Deeplearning-based vehicular channel estimator in high mobility environments

Abstract

In recent years, deeplearning has almost invaded the world of telecom electronics and other fields, given the spectacular results it achieves in terms of improving the performance of digital processing chains. Wireless Access in Vehicle Environments (WAVE) technology has been developed, and IEEE 802.11p defines the Physical Layer (PHY) and Media Access Control (MAC) layer in the WAVE standard. However, the IEEE 802.11p frame structure, which has a low pilot density, makes it difficult to predict wireless channel properties in a vehicle environment with high vehicle speeds (high Doppler frequency), thus system performance are degraded in realistic vehicle environments. The motivation of this article is to improve channel estimation and tracking performance without modifying the IEEE 802.11p frame structure. Therefore, we propose a channel estimation technique based on deeplearning that can perform well over the entire range of SNR values, the effects of ISI and ICI interference remain inescapable phenomena. The improvement brought by the LS channel estimation methods, MMSE and linear equalizers, cubic spline, linear DFT ET and cubic spline DFT interpolation are reviewed, these interpolation techniques contribute to the reduction of the BER in the chain. The different vehicular channel environment scenarios are split, simulations of the new estimation DNN method are performed on examples of high mobility channels, and compared to the LS and MMSE methods. A strong immunity of the proposed estimator against the high mobility of the channels is observed.