Deep Neural Network based Channel Estimation in VANET

In recent years, vehicular communication has gained huge interest among industry and researchers for combating traffic overload and high risk of accidents. A major challenge in these communication systems is to handle the distortions and channel variations caused due to high mobile environment. Several conventional data pilot assisted (CDPA) channel estimation such as least square (LS), spectral temporal averaging (STA), constructed data pilot (CDP), etc. as well as deep learning based estimation schemes have been evolved for addressing the same. In this paper channel estimation scheme for vehicular communication system using autoencoder (AE) has been proposed. The proposed technique is implemented using leaky rectified linear unit (Leaky ReLU) activation function as compared to the existing scheme which has used ReLU function. The performance of the proposed scheme has been compared with CDPA and also with deep learning channel estimation schemes on the basis of bit error rate (BER) and mean square error (MSE). The result shows that the proposed scheme performs better than the CDPA channel estimation schemes even in the lower signal-to-noise ratio (SNR) regime. In addition comparison between the existing deep learning channel estimation scheme and proposed scheme shows that deep learning based on leaky ReLU gives better performance than ReLU function based estimation.