HOW VEHICLE SENSORS CAN BE QUANTUMIZED (AND WHY IT MATTERS)
Key players insights
OPTIMIZING SENSOR PLACEMENT IS KEY TO SAFETY AND CONVENIENCE FOR DRIVERS AND PEDESTRIANS
Picture a bright summer day—perfect weather for a picnic by the sea. You and your friends pack the car and head out. The road ahead? Clear as the sky above. Suddenly, a person walks out from behind a delivery truck, right in front of your car.
We all remember this from driving school: Every car has blind spots. Even with mirrors and cameras, you simply can’t see everything from every angle. On top of that, every human has a certain reaction time, making some events practically invisible. If you see them, it’s likely too late.
SENSOR TECHNOLOGY IS IMPORTANT—BUT EXPENSIVE
Today, many drivers rely on sensors that warn when people and objects veer too close to vehicles. All too often, the warnings come too late, but some manufacturers are using innovative sensor technology to improve precision.
In addition to a front collision warning, BMW vehicles now include rear-end collision and personal injury warnings. The system’s speed range has also been extended, with braking intervention up to 250 kilometers per hour. If an obstacle is detected, the brakes are applied up to full braking before a collision becomes unavoidable. All this is made possible by sensor technology, a crucial factor in advancing automated driving and increasing safety for people inside and outside the vehicle.
Vehicles with automated driving and driver assistance capabilities require a large number of sensors to observe surroundings. However, each sensor adds costs: LiDar and radar, for instance, are expensive, complex systems. For automotive manufacturers like BMW, it’s essential to minimize the cost of sensor setup for every single car during development.
AWARD-WINNING OPTIMIZATION THROUGH QUANTUM COMPUTING
Vehicle sensors and the systems they support must gather and process as much data as possible from the environment, resulting in vast computational challenges that classical computers simply can’t handle within a reasonable timeframe. Quantum computers, however, can speed up the process, helping to find optimal sensor position and shorten car development cycles.
Working with Amazon Web Services (AWS) and BMW, Accenture set out to reposition the sensors on a vehicle for maximum coverage while minimizing their number. Using quantum computing, our Quantum team delivered a process for optimizing sensor position called “quantumization”. The solution won first place in the 2021 BMW Quantum Computing Challenge and is one of four use cases the auto manufacturer has developed for quantum computing.
EASING THE COMPUTATIONAL BURDEN
Increasing safety through quantumization is a computationally intensive task. Our quantum experts addressed this issue with a three-pronged approach:
- Mathematical formulation
- Data preparation
- Visualization
- MATHEMATICAL FORMULATION
The first task was to analyze the use case and formulate it into a mathematical problem.
This formulation was modeled in various computational algorithms, which were combined and run one after another. The team also extended the given formulation to create a more sophisticated problem description - resulting in the modeling of a better solution.
1. DATA PREPARATION AND PROBLEM REDUCTION
Data preparation for quantum algorithms is critical. Data must be preprocessed so that the algorithm can understand the given data input. The data set BMW provided was too large to run on available quantum computers, which were small and prone to “noise”, or interference in the form of stray atoms, vibrations, etc. By building various functions to reduce the degrees of freedom, our team reduced the size of the quantum computer required, making it possible to use smaller computers.
2. VISUALIZATION
The final primary working stream was centered around visualizing the vehicle, its sensors and their coverage. Here, the experts from leading CGI studio Mackevision took over. Their cutting-edge application made it possible to perform two pivotal steps in the process:
- Depict a true-to-original BMW with optimal sensor setup coverage
- Efficiently calculate the surroundings covered by the newfound sensor setup
IMPACT
The sensor positioning optimization initiative was the first end-to-end solution based on a real-world quantum computing use case. With the data set and car coordinates from Mackevision, the team was able to calculate and preprocess all possible coverage outcomes, hand inputs over to algorithms and ultimately decide which outcome to visualize and build.
WHAT’S NEXT
Sensor quantumization is one of four real-world quantum computing use cases that BMW and AWS have published, alongside pre-production vehicle configuration, material deformation production and automated quality assessment. Not only are they all crucial to BMW’s production process, they’re providing valuable data to accelerate growth of the quantum ecosystem. As a result, all are set to be further explored in follow-up projects.
By providing a clear, continuously updated picture of the vehicle, its environment and the activities of drivers and passengers, smart (or “quantumized”) sensor technologies are a key enabler of auto industry transformation—for a safer, more cost-effective vision of tomorrow’s mobility landscape.
