Mobility modeling and management in wireless networks

We develop efficient and accurate estimation algorithms for tracking the location and dynamic motion of mobile units in wireless cellular and ad hoc networks. The underlying mobility model, for both types of wireless network, is based on a first-order autoregressive (AR-1) model of mobility that accurately captures the characteristics of realistic user movements in wireless networks. Estimation of the model parameters is performed by applying the Yule-Walker equations to the training data. Estimation of the mobility state, which consists of the position, velocity, and acceleration of the mobile station is accomplished via an extended Kalman filter using measurements from the wireless network. The integration of mobility state and model parameter estimation results in an efficient and accurate real-time mobility tracking scheme that can be applied in a variety of wireless networking applications. The mobility estimation scheme can also be used to generate realistic mobility patterns to drive computer simulations of mobile networks. We validate the proposed mobility estimation scheme using mobile trajectories collected from drive-test data obtained from a live cellular network.; The mobility estimation algorithms are used to devise mobility management schemes for ad hoc networks and cellular networks. For the ad hoc network scenario, we develop a distributed mobility tracking algorithm which allows the mobile nodes to cooperatively track their mobility states using only received signal strength measurements and position estimates that are periodically exchanged among neighboring nodes. The scheme makes use of a novel coordinate adjustment scheme to derive a global coordinate system from different local coordinate systems. We further develop a scheme for predicting link availability which can be incorporated into ad hoc routing protocols to improve route stability and overall network performance. In the cellular network scenario, we propose a handoff trigger scheme to anticipate the onset of a handoff in advance using the predicted mobility state. Such a scheme can be used to support seamless handoffs and more efficient resource utilization in a cellular network.