Abstract:

In wireless heterogeneous networks,the access control mechanism plays a vital role in ensuring the net-work performance.This paper proposes a Q- learning- based network selection algorithm for the access control of wireless heterogeneous networks.In the algorithm,the agent takes advantage of the past experience to implement a Q- learning algorithm by means of value iteration,thus obtaining the optimal strategy.Moreover,before the system makes decisions,the statuses of the wireless heterogeneous network of WLAN/WIMAX are analyzed in detail from a new perspective,thus providing accurate input bottom layer parameters for the Q- learning algorithm.Simulation results show that the proposed algorithm outperforms the traditional Markov Decision Process (MDP) algorithm in terms of the call- blocking probability and the system reward.

Key words: network selection, resource management, reinforcement learning, Q- learning