WBGM-IC: Joint uplink–downlink resource optimization for many-to-many D2D sharing in 6G networks

Abstract

Device-to-device (D2D) communication is envisioned to support distributed artificial intelligence in 6G networks by enabling data sharing and collaborative learning among terminals. To mitigate co-channel interference introduced by D2D communication, a complex “many-to-many” D2D resource-sharing scenario within cellular networks is investigated. We propose a joint uplink–downlink resource allocation mechanism based on weighted bipartite graph matching and interference clustering to optimize resource coordination. In the considered resource-sharing model, a single channel can be shared by various D2D users (DUs), and each DU is allowed to access multiple channels concurrently. A joint optimization framework for uplink–downlink channel assignment and power control is formulated, comprising two stages. In the first stage, a weighted bipartite graph matching-based resource allocation (WBGM-RA) algorithm is employed to allocate channels to cellular users (CUs) to maximize the system sum rate. In the second stage, an interference clustering-based resource allocation (IC-RA) algorithm is developed, where an interference matrix is constructed to represent inter-user interference relationships. Based on this, the transmit power of DUs is optimized while ensuring that the communication quality of CUs is not compromised. Experimental results demonstrate that, under the condition that the data rates of CUs and DUs are at least 2 bps/Hz, the proposed scheme significantly outperforms JUDRA and JUAD in terms of system sum rate, number of supported communication links, and DU channel access ratio—achieving at least 27% and 31% gain in sum rate, 62% and 112% increase in links support, and $8.7\text{\%}\sim 21.5\text{\%}$ and $13.6\text{\%}\sim 122\text{\%}$ improvement in DU access ratio, respectively. This work may serve as a preliminary step toward more efficient spectrum utilization and enhanced system capacity.