In this paper, we present the detailed design and implementation procedures for an advanced driver assistance system (ADAS) based on an open source automotive open system architecture (AUTOSAR). Due to the increasing software complexity of ADAS, portability, component interoperability, and maintenance are becoming essential development factors. AUTOSAR satisfies these demands by defining system architecture standards. Although commercial distributions of AUTOSAR are well established, accessibility is a huge concern as they come with very expensive licensing fees. Open source AUTOSAR addresses this issue and can also minimize the overall cost of development. However, the development procedure has not been well established and could be difficult for engineers. The contribution of this paper is divided into two main parts: First, we provide the complete details on developing a collision warning system using an open source AUTOSAR. This includes the simplified basic concepts of AUTOSAR, which we have organized for easier understanding. Also, we present an improvement of the existing AUTOSAR development methodology focusing on defining the underlying tools at each development stage with clarity. Second, as the performance of open source software has not been proven and requires benchmarking to ensure the stability of the system, we propose various evaluation methods measuring the real-time performance of tasks and the overall latency of the communication stack. These performance metrics are relevant to confirm whether the entire system has deterministic behavior and responsiveness. The evaluation results can help developers to improve the overall safety of the vehicular system. Experiments are conducted on an AUTOSAR evaluation kit integrated with our self-developed collision warning system. The procedures and evaluation methods are applicable not only on detecting obstacles but other variants of ADAS and are very useful in integrating open source AUTOSAR to various vehicular applications.
Design and Implementation of Driver Assistance System using Machine Learning
