Traditionally, automotive electronic control units (ECUs) have been quite resource constrained due to cost limitations. However, the development rate and the complexity of automotive software is starting to exceed the capacity of existent ECU hardware.
Partly for this reason and partly motivated by falling hardware prices, high performance hardware solutions, similar to those used in applications such as mobiles, media and networking, will have to be considered. Also, following the trend of shorter development and product life times, new hardware solutions will be introduced more fre- quently in the near future.
For these reasons, there is a need for a realistic test and evaluation platform for the research and development of future automotive ECU architectures. At the very least, such platform should consist of a network of embedded systems, representative for the future types of automotive ECUs. It should conform to existing standards, be easily extendable, and preferably open-source.
In this work, first steps towards such a platform are taken. Briefly put, this paper presents the experiences during port- ing of an operating system (OS), commonly used in automotive applications, to Raspberry Pi, a cheap and widely available high performance ARM-based embedded platform. For the sake of compliance with existing standards, the OS of choice relies on AUTOSAR, the prevailing software architecture standard in the automotive industry.
In this design ArcticCore – one of the very few open-source AUTOSAR implementations – was chosen. ArcticCore contains among other things a real-time OS, memory management, and ECUAL software. For some hardware architectures, mostly those that are currently popular in the automotive industry, such as Freescale MPC5xxx, also the µCAL software is implemented. This work is based on AUTOSAR 3.2, with the focus on extending existing ArcticCore source code to make it runnable on Raspberry Pis.
The results address several concerns. Firstly, experiences from the porting work, including typical pitfalls and oppor- tunities, are presented, to serve as a basis for future porting work of AUTOSAR to new hardware.
And secondly, an embryo of an open automotive hardware platform is demonstrated through an experimental setup, consisting of interconnected AUTOSAR-compliant Raspberry Pis, communicating through a Controller Area Network (CAN) bus, typically used in automotive applications.
To read this external content in full, download the complete paper from the open online author archives at SIGBED Review