We have seen major progress in Advanced Driver Assistance Systems (ADAS) and Automated Driving (AD) in recent years and enthusiasm from consumers for these vehicles seems to have grown during the pandemic. The technology that powers such vehicles has significantly changed the E/E architecture of modern cars: they have moved away from multiple, small electronic control units (ECU) spread over the car to a single or a handful of high-performance, large SoC domain controllers. They are often heterogeneous SoCs with dedicated hardware for graphics, image recognition, neural networks, or signal processing workloads.
The software stack needed to operate domain controllers is incredibly complex and the amount of code is growing. For example, resources and functionalities for instrument clusters, camera monitoring system, navigation, or entertainment are generally isolated in virtual machines that can run RTOS, Android or Linux.
Automated driving software have a huge amount of code: a SAE level 3 vehicle (which requires human supervision and input) has at least 300 million lines of code, and a fully autonomous level 5 vehicle might have more than a billion. The effectiveness to analyze the performance and optimize code across the system is key.
You can add two events to your Embedded World calendar in case you would like to learn more:
Many other Arm colleagues will attend the Embedded World conference this year on various topics such as IoT or machine learning. Check out our event page to know more.
We look forward to meeting you there virtually.
Visit Arm at Embedded World