An Intelligent QoS Model for Service Differentiation in Mobile Ad Hoc Networks

L. Khoukhi and S. Cherkaoui (Canada)


Mobile ad hoc networks, QoS, admission control, feedforward neural networks, service differentiation.


In this paper, we develop an Intelligent QoS model based on neural networks for service differentiation in mobile ad hoc networks (GQOS). GQOS is composed of two plans: first, the GQOS kernel plan is used for routing and quality of service (QoS) support. It is responsible of route discovery, resources reservation, admission control, adaptation and QoS violation detection and recovery. Second, the intelligent learning plan allows to learn a traffic class selection that corresponds to the user requirements in terms of QoS metrics. The intelligent plan uses a neural network algorithm in order to assure a behavior that adapts to the changes in the network and to train different GQOS kernel operations. The QoS violation detection and adaptive recovery is assured by a mechanism based on the prediction of the arrival time of data packets. On the other hand, a priority queuing mechanism is integrated in GQOS in order to regulate traffic and to prevent congestion. The simulation results show that our model outperforms the SWAN model at low mobility. It shows that the Feedforward Neural Network used is able to classify all the traffic according to the QoS requirements after a small period of time, which means efficient learning.

Important Links:

Go Back