Joint Optimization in Pinching-Antenna-Assisted NOMA-MEC Systems via DRL

Abstract

This work investigates the integration of pinching antennas into latency-sensitive mobile edge computing (MEC) systems, where a hybrid non-orthogonal multiple access (NOMA) strategy is employed for task offloading to minimize energy consumption. An energy minimization problem is formulated to jointly optimize the pinching antenna position, transmit power, and time allocation. Due to the non-convexity caused by highly coupled optimization variables, an entropy-based deep reinforcement learning (DRL) approach is developed. Simulation results show that the proposed soft actor-critic (SAC) framework achieves near-optimal performance under dynamic channel conditions while maintaining stability and generalization. Meanwhile, with the integration of pinching antennas, hybrid NOMA outperforms other strategies in latency-critical scenarios. © 2026 IEEE.