This paper introduces a method for selectively preprocessing and recording sensor data for engineering testing purposes in vehicles. In order to condense data, methodologies from the domain of sensor networks and stream processing are applied, which results in a reduction of the quantity of data, while maintaining information quality. A situation-dependent modification of recording parameters allows for a detailed profiling of vehicle-related errors. We developed a data-flow oriented model, in which data streams are connected by processing nodes. These nodes filter and aggregate the data and can be connected in nearly any order, which permits a successive composition of the aggregation and recording strategy. The integration with an event-condition-action model provides adaptability of the processing and recording, depending on the state of the vehicle. In a proof-of-concept system, which we implemented on top of the automotive diagnostic protocols KWP and UDS, the feasibility of the approach was shown. The target platform was an embedded on-board computer that is connected to the OBD-II interface of the vehicle. As the scope of recording can be adjusted flexibly, the recording system can not only be used for diagnostic purposes, but also serves objectives in development, quality assurance, and even marketing.
展开▼
