ROS2 versus AUTOSAR: Automated Parking System case-study

Vehicles are complex systems as they combine several engineering disciplines, such as mechanical, electric, electronic, software and telecommunication. In the last decades, most innovations in the automotive domain have been achieved as a combination of electronics and software. Consequently, the software development and deployment has resulted a highly sophisticated engineering process to manage and to integrate. With the introduction of artificial intelligence, automated driving has become a reality. However it has additionally increased the requirements on the system design. One widely accepted approach to manage complexity is to divide the system into subsystems through a well-defined architecture. The architecture of an autonomous system must be suitable to guarantee that the self-driving functionality remains safe in a broad range of operational domains. The challenge is how to design the architecture of the system to be reliable and resilient to changing context. The automotive industry has well established standards and development practices, but it is open to explore and integrate solutions from other domains like Internet of Things and Robotics. In the area of autonomous systems, the capabilities of the robotics middleware ROS2 have been used for prototyping purposes. It is an open question whether ROS2 is suitable for automotive safety relevant applications. This master thesis addresses this challenge through evaluating the possible application of ROS2 in the automotive domain. The development consists of implementing an architecture for an autonomous driving function case-study, an Automated Parking System, which adapts to its context by switching between different operational modes. The Automated Parking System has been implemented and validated in a simulation environment. The experiment results show which benefits bring ROS2 compared with the automotive standardised architecture AUTOSAR.