Learning-Based Hybrid Neural Receiver for 6G-V2X Communications

TL;DR

This paper introduces a hybrid neural receiver for 6G-V2X communications that replaces multiple physical layer blocks using Transformer and GNN, demonstrating superior performance across various scenarios and multimodal data.

Contribution

The paper presents a comprehensive hybrid neural receiver model that replaces entire physical layer blocks in wireless communication, a novel approach in the field.

Findings

Outperforms state-of-the-art neural receivers by ~0.5 dB

Effective across diverse vehicle speeds and channel conditions

Adaptable to multimodal data like images and radar

Abstract

Neural receiver models are proposed to jointly optimize multiple functionalities of wireless receivers; however, a comprehensive receiver model that replaces the entire physical layer blocks has not yet been presented in the literature. In this work, we introduce a novel hybrid neural receiver (H-NR) built on Transformer encoder blocks and Graph Neural Network (GNN), as part of an end-to-end wireless communication framework. In our communication framework, we assume vehicle to network (V2N) uplink scenario where information is transmitted by vehicle and received at the base station (BS). Our proposed H-NR model replace OFDM resource grid demapping, channel estimation, signal equalization, demodulation, and channel decoding. To test the adaptability of our proposed model on unseen conditions, we evaluate its performance for various scenarios, including a vehicle speed of range [0-60] km/h, a carrier frequency of 5.9GHz, and a cluster delay line (CDL) channel model. Furthermore, we assess the performance of our proposed H-NR on multimodal data, such as images, audio, GPS, radar, and LiDAR, to examine its adaptability in real-world use cases. The simulation results clearly demonstrate that our proposed model outperforms the state-of-the-art neural receiver by approximately 0.5 dB in terms of reconstruction and error correction.