Mobile Intelligence

11.2 MOBILITY PREDICTION USING MOVEMENT HISTORIES

Mobility prediction, which is used for predicting the future trajectory of a given moving object, is a widely researched topic in mobile computing. Consider a mobile user in a cell-based mobile phone network who is moving continuously while making a call. The underlying network system has to transfer his calling status between cells [70]. If the next cell to which a mobile user will move can be predicted, then an efficient resource reservation and quick handover between base stations can be achieved. So far, various mobility prediction methods have been proposed. A study [6] provides a good survey of this topic. It roughly classifies the approaches to mobility prediction into two categories:

• Domain-independent methods: Locations or cells are treated as symbols, and only location names are considered, without taking other semantics into account.
• Domain-specific methods: Additional information, such as coordinates, directions, and velocities of moving objects, road networks and map information, and/or facility locations are used.

In this subsection, we introduce some selected mobility prediction models that utilize movement histories, since they are related to the concept of data mining. We focus in particular on Markov predictors and their extensions.

11.2.1 Domain-Independent Markov Predictors

We first describe the most basic type of mobility predictors, called Markov predictors, and their variants.

11.2.1.1 Markov Predictors ...