Quantum computing could well become transformational for the automotive industry, among other sectors – that much is clear despite the current immaturity of the technology. The opportunities are discussed in a recent Capgemini report, Quantum technologies: How to prepare your organization for a quantum advantage now
The BMW Group is among the first automotive companies to take a practical interest in the potential of quantum computing. In summer 2021, the BMW Group issued a Quantum Computing Challenge, in collaboration with AWS, to crowdsource innovation around four specific use cases where it believed quantum computing could benefit its business by solving complex computational problems.
As a long-standing partner of BMW, Capgemini had the opportunity to compete in the BMW Group Quantum Computing Challenge using the expertise of its own established quantum computing community and lab. We welcomed this opportunity to collaborate with the BMW Group in this exciting area, and to enable our quantum community to compete with some of the world’s other best brains in this field.
The focus was on four specific challenges where it was believed that quantum computing could deliver an advantage over classical computing methods: optimization of sensor positions for automated driving functions, simulation of material deformation in the production process, optimization of pre-production vehicle configuration, and machine learning (ML) for automated quality assessment.
Of these use cases, Capgemini focused on machine learning (ML) for automated quality assessment.
The BMW Group’s statement of the use case
“Due to the rapid development of hardware and software, the past decades have drastically shifted quality control from manual examination towards automated inspection. In light of the required human expertise to hand-tune algorithms, machine learning (ML) techniques promise a more general and scalable approach to quality control. The remarkable success of convolutional neural networks (CNNs) in image processing has revolutionized automated quality inspection. Of course, any technology has its limitation, and for CNNs, it is computation power. As high-performance CNNs usually assume large datasets, datacenters ultimately end up with large numerical workloads and expensive GPUs. Quantum computing may one day break through classical computational bottlenecks, providing faster and more efficient training with higher accuracy.”
The challenge’s first round focused on proposing ideas for applying quantum technology to the chosen use case. Capgemini’s submission was well received, and the team was one of a handful chosen (from around 70 participants) to compete in the second, and final, round. Here, the team had the opportunity to work with BMW’s live data under a non-disclosure agreement.
Capgemini’s multidisciplinary approach
Capgemini was delighted to make it through to the final, especially given that most of our competitors were quantum pure plays that had been working on these technologies for a long time. Our success was due to strong collaboration across a wide-ranging team comprising quantum and automotive experts.
Unlike most other competitors who focused on the specific issue of quantum machine learning (QML), Capgemini considered the breadth of the quality assurance process. To support this holistic approach, the team was expanded to bring in different types of expertise from across the business when needed. Our colleagues at Cambridge Consultants, some of whom are quantum specialists, played a pivotal role alongside our experts in automotive, classical ML, and several other areas.
Benefits of our approach
This wide-ranging approach enabled the team to develop a pathbreaking QML model in just a few weeks, as we’ll describe in our next article. What’s more, our holistic, multidisciplinary perspective meant that, in the same timescale, we could take a wider look at the applicability of quantum to automotive quality assessment more generally, identifying some new opportunities.
For example, we considered quantum sensing, and how it could help with problems such as obscuration of images by stray particles and relieving potential bottlenecks around the ML model. We’ll discuss this approach in a future article in this series.
To help BMW assess scalability and viability, Capgemini also laid the foundations for a roadmap for quantum adoption – a topic that will again be covered in a future article.
The project has revealed the potential for applying quantum computing techniques to real problems, now – without waiting for quantum hardware to mature. This is true both in automotive contexts and also in other industries, such as aerospace and life sciences.
More generally, BMW’s Quantum Computing Challenge has provided a great example of how industry can tap into expertise around the world to help solve its biggest problems and leverage complex technology.
We hope we’ve communicated our excitement about taking part in BMW’s challenge and shown you what we achieved at a general level. Our next article will focus on BMW’s central requirement: applying QML to quality assessment.
Authors include: Julian van Velzen, Edmund Owen, Christian Metzl, Barry Reese and Joseph Tedds.
About Julian Velzen
Julian likes to pioneer. Equipped with a master degree in physics, he put Capgemini's quantum technology efforts on the map, and now leads the computing futures (bits/qubits/neurons) domain from within the group's CTIO++ community. Furthermore, he initiated and led project FARM, a big data solution for small-holder farmers in developing countries.
More on Julian Velzen.