Abstract
Modern vehicles contain a highly complex network of hardware and software subsystems (1). To be able to locate faults or to evaluate the behaviour of subsystems, the communication on the vehicle’s networks is being recorded by measurement systems (13) – so called data loggers. This is for example done during road trials or Hardware-in-the-loop (HiL)-tests. The amount of data resulting from each recording is huge – easily in the region of several million data points. Analyzing the recordings is very time-consuming. This paper discusses an approach for the analysis of mass data resulting from vehicle tests.
The recordings can be viewed as time series data (2). In order to uncover faults or evaluate the performance of algorithms, this research work proposes an approach to interactively explore the data recordings, which is referred to in this work as interactive knowledge discovery from multivariate time series.
The traditional way of presenting this type of data – plotted w.r.t. time or as individual scatter plots – is not sufficient due to the complexity of the data. In this work, a combination of various techniques from the fields of visual data exploration (7) (8) and temporal data mining (4) is applied. The approach combines and enhances the two existing visual data exploration techniques “parallel coordinates” (10) and “scatter plot matrix” (2) to cope with time series resulting from vehicle tests. Additionally a facility to query a time series by graphically formulating a search pattern is integrated. This enables the user to interactively analyse the recordings by formulating sophisticated filtering and querying operations. The approach was implemented and shall be named “Automotive Trace Miner”.
Keywords: vehicle electronics, visual data mining, temporal data mining, time series, automotive trace analysis