Mobility Prediction in MANET using Neural Learning Methods

The system is developed for exposing the correlation between the changes in mobile node location, speed and direction in order to effectively predict its mobility. Also, the life of node battery is improved by predicting routing tables in order to reduce data exchange in MANET. For these purpose, a neural learning based framework is proposed in which the future changes in the network topology are efficiently predicted. This framework is based on the architectures of the standard Multi-Layer Perceptron (MLP) and the Extreme Learning Machine (ELM). The ELM does not require any parameter tuning and is unbiased to initial weights also this predictor capture better correlation between Cartesian coordinates of arbitrary MANET nodes leading to more realistic and accurate prediction. Also, the complex-valued representations are achieved to reveal the correlation between the changes in mobile node location, speed and direction. This proposed predictor is also used to predict the routing tables which are used to reduce the data exchange in MANET.