Comparison of Deep Learning-Based Channel Estimator and Classical Estimators in VANET

A network-based deep learning (DL) channel estimation algorithm is proposed for selective fading channels using a deep neural network (DNN). The DNN can not only exploit the channel variable functions from previous channel estimates, but it can also use least squares estimation to further improve channel estimation performance. Before the dynamic channel can be tracked online, simulated offline data is used to train the DNN. Its purpose is to refine the initial parameters of the DNN over several training sessions cycles. In this paper, we compare old channel estimation methods LS, MMSE, LINEAIR, DFT, SPLINE CUBIC SPLINE with the new method based on DNN for different environmental scenarios.