AI/ML for D2D Communication in 5G/6G Networks and Beyond: Concept, Applications, Challenges, and Future Directions

Device-to-Device (D2D) communication has emerged as a pivotal technology for fifth-generation (5G) and sixth-generation (6G) wireless networks, enabling direct communication between proximate devices while bypassing the base station. This paradigm promises significant improvements in spectral efficiency, energy efficiency, system capacity, and ultra-low latency communications essential for emerging applications including Vehicle-to-Everything (V2X), Industrial IoT, and extended reality. However, realizing the full potential of D2D communication requires addressing numerous technical challenges including device discovery, mode selection, interference management, power control, security, radio resource allocation, cell densification and offloading, quality of service provisioning, millimeter-wave integration, and seamless handover. The complexity and dynamic nature of these challenges have catalyzed extensive research into Artificial Intelligence (AI) and Machine Learning (ML) solutions. This survey provides a structured, multi-dimensional review of AI/ML approaches for D2D communication across classical, advanced, and emerging paradigms. We organize the literature into three broad generations: classical techniques (e.g., Fuzzy Logic, Q-Learning, Neural Networks, GA/PSO/ACO), advanced paradigms (e.g., DRL, GNNs, FL, MARL), and emerging technologies (e.g., LLMs, Generative AI, Semantic Communication, RIS). Rather than cataloguing methods in isolation, we synthesize the literature through cross-cutting analytical perspectives including challenge coverage, performance and complexity trade-offs, convergence and scalability trends, autonomy readiness, and research maturity across more than 170 references, including recent works from 2023 to 2025 relevant to 6G-oriented D2D systems. Performance figures are reported as indicative ranges synthesized from heterogeneous studies and are not used for direct cross-paper ranking without normalization to common experimental conditions. We also discuss the DAI/BDIx framework as one possible path toward autonomous distributed D2D control and note that initial ADROIT6G testbed validation has now been reported for BDIx-enabled control blocks in a beyond-5G stand-alone environment, while recent cluster-based cell-free D2D/6G results further broaden the supporting empirical base. Finally, we identify twelve open issues and outline a roadmap toward intelligent, distributed, and deployment-aware D2D systems in 6G networks.