Available bandwidth prediction using wavelet neural network in mobile ad-hoc networks

Many multimedia applications over Mobile Ad hoc NETworks (MANETs) require Quality of Service (QoS) to meet real-time services. Accurate Available Bandwidth (AB) prediction and allocation of AB are important component for QoS provisioning which is affected by many factors such as latency, bandwidth, reliability, packet-loss, memory size, buffer cache, available capacity, and CPU speed. Media Access Control (MAC) protocol is responsible for efficient usage of AB in MANET to provide QoS. In this paper, we propose a novel AB prediction mechanism in MANET that is necessary for efficient AB allocation to support real-time and multimedia communication. AB prediction mechanism is being designed with wavelet neural networks. Simulation result shows that the predicted resource closely match with the actual values. Maximum variation between predicted AB and real AB is approximately 20%.