Towards Heterogeneous Computing Platforms for Autonomous Driving

摘要

This paper introduces a new project: Research and Development of System on a Chip and Software Platforms for Autonomous Driving, started in April 2019. State-of-the-art high-end embedded platforms have trade-offs between general-purpose processing and power consumption. In order to solve the trade-offs, this project develops heterogeneous computing platforms for autonomous driving. In particular, the heterogeneous computing platforms have a heterogeneous System on a Chip (SoC), which consists of Accelerated Processing Units for AI and robotic algorithms, multi-core CPUs for general-purpose processing, and many-core CPUs for real-time processing. This project also develops software platforms including compilers, operating systems, middleware, and applications for the heterogeneous SoC. The authors believe that this project contributes to early realize autonomous driving.