Simulated Annealing Based Channel Estimation for Pinching-Antenna Systems: A Comprehensive Performance Analysis

Pinching-Antenna Systems (PASS) enable dynamic antenna reconfiguration for 6G networks through dielectric particle attachment but create challenging channel estimation problems due to sequential switching and ill-conditioned diagonal measurement matrices. This paper presents a simulated annealing (SA) approach achieving competitive estimation performance without requiring channel statistics or calibration data. The proposed phase-aware algorithm employs Cartesian coordinates together with an adaptive cooling process to maintain stable and reliable convergence. According to simulation results, the SA method attains an NMSE of about 3.37 dB, which is very close to the 3.15 dB achieved by regularized least squares (RLS) under the same test settings, and roughly corresponds to 85% of the performance level of LMMSE, even though it does not require any knowledge of noise variance or covariance. The obtained performance is clearly influenced by the constraint boundaries: when tight limits are used (for instance, in the range [-0.5, 0.5]), the NMSE reaches -0.78 dB, whereas for looser limits ([-3.0, 3.0]) it increases to around 4.70 dB. Additional tests over different SNR values and array sizes (N ranging from 8 to 64) confirm that the algorithm scales consistently and remain highly robust at low SNR levels. Overall, these findings indicate that SA offers a practical approach with reduced statistical dependence, requiring neither channel covariance nor noise variance information, and is suitable for dynamic PASS systems, particularly in scenarios where rapid structural reconfiguration prevents the use of any stable statistical modeling.