Detecting convoys in networks of short-ranged sensors

Detecting groups of vehicles travelling together as a convoy is an important problem in military and law enforcement applications. License plate recognition sensors provide discrete, irregularly sampled, time series information about where vehicles are travelling. With this irregular time series, we would like to determine when vehicles travel as a convoy. We construct a semi-Markov process to model network traffic and utilize the Markov property to develop a sequential hypothesis test. This requires defining two models for how vehicles travel through the network and testing the likelihood between them. The main contribution of this work is the modeling of the alternate hypothesis of when two vehicles are traveling as a convoy. We present performance results based on simulated data showing the tradeoff between false-positives and true detections.