How quickly can a computer make sense of what it sees without losing accuracy? And to what extent can you perform car-related AI tasks on hardware with limited computing resources? Aiming to answer these questions, the Zenseact-CERN collaboration shows great promise for technological advancements in autonomous driving, helping to improve cars’ ability to avoid accidents.
Gothenburg and Meyrin, 25 January 2023. Zenseact announces the end of a three-year research project around deep learning methods and processes in collaboration with CERN (the European Laboratory for Particle Physics). Focusing on computer vision, an AI discipline dealing with how computers interpret the visual world and then automate actions based on that understanding, the purpose of the collaboration was to make Deep Learning faster and more accurate, enabling autonomous cars to make correct decisions faster.
In the interest of the research carried out at car safety software developer Zenseact, the project focused on real-life problems related to autonomous vehicle technology.
“Deep Learning has strongly reshaped computer vision in the last decade, and the accuracy of image recognition applications is now at unprecedented levels. But the results of our research show that there’s still room for improvement when it comes to running the Deep Learning algorithms faster and more energy-efficient on resource-limited on-device hardware. Simply put, machine learning techniques might help drive faster decision-making in autonomous cars,” Christoffer Petersson, research lead at Zenseact, says.
Advances in Deep Learning for computer vision have had a crucial impact on the development of autonomous vehicles, enabling the perception of the car’s environment at ever-increasing levels of accuracy and detail. Deep Neural Networks are used for finding patterns and extracting relevant information from camera images, such as the precise location of the surrounding vehicles and pedestrians.
For an autonomous vehicle to drive safely and efficiently, it must be able to react fast and make quick decisions. This imposes strict runtime requirements on the neural networks employed to run on the embedded hardware in the vehicle. By compressing the neural networks, e.g., by using fewer parameters and bits, the algorithms can be executed faster and use less energy.
For this process, Field Programmable Gate Arrays (FPGA) – configurable integrated circuits used in various science and technology domains – were chosen as the hardware benchmark. Given the limited computing resources available to FPGAs, it is essential to reduce, to a minimum, the required computing resources while preserving accuracy. Since FPGA hardware requires effective compression, they provide a challenging problem.
The main result of the FPGA experiment was a practical demonstration that computer vision tasks for automotive could be performed with high accuracy and short latency, even on a processing unit with limited computational resources.
“The project clearly opens up for future directions of research. The developed workflows could be applied to many industries, for example, automotive.” Christoffer explains.
Many of the challenges faced by future scientific experiments and the automotive industry’s technological challenges require processing large amounts of data in real-time, often through edge computing devices with strict latency and power consumption constraints.
The joint team of Zenseact and CERN researchers carried out this project within an open-source software environment. The collaboration reveals that the largest physics experiment in the world could clearly help autonomous driving. Results show great promise for future speed and accuracy increases in image recognition for autonomous vehicles, helping to improve cars’ ability to avoid accidents. For CERN, it has also been a fruitful collaboration.
“The research is of significant importance for CERN. With AI growing in relevance for particle physics, future development of this research area could be a major contribution even to progress in multiple other areas in society,” Maurizio Pierini, Physicist at CERN, says.
We’re a software company dedicated to revolutionizing car safety. By designing the complete software stack for autonomous driving and advanced driver-assistance systems, from sensing to actuation, we’re fighting to end car accidents and make roads safe for everyone. Zenseact was founded by Volvo Cars, and the teams of more than 500 developers and engineers are based in Gothenburg, Sweden, and Shanghai, China.
Open access links to scientific papers written within the project:
Tel: +46 721799103