Channel estimation enhancement in vehicular communication using deep neural network

The concept of Vehicular ad-hoc network (VANET) can significantly improve the safety and traffic management of the vehicular communication. It is considered to be favorable technology for intelligent transportation systems (ITS). However, the fundamental challenge is to address the channel distortions and variation due to high mobile environment. To resolve this issue, an optimization technique that assimilates deep learning into the existing IEEE 802.11p systems is implemented in this work. Deep neural network (DNN) is purely a data-based technique which serves better where channel modelling is difficult. The channel estimation is realized employing DNN which tracks the channel characteristics efficiently. DNN is first trained offline as per channel conditions and finally trained network reconstructs the transmitted data according to the input in testing or online stage. The goal here is to compensate for the error introduced in the pilot-aided channel estimation schemes. This is done by appropriate selection of hyper-parameters of DNN which surges the capacity of the network considerably. The proposed work is compared with existing pilot-aided conventional methods and deep learning-based estimation techniques according to bit error rate (BER) performance. The simulation results demonstrate that the propounded method is more superior to the earlier channel estimation schemes. The proposed method is examined deeply in multiple scenarios to test the strength.