Fusion: A Safe and Secure Software Platform for Autonomous Driving

Abstract

The vastly increasing amount of software in vehicles, its variability and complexity, as well as the computational requirements, especially for those built with autonomous driving in mind, require new approaches to the structure and integration of software. The traditional approaches of single-purpose embedded devices with integrated software are no longer a suitable choice. New architectures introduce general purpose compute devices, capable of high-performance computation, as well as high variability of software. Managing the increasing complexity, also at runtime, in a safe and secure manner, are open challenges. Solving these challenges is a high-complexity development and integration effort requiring design-time and runtime configuration, approaches to communication middleware, operating system configuration, such as task scheduling, monitoring, tight integration of security and safety, and, especially in the case of autonomous driving, concepts for dynamic adaption of the system to the situation, e.g., fail-operational concepts. We present Fusion, a next-generation software platform supporting the development of autonomous driving systems.