Adaptive Pinching Antenna Optimization via Meta-Learning for Physical-Layer Security in Dynamic Wireless Networks

This paper develops a gradient-based meta-learning framework for real-time control of waveguided pinching-antenna systems under user-location uncertainty and physical-layer security (PLS) constraints. A probabilistic system model is introduced to capture the impact of imperfect localization on outage performance and secrecy. Based on this model, a joint antenna-positioning and transmit-power optimization problem is formulated to satisfy probabilistic reliability and secrecy requirements. To enable rapid adaptation in highly dynamic environments, the proposed approach employs model-agnostic meta-learning (MAML) to learn a transferable initialization across diverse mobility and channel conditions, allowing few-shot online adaptation using limited pilot feedback. Simulation results demonstrate that the proposed framework significantly outperforms Reptile-based meta-learning, non-meta reinforcement learning, conventional optimization, static antenna placement, and power-only control in terms of outage probability, secrecy performance, and convergence latency. These results establish meta-learning as an effective tool for secure and low-latency control of reconfigurable pinching-antenna systems in non-stationary wireless environments.