Mobile ad hoc network proactive routing with delay prediction using neural network

Existing MANET routing protocols rely heavily on hop count evaluation. Although this is simple and efficient, it sacrifices the potential performance gains obtainable by considering other dynamic routing metrics. In this paper, we propose a delay prediction mechanism and its integration with a MANET proactive routing protocol. We demonstrate our approach of predicting mean queuing delay as a nonstationary time series using appropriate neural network models: Multi-Layer Perceptron or Radial Basis Function. To support MANET proactive routing, our delay prediction mechanism is devised as a distributed, independent, and continuous neural network training and prediction process conducted on individual nodes. We integrated our delay prediction mechanism with a well-known MANET proactive routing protocol—OLSR. The essential part of this integration is our TierUp algorithm, which is a novel node-state routing table computation algorithm. The structure and the key parameters of the resulting extended OLSR, called OLSR_NN, are also discussed. Our simulation shows that because of its capability of balancing the traffic, OLSR_NN is able to increase data packet delivery ratio and reduce average end-to-end delay in scenarios with complex traffic patterns and wide range of node mobility, compared to OLSR.