Efficient Mobility Pattern Stream Matching on Mobile Devices

The increasing amount of mobile phones that are equipped with localization technology offers a great opportunity for the collection of mobility data. This data can be used for detecting mobility patterns. Matching mobility patterns in streams of spatiotemporal events implies a trade-off between efficiency and pattern complexity. Existing work deals either with low expressive patterns, which can be evaluated efficiently, or with very complex patterns on powerful machines. We propose an approach which solves the trade-off and is able to match flexible and sufficiently complex patterns while delivering a good performance on a resource-constrained mobile device. The supported patterns include full regular expressions as well as relative and absolute time constraints. We present the definition of our pattern language and the implementation and performance evaluation of the pattern matching on a mobile device, using a hierarchy of filters which continuously process the GPS input stream.