A Deep Learning-Assisted Multi-Relay DCSK Communication System

Abstract

This paper proposes a novel multi-relay deep learning-assisted differential chaos shift keying (MR-DL-DCSK) communication system to enhance the capabilities of existing chaos-based cooperative communication systems. Channel quality significantly affects transmission reliability. However, existing channel quality evaluation methods require channel state information (CSI). To address this limitation, a deep neural network (DNN) classifier is employed at the receiver in this paper to perform joint channel quality assessment and symbol demodulation. We propose a channel quality-aware relay coordination strategy: at the relay stage, all relays assess their channel qualities using the DNN-output probability distribution, and relays with lower channel quality align their decoded bits with the bits from the relay with the highest channel quality before forwarding; at the destination stage, the destination selects the signal with the highest channel quality probability for final demodulation. This joint detection approach enables reliable demodulation without requiring explicit CSI, while the channel quality-aware relay coordination mechanism ensures that signals from the most reliable links are prioritized for final decision. Comprehensive simulation results demonstrate that the proposed multi-relay DL-DCSK system achieves superior bit error rate performance. Furthermore, the system exhibits excellent generalization capability when tested on vehicle-to-vehicle (V2V) communication channels modeled by the double-generalized gamma distribution, validating its practical applicability in diverse wireless environments.