Winners announced in the BMW Group Quantum Computing Challenge
The four winning teams of the BMW Quantum Computing Challenge were announced this morning at the annual Q2B conference in Santa Clara, California. The challenge, focused on discovering potential quantum computing solutions for real-world use cases, was a collaboration between the BMW Group and the Amazon Quantum Solutions Lab Professional Services team.
“We at the BMW Group are convinced that future technologies such as quantum computing have the potential to make our products more desirable and sustainable,” said Dr. Peter Lehnert, Vice President BMW Group Research and New Technologies Digital Car. “We have succeeded in reaching the global quantum computing community with our crowd-innovation approach and enthusing them about automotive use cases. We look forward to continuing to work with the winners.”
BMW Group Quantum Computing Challenge Use Cases
The BMW-AWS team selected four use cases for the competition: pre-production vehicle configuration, material deformation in production, vehicle sensor placement, and machine learning for automated quality assessment. Challenge participants were given AWS credits to use on Amazon Braket where they had access to available QPUs from D-Wave, IonQ, and Rigetti, and also quantum circuit simulators.
Use Case: Pre-Production Vehicle Configuration: One QuBit eNTiTy
The goal of this use case was to perform a maximum number of required tests on a minimum number of vehicles, while accounting for buildability and scheduling constraints to optimize pre-production vehicle testing. The winning team of One QuBit eNTiTy (comprised of researchers from 1Qbit, NTT Research, and NTT Data) provided a comprehensive solution strategy, including both hybrid (quantum-classical) approaches with near-term impact, and also a long-term quantum-native solution for fault-tolerant quantum hardware. The latter relies on a combination of the Duerr–Høyer algorithm for Quantum Minimum Finding and Quantum Amplitude Amplification, for which One QuBit eNTiTy worked out a thorough scaling analysis. Complementary to this long-term approach, the One QuBit eNTiTy team also developed a native and highly modular hybrid optimization technique, utilizing a fast classical or quantum MaxSAT solver as a sub-solver. The solution also comes with plugins for quantum-inspired devices, such as coherent Ising machines. Preliminary numerical experiments show promising results worth further investigation.
Use Case: Simulation of Material Deformation in Production: Qu&Co
This use case involved the development of novel quantum algorithmic approaches to model and numerically simulate material deformation, as relevant for the accurate prediction of material properties in the pre-production phase of vehicle component manufacturing. Mathematically this problem can be generalized to solving non-linear partial differential equations. The winning team from quantum computing startup Qu&Co provided a detailed, NISQ-ready solution strategy to this use case, based on differentiable quantum circuits. Moreover, the Qu&Co team included in their proposal promising benchmark comparisons to exact results, and results based on classical neural networks.
Use Case: Vehicle Sensor Placement: Accenture
Modern vehicles come with sensors to help provide safety and convenience to drivers. Vehicles need these sensors to gather data from as large a portion of their surroundings as possible, but each additional sensor adds costs. The goal of this use case was to optimize the positions of sensors to allow for maximum coverage while keeping the required number of sensors as low as possible. The Accenture team provided a holistic workflow for prototyping, from user input all the way to the final result, involving a detailed pipeline for (i) the definition of the input data, (ii) pre-processing steps, (iii) optimization of the underlying MaxCover problem and (iv) visualization of the results with an advanced sensor distribution visualization app. For the actual optimization problem, the Accenture team developed a general framework including four classes of algorithmic approaches. While classical custom algorithms delivered the best results today, the framework from the Accenture team comes with plugins for quantum methods to be elaborated in the future.
Use Case: Machine Learning for Automated Quality Assessment: QC Ware
Manufacturers today heavily use classical deep-learning algorithms to assess vehicle parts for cracks and scratches caused by the metal-forming process, based on the successful segmentation and classification of imperfections. In this use case, the goal was to explore novel quantum or hybrid classical-quantum machine learning (QML) approaches that have the potential to provide more efficient training with higher accuracy to help improve the automated quality assessment of vehicles. The winning team from QC Ware provided a comprehensive solution involving three novel quantum algorithms for automated image classification. They were based on improvements for linear algebra routines to implement faster convolutional products compared to classical counterparts. In addition to complexity-theoretic arguments, the team presented extensive classical numerical simulations on two different image datasets. While today QML will not be able to solve this classification task better than established classical methods, the work from the QC Ware team allows us to understand where and how we can use quantum to enhance deep learning techniques in the future.
About the competition
Submissions were received from more than 70 teams globally, spanning from quantum software startups to enterprise companies. The jury evaluated the submissions for comprehensibility, feasibility, scalability, innovation, and benefit for the BMW Group. From the initial submissions, 15 finalists were selected before the final selection of the four winners.