DQN-Based Transmit Power Control in V2V Communications Using Sensor Images

This paper proposes a way to optimize the transmit power of a vehicle based on the Deep-Q-Network (DQN) model with sensor images in a vehicle-to-everything (V2X) environment. The road situation is observed and represented as images, and then the transmit power of a vehicle is decided to meet the target packet reception ratio (PRR) via the DQN model. The proposed scheme solely depends on the sensor images unlike many previous works based on channel information. The performance is evaluated in terms of the learning speed and success rate. The proposed scheme can minimize transmit power while satisfying the target PRR in various V2X scenarios.