On-line Detection of Collective Mobility Patterns through Distributed Complex Event Processing

Applications such as fleet management, mobile task force coordination, logistics or traffic control can largely benefit from the on-line detection of collective mobility patterns of vehicles, goods or persons. However, collective mobility pattern analysis is exponential by nature, requires the high-throughput processing of large volumes of mobile sensor data, and thus generates huge communication and processing load to a monitoring system. Considering the benefits of the event-based asynchronous processing model for on-line monitoring applications, in this paper we argue that several collective mobility patterns can be elegantly described as a composition of reusable Complex Event Processing (CEP) rules, and specifically focus on the detection of the cluster mobility pattern. We also present a DDS-based mobile middleware that sup- ports a distributed deployment of these CEP rules for such collective mobility pattern detection. As means of evaluating our approach we show that using our middleware it is possible to detect this mobility pattern for thousands of mobile nodes, with a latency that is adequate for most monitoring applications.