-concept project with D-Wave Systems to investigate the readiness of quantum computing for industry. The goal was to showcase a small prototype program at CeBIT 2017 to highlight why quantum computing was interesting for a company like Volkswagen. The result was a traffic-flow optimization program that used GPS coordinates of 418 taxis in Beijing to resolve traffic congestion. Through this project our researchers in San Francisco and Munich learned a fundamentally different paradigm of programming for quantum computers: instead of writing many lines of code for a computer to step through, we simply set the appropriate initial conditions and constraints for the quantum computer, and let it evolve naturally. When this process is performed correctly, reading out the solutions provided by the quantum computer resolved the traffic congestion. The importance of this work for Volkswagen was two-fold. Firstly, by doing the development work ourselves we retained the experience and knowledge required to work with quantum computers. Secondly, and most importantly, we were able to translate data taken from a real-world system (GPS coordinates) and use this to write a program that solved a difficult problem using a *real *quantum processor, which had never been done before. From an industrial perspective, this project proved vital in forming Volkswagen’s path forward into quantum computing. After learning that it was indeed possible to use quantum computers for problems that are relevant to Volkswagen, and with the strong support from Volkswagen CIO Martin Hofmann, we were able expand our research and development efforts. Today, we focus on applying our knowledge to research areas where we see the best fit between quantum computing and Volkswagen.

One of the most promising applications of quantum computing is quantum machine learning. Machine learning algorithms are robust against noisy data by design, and often can use this noise to improve performance. Because near-term quantum devices aren’t advanced enough to work in ideal conditions, this makes them perfect for quantum machine learning. Most recent proposals in the quantum machine learning community focus on taking existing classical algorithms and translating them to the realm of quantum computers. This can either mean extracting certain parts of an algorithm that can be sped up by executing them on a quantum device, or finding purely quantum analogs of existing algorithms. In-between these two there are the so-called hybrid quantum classical algorithms, which jointly use both types of computers. One of our main research topics is implementing these machine learning networks directly on a quantum processer (called a “quantum neural network”). By doing this on a quantum processor instead of a regular computer, we can encode much more information in our quantum network, due to how the qubits interact with each other. In our work we research various methods of controlling the quantum gates which govern the interactions between the qubits to give us the best results for our quantum machine learning algorithms. Along with developing the algorithms themselves, we also look for tasks where quantum machine learning algorithms are more suitable than classical machine learning. This allows us to build solutions inside Volkswagen using current machine learning techniques, but including the power of quantum machine learning. These solutions can in turn be used for practical applications. Problems like predicting maintenance for machines in our factories, price changes in the automotive market, or prediction of traffic congestion are all still intractable even for current machine learning solutions, making them good candidates for quantum machine learning algorithms. Together with researchers from Google, we are working towards creating quantum machine learning algorithms that will enable us to do all these applications and more on the quantum computers of today and those that are still to come.

Applications of quantum computing are not all confined to the theoretical. A good example of this is optimization problems, another proposed sweet-spot for quantum computers. In business, as they say, time is money, and so everything can be boiled down to an optimization problem. Every decision in a company is weighed against resources, financial cost, and opportunity cost- and the longer it takes to make a decision the higher the cost. Anything from amending logistics schedules to mobility solutions to product pricing can be formulated as an optimization problem. For this reason we’ve been working with our partners at D-Wave and Google to learn how we can use their quantum processors to solve our internal optimization problems. We have already produced multiple prototypes of optimization solutions using quantum computing. At the Web Summit 2018, we showed how we can use a D-Wave quantum processor to assign taxis on-demand in Barcelona, and at CeBIT 2018 we showed how one could optimize a car mirror’s shape to reduce acoustic noise with a hybrid quantum-classical algorithm. Recently we also published a first-of-its kind demonstration of quantum materials simulation of small molecules, which will allow us to investigate new materials and their properties. We showed how a D-Wave quantum processor could simulate properties of these molecules by translating the simulation to an optimization problem. With our collaborators at Google, we’ve been working on solving complex optimization problems with a significantly different approach. We are researching methods to compile quantum circuits in a way that exploits the inner structure of these optimization problems to compress the circuits. This will allow us to make even more efficient use of quantum computers when solving these problems, much like compilers work in regular computers today. We can then train these quantum circuits to solve problems like traffic flow, quantum machine learning, and high-dimensional optimization of car parts. While many of these projects are small relative to the size of our company, the knowledge and experience they provide help us prepare Volkswagen for the arrival of bigger quantum computers in the future.

## Volkswagen’s future in quantum computing

From Richard Feynman’s proposal in the 1980’s, to the device prototypes of the early 2000’s, to the quantum processors of today, there is no doubt that quantum computing has advanced enormously over the years. The way this technology works is a fundamental shift in how we view computing, requiring a different approach to both designing and programming these computers. It seems unlikely that quantum computers will ever replace traditional computers in the workplace, but they are poised to be great companions to classical computers, accelerating the way we view computing and allowing us to perform tasks we were unable to before. The truth about quantum computing is that it’s too early to know exactly what the first “killer application” will be. That the technology holds enormous potential has become clear over time, and at Volkswagen we have invested deeply in quantum computing to position ourselves for the future. We have to be ready today for the technology of tomorrow. No matter what field this “killer application” of quantum computing will come from, at Volkswagen we must find a way to use it to benefit everyone. We can use our optimization solutions for smart mobility platforms that will reduce traffic congestion in large cities. Quantum machine learning algorithms could revolutionize autonomous cars, making them safer and more reliable at a lower cost. With new material simulations we can research better components for our products, reducing our waste production and impact on the environment. After decades of research, we are finally reaching the point where quantum computers are leaving the research labs and becoming accessible to industrial early adopters. For us at Volkswagen it’s not enough to be users of this technology, we need to become experts in it to help ourselves and others accelerate the growth of the quantum computing community. This is why we are constantly expanding our quantum computing efforts by partnering with industry leaders, attending conferences and events around the world. Internally within our company, we work closely with experts in related fields to understand what their needs are, and how we can help Volkswagen grow using quantum computing wherever possible. We know that these quantum devices, although small today, are capable of doing extraordinary things in a fundamentally different way. We will continue to showcase their use in industry applications like we’ve done with mobility optimization and materials research, while we prepare for the quantum computers of the future. Whichever path leads in the end to the quantum computing revolution, Volkswagen is committed to leading the way.

### Über den Autor / die Autorin:

**Sheir Yarkoni** is a quantum computing researcher at the Volkswagen Data:Lab in Munich. Before joining Volkswagen he worked for D-Wave Systems with a focus on benchmarking quantum processors for industry applications. His work includes doing a Ph.D. in hybrid quantum-classical algorithms for applications at Volkswagen and Leiden University.

**Andrea Skolik** gained industry experience in many application areas before joining the Data:Lab including machine learning, e-commerce, and finance, and holds a Masters in computer science focusing on machine learning and robotics. She is currently a Ph.D. student at LMU in Munich researching quantum machine learning.

**Florian Neukart** is the principal quantum computing researcher for Volkswagen, based at the CODE:Lab in San Francisco. He holds a Ph.D. and several Masters degrees in computer science. Before moving to the CODE:Lab, Florian brought quantum computing to Volkswagen as co-founder of the Data:Lab in Munich. Florian is also an Associate Professor at Leiden University for quantum computing.

**David von Dollen** is a Senior Data scientist based in the Volkswagen CODE:Lab in San Francisco, holding a Masters in Computer Science and machine learning from Georgia Institute of Technology. His work involves accelerating machine learning algorithms with new techniques using quantum computers, and is working towards a Ph.D. at Leiden University.

**Martin Leib** joined the Volkswagen Data:Lab quantum computing team after post-doctoral stints in Japan, Scotland, and Austria. He obtained his Ph.D. in physics, developing superconducting processors at TU Munich. Martin leads the Data:Lab research in optimized quantum circuits for quantum machine learning and optimization.

**Michael Streif** is a physicist currently doing his Ph.D. with University of Freiburg at Volkswagen. Before that he studied in Freiburg and Oxford. His current research involves understanding the physical effects governing quantum processers and their applications to optimization problems.