Anti-Interference Filter Generator Based on MGDA-UB Strategy for V2X

Abstract

In V2X communication, a large number of vehicles and sensor devices communicate with each other, and they adopt competitive allocation strategy in the resource pool. However, the high PAPR of signals in OFDM systems can lead to undesirable spectrum expansion when power amplifiers are used, thereby resulting in an increase in ACLR. Traditional filter design is difficult to simultaneously optimize the EVM and ACLR, which is deteriorated by the subcarrier of narrowband data located at the edge. In this paper, the deep learning is introduced into anti-interference filter design, and the method based on Swin-Transformer is used to optimize EVM, ACLR, PAPR and other indicators under the multi-task learning framework. Balance the task weights in the multi-task learning process through MGDAUB weight optimization strategy. The experimental results show that the trained filter generator can meet the requirements of multi-task targets. This paper provides a promising solution to the problem of Filter design in V2X communication.