Improved Markov predictor in wireless networks

Abstract

As wireless networks have been an integral part of our daily life, mobility prediction techniques have become one of the main topics in current research efforts. An accurate prediction of the next cell to which the mobile users are going can greatly improve the performance of wireless applications, such as map resource allocation, congestion control, quality of service and mobility management. It has been shown that the Markov predictor is a good mobility predictor in actual wireless local area network environments. However, from the standpoint of conditional entropy, the authors analyse that the Markov predictor has the disadvantage of performing worse when the location history is lacking or an approximate tie has happened. As a consequence, a novel improved Markov predictor is proposed, and simulations are conducted to evaluate the performance of the proposed scheme. The simulation results show that the improved Markov predictor solves not only the disadvantages of Markov predictor due to the lack of location history information, but also the expansion of state space in multiple-order Markov predictors.