Robust Estimation of VANET Performance-Based Robust Neural Networks Learning

Vehicular ad hoc network (VANET) can manage live traffic and send emergency messages to the base station in any smart city and is emerging as a connectivity network. In VANET, every vehicle acts as a sensor node, which collects the surrounding information and sends information to the base station. VANET network is created when communication between cars with wireless transceiver is needed. Despite the fact that VANET and mobile ad hoc network (MANET) have some similarities; the dynamic nature of VANET has posed a challenge on routing protocols designing; VANET is composed of models based communication among vehicles and vehicle with a high mobility feature. Presently the artificial neural networks is often used in several fields. Neural networks are usually trained by conventional backpropagation learning algorithm that minimizes the training data mean square error (MSE). The goal of this paper is to investigate VANET performance in terms of packet loss rate and throughput using robust neural networks learning based on the robust M-Estimators performance function instead of the traditional MSE performance function. Robust M-estimators performance functions outperform the traditional MSE performance function in terms of RMSE and training speed as simulation results show.