Using Neural Network and Levenberg–Marquardt Algorithm for Link Adaptation Strategy in Vehicular Ad Hoc Network

Abstract

Vehicular Ad Hoc Network (VANET) was initiated about two decades ago in view of saving lives by mitigating and reducing the number of accidents and incidents on public roads. Moreover, this objective can only be achieved if VANETs mobiles regularly exchange Road State Information (RSI) with their neighborhood and take decisive actions based on the RSI received. Therefore, it becomes paramount to ensure that the transmitted message is well received. And this is only possible if the quality of the sharing medium or link is controlled, and transmission performed while taking into consideration the Channel State Information (CSI). The CSI provides information related to channel quality, Signal-to-Noise Ratio (SNR), and so forth. The process of adapting the payload as a function of the CSI is called Link Adaptation (LA). Several LA works have already been published in VANETs, but almost without serious consideration of the effect of the relative mobility amongst the nodes. Hence, while taking into consideration the Doppler Shift induced by the relative velocity, the current work presents a link adaptation strategy using a Neural Network (NN) and the Levenberg-Marquardt algorithm in VANETs. The simulation results definitively demonstrate that the NN approach outperforms its counterparts by a significant margin. It achieves a performance of 1075% in transmission duration, 180% in transmitted bit, and 115% in model efficiency when compared to the Cte, ARF, and AMC algorithms, respectively.

Keywords

• VANET
• machine learning
• link adaptation
• WAVE
• V2V
• V2I