Accelerating AUTOSAR-based Data Transformation

Abstract

In traditional protocols (i.e., CAN/FlexRay), the exchanged data between ECUs are of basic types (e.g., integer/float types for engine-speed/temperature values). In addition, static frame layout is sufficient for typical use cases, where no need to dynamically change the sent/received data. However, lately after the emergence of autonomous driving and with the availability of Ethernet communication with its large bandwidth capabilities in automotive ECUs, the need for a new communication strategy is raised. One of the new communication strategies that have taken place is the Service Oriented Communication. In this thesis, we explore an approach to accelerate one of the AUTOSAR functions that is part of any Ethernet-connected AUTOSAR-based Electronic Control Units. Based on the AUTOSAR standard, a so-called SOME/IP (Scalable Service-Oriented MiddlewarE over IP) transformer is used to serialize application data before they are communicated between ECUs over an Ethernet interface. With future autonomously driven, fully connected vehicles, the need for communication will significantly increase either between different sensors distributed across the vehicle or even between the vehicles themselves. In our approach, we used a Hardware Co-Processor instead of the conventional Software implementation used by all AUTOSAR stack suppliers for data serialization. Having a dedicated hardware for serialization makes the architecture more reliable. Our approach achieves up to 50x speedup over the Software implementation, which in turn helps to cope with the increased demand for data communication required in modern vehicles. This approach also gives more CPU room for other real-time applications to utilize the CPU.