Spectral Efficiency Maximization for V2V Communication Underlaid Cellular Uplink Using Deep Neural Networks

Abstract

Vehicle-to-vehicle (V2V) communication has been considered as a key technology of the intelligent transportation system because it has emerged with significant benefits such as improving driver safety and reducing traffic congestion and accidents. Although the V2V technology has provided some key advantages, the challenge still exists. Since V2V communication enables the transceiver pairs to exchange emergency information in the same cellular frequency band, the interferences of V2V links and vehicle-to-cellular (V2C) links should occur. Therefore, in our study, we tackle the interference problem by optimizing the transmit powers of the V2V users and the cellular users. The problem-solving process begins with formulating the optimization problem with linear constraints, where the objective function is the sum of data rates, and the transmit powers of all transmitters are the control variables. Then, we design a proper deep neural network (DNN) to solve the optimization problem. DNN obtains the optimal solution via training the neural networks in a way to minimize the loss function. The simulation results show that the proposed DNN algorithm is better than those of weighted minimum mean squared error (WMMSE), fixed transmit power, and Dinkelbach’s methods, and particularly achieves near-global optimum with lower computation complexity than the exhaustive search (ES).