Reliability-Margin KBGWO for Robust RFID Network Planning

Abstract

This brief paper presents a compact reliability-margin framework for robust radio-frequency identification (RFID) planning in healthcare environments. Deterministic RFID network planning can overestimate indoor healthcare coverage when walls and propagation variability perturb the effective link margin. A reliability-margin Knowledge-Based Grey Wolf Optimizer (KBGWO) is proposed for criticality-aware RFID network planning. Grid readability is estimated from sample-wise Monte Carlo read indicators under stochastic path loss, wall loss, and shadowing. The objective maximizes criticality-weighted reliable coverage and lower-tail reliability while penalizing redundant-read risk, cost, and shortfall from a reliable-coverage target. A small hospital-pilot readability validation compares measured ultra-high-frequency RFID (UHF-RFID) read probabilities with simulated probabilities over line-of-sight (LOS), one-wall, two-wall, metallic-obstructed, and service-core non-line-of-sight (NLOS) locations. The pilot gives mean absolute error (MAE) = 0.029, root mean square error (RMSE) = 0.031, and correlation $r=0.988$ . Simulations on a hospital-inspired grid show that KBGWO increases reliable coverage from 77.8% to 85.9% and lower-tail reliability from 71.3% to 91.3% versus deterministic KBGWO, at 29% higher cost. Additional checks include external reimplemented baselines, weight sensitivity, Monte Carlo sample stability, ablation, attenuation sensitivity, convergence behavior, perturbed layouts, and raw seed reporting.