Toward Neural Networks Solution for Multimedia Support in Mobile Ad hoc Networks

Abstract

In this paper, we detail an integrated model for quality of service (QoS) support with service differentiation based on neural networks in wireless mobile ad hoc networks. The model aims at minimizing end-to-end delay for real-time traffic when the network has low to medium dynamics. The use of multi-service classes allows per-flow end-to-end QoS support. The operation of the model is based on a set of core functionality that is augmented with an intelligent learning layer. Core operations ensure different tasks of routing and QoS support control such as route discovery, resources reservation, admission control, traffic adaptation and detection and recovery of QoS violation. The added intelligent layer learns from the different operations performed by the core functionality by means of a Multilayered Feedforward Neural Network according to user requirements and network state. The integrated model showed in simulations that the core functionality together with the intelligent learning improve the quality of data delivery in the network comparatively to models such as SWAN. Te effect of the learning rate on the convergence of network training was also examined. Using intelligent neural networks tools to support the multimedia services in ad hoc networks is a promising avenue.