Efficient Location Management by Movement Prediction of the Mobile Host

摘要

The mobile host's mobility profile, in a Personal Communication Network (PCN) environment, is modeled. It is argued that, for a majority of mobile hosts (MHs) for most of the time, the movement profile repeats on a day-to-day basis. The next movement strongly depends on the present location and the time of the day. Such a pattern for individual MHs is learned and modeled at the Home Location Register (HLR), and downloaded to the mobile terminal which can verify its correctness real-time. The model is not static and re-learning is initiated as the behavior of the mobile host changes. The model assumes that the past patterns will repeat in future, and a past causal relationship (i.e., next state depends on previous state) continue into the future. This facilitates the system to predict to a high degree of accuracy the location of the MH. As the model is trained up, the frequency of updates decreases as well as the probability of success in paging improves. The movement-pattern model is continuously verified locally, so that any deviation is immediately detected. The validity of the proposed model is verified through simulations.