Research on Collaborative Optimization of Near-Field ISAC Systems Based on Multi-Waveguide Pinching Antenna Arrays

In this paper, a near-field integrated sensing and communication (ISAC) using a multi-waveguide pinching antenna system (PASS) is proposed, where discrete radiating particles are flexibly clamped along multiple dielectric waveguides to realize reconfigurable near-field beam focusing. To accurately characterize the underlying propagation mechanism, a spherical-wave channel model is established by jointly accounting for free-space distance-dependent attenuation, waveguide propagation loss, and phase accumulation. Based on this model, a constrained joint design problem is formulated to simultaneously optimize pinching antenna (PA) positions, communication-and-sensing power allocation, and quantized phase-only beamforming under total power, minimum power, minimum inter-element spacing, and finite-resolution hardware constraints. The resulting problem is a high-dimensional nonconvex mixed-integer optimization problem. To address it, a collaborative alternating particle swarm optimization-genetic algorithm (PSO-GA) framework is developed, where PSO is employed to optimize continuous variables, including PA positions and power allocation, while GA is used to search over discrete beamforming phases. In addition, feasibility is maintained through projection and power normalization, and a physics-informed initialization strategy based on MRT and weighted ISAC beamformers is introduced to accelerate convergence. Furthermore, exact channel gradients, a closed-form Fisher information matrix (FIM), and corresponding Cramér-Rao bound (CRB) expressions are derived to analytically characterize geometry-dependent sensing performance. Simulation results for an eight-waveguide, 64-element PASS operating at 28 GHz with 500 MHz bandwidth demonstrate that the proposed design achieves about 6.1 Gbps communication rate while maintaining about 0.01 m localization accuracy. The results also show that the proposed method attains a favorable communication-sensing tradeoff, converges faster than standalone heuristics, and remains robust under discrete configurability, waveguide attenuation, and imperfect channel state information (CSI).