Deep Multi-modal Neural Receiver for 6G Vehicular Communication

Deep Learning (DL) based neural receiver models are used to jointly optimize PHY of baseline receiver for cellular vehicle to everything (C-V2X) system in next generation (6G) communication, however, there has been no exploration of how varying training parameters affect the model's efficiency. Additionally, a comprehensive evaluation of its performance on multi-modal data remains largely unexplored. To address this, we propose a neural receiver designed to optimize Bit Error Rate (BER) for vehicle to network (V2N) uplink scenario in 6G network. We train multiple neural receivers by changing its trainable parameters and use the best fit model as proposition for large scale deployment. Our proposed neural receiver gets signal in frequency domain at the base station (BS) as input and generates optimal log likelihood ratio (LLR) at the output. It estimates the channel based on the received signal, equalizes and demodulates the higher order modulated signal. Later, to evaluate multi-modality of the proposed model, we test it across diverse V2X data flows (e.g., image, video, gps, lidar cloud points and radar detection signal). Results from simulation clearly indicates that our proposed multi-modal neural receiver outperforms state-of-the-art receiver architectures by achieving high performance at low Signal to Noise Ratio (SNR).