D2D- and NOMA-Enhanced Multi-Cell RAN Slicing for Maximizing Service Capacity Under Security Isolation Constraints

Abstract

In radio access network (RAN) slicing, achieving both high service capacity and strong security isolation is critically important but inherently conflicting. While fine-grained orthogonal resource allocation can balance this tradeoff for small-scale networks, it fundamentally fails to support the rapid increase in the number of users, particularly for massive machine-type communications (mMTC). Although non-orthogonal multiple access (NOMA) has been widely adopted to mitigate this by improving spectral efficiency, existing approaches largely overlook the massive potential of exploiting the heterogeneous capabilities of users to further expand service capacity. To address this, we propose a device-to-device (D2D)- and NOMA-enabled multi-cell RAN slicing scheme. Specifically, by leveraging user heterogeneity, our scheme enables capability-rich enhanced mobile broadband (eMBB) users to serve as D2D relays for power-constrained mMTC users, effectively expanding spatial access dimensions. We formulate a resource optimization problem aimed at maximizing the system satisfaction ratio (reflecting service capacity), subject to pre-set system isolation ratio constraints for security and per-slice satisfaction constraints for fairness. Given the complex co-channel interference in multi-cell environments, we design a multi-agent deep reinforcement learning-based subchannel allocation solution to solve the problem. Simulation results demonstrate that the proposed scheme not only enhances slice service capacity under slice security constraints, but also effectively reduces both the energy consumption of mMTC users and the overall system energy consumption.