A Practical User Mobility Prediction Algorithm for Supporting Adaptive QoS in Wireless Networks Jonathan Chan and Aruna Seneviratne School of Electrical Engineering & Telecommunications The University of New South Wales Sydney 2052, Australia [jchan, aruna]@ee.unsw.edu.au Abstract A number of user mobility prediction algorithms have been reported in the literature. These may be used for resource reservation and service pre-configuration/ adaptation in future wireless networks to provide QoS guarantees. However, our analysis of some of these techniques using measured cellular performance shows that these models do not accurately represent the mobility patterns of users. As a result, resource reservation schemes need to reserve excessive resources, and the pre- configuration/adaptation does not work well. To over come this, we propose an adaptive user mobility prediction algorithm that limits the reservation and configuration procedure to a subset of cells around the user. The viability and effectiveness of the proposed scheme is then demonstrated through a simulation based on measured data. 1. Introduction Future wireless systems will be required to support the increasingly nomadic lifestyle of people. This support will be provided through the use of multiple overlaid networks which have very different characteristics [ 1]. Moreover, these networks will be required to support the seamless delivery of today’s popular desktop services, such as web browsing, interactive multimedia and video conferencing to the mobile devices. Thus one of the major challenges in the design of these mobile systems will be the provision of the quality of service (QoS) guarantees that the applications demand under this diverse networking infrastructure. We believe that it is necessary to use resource reservation and adaptation techniques to deliver these QoS guarantee to applications. However, reservation and pre-configuration in the entire service region [ 2] is overly aggressive, and results in schemes that are extremely inefficient and unreliable. To overcome this, the mobility pattern 1 of a user can be exploited. If the movement of a user is known, the reservation and configuration procedure can be limited to the regions of the network a user is likely to visit. Recently, a number of schemes that apply user movement prediction to various aspects of mobility management have been reported in the literature. It has been shown through modelling and simulation, that the use of movement prediction is effective to enhance the performance of resource reservation [ 3, 4, 5, 6, 7], handover management [8, 9, 10], and location management [8, 11, 12] schemes. Furthermore, it has been shown that movement prediction can be used for adaptive resource management in wireless systems [10, 13]. Although the use of movement prediction seems to be a promising approach for improving the efficiency, reliability and adaptivity of wireless networks, the actual user mobility patterns are not yet well understood. The above performance studies and others reported in the literature, have used simplified movement models that do not accurately characterise user mobility and consequently lead to unrealistic conclusions. Recently we analysed mobility traces obtained through the use of infrared sensors within a building [14]. This showed that, despite being in a well-defined and stable indoor environment, there was a significant amount of variation, and user movement could not be accurately predicted. To overcome the inaccuracy of movement prediction, an adaptive prediction algorithm was used [9]. 1 Mobility pattern is also known as movement pattern in the literature. These terms are used interchangeably in this paper.