Improvements in sensor technology and data processing rates are leading to the collection of vast databases of time series data. These data sets are spawning new applications that are transforming the way we live, work, and learn. In particular in the domain of vehicles, the rapid expansion of sensors to monitor both the state of the system and the occupants has led to an increase in the available data, but the research is still inconclusive on how to best handle this data.
This paper develops one data representation that is scalable in dimension and efficiently stores/retrieves multi-attribute time series in the presence of noise. Here the proposed data representation is a multi-input multi-output autoregressive model (MIMO ARX) with an exogenous input. MIMO ARX models are an advantageous data representation because they are a dimension-reducing representation that inherently describes the inter-dependencies in the data while enabling the creation of efficient noise mitigation approaches. Tests using real-life vehicle data show the effectiveness of these data representations in the application of passenger vehicle localization.