Fuzzy Q-Learning Admission Control for WCDMA/WLAN Heterogeneous Networks with Multimedia Traffic

In this paper, admission control by a fuzzy Q-learning technique is proposed for WCDMA/WLAN heterogeneous networks with multimedia traffic. The fuzzy Q-learning admission control (FQAC) system is composed of a neural-fuzzy inference system (NFIS) admissibility estimator, an NFIS dwelling estimator, and a decision maker. The NFIS admissibility estimator takes essential system measures into account to judge how each reachable subnetwork can support the admission request's required QoS and then output admissibility costs. The NFIS dwelling estimator considers the Doppler shift and the power strength of the requested user to assess his/her dwell time duration in each reachable subnetwork and then output dwelling costs. Also, in order to minimize the expected maximal cost of the user's admission request, a minimax theorem is applied in the decision maker to determine the most suitable subnetwork for the user request or to reject. Simulation results show that FQAC can always maintain the system QoS requirement up to traffic intensity of 1.1 because it can appropriately admit or reject the users' admission requests. Also, the FQAC can achieve lower blocking probabilities than conventional JSAC proposed in and can significantly reduce the handoff rate by 15-20 percent.