Resource reservation in wireless networks based on pattern recognition

Abstract

Resource reservation is very important for handoff control in wireless networks. Many researches have aimed to predict the user's destination cell based on its movement pattern for efficient resource reservation. In the future networks with small size cells, handoffs will occur more frequently and the user's movement will be more like a random process, so it is not practical to predict the accurate destination of a user. We propose a statistical strategy for resource reservation through the estimation of a user's transfer probabilities, which represent the possibilities of the user leaving the current cell and entering the neighboring cells. The resources reserved for a user in each base station are proportional to the user's transfer probabilities. A mathematical model is proposed to obtain the transfer probabilities of a user from the initial states (position, velocity and direction) through simulation of the user's movement. Neural networks are developed to predict the transfer probabilities of a user from the initial states and facilitate efficient resource reservation.