Learning movement patterns in mobile networks: a generic method

Predicting terminals movements in mobile networks is useful for more than one reason, in particular for routing management. A way to do such prediction is to learn the movement patterns of mobile nodes passing by an access router.
In this paper, the information (e.g. layer 2 measurements) related to the different paths followed by mobiles are learned using a hidden Markov model. Simulations have been done using this method and show it can handle different layer~2 signals and collect statistical information when no such signal is available. Furthermore, the method works when no information is available and can be extended so as to guess the timing of the handoffs.