User-driven Call Admission Control for VoIP over WLAN with a Neural Network based cognitive engine

In this paper we deal with the problem of user-driven Call Admission Control for Voice over IP communications in a Wireless LAN environment. We argue that state-of-the-art solutions to this problem are suboptimal, since they leverage on analytical models whose assumptions are not necessarily verified in the scenario considered. To overcome this problem, we propose a cognitive solution based on Multilayer Feed-forward Neural Networks. According to our solution, the mobile station learns from past experience how application-layer service quality depends on the wireless link conditions. Our performance evaluation, carried out both by simulation and testbed experiments, shows that this solution effectively outperforms state-of-the-art strategies in performing a correct admission decision.