Meta-Heuristics Methods of Parametric Propagation Model tuning in Modern Cellular Network Planning and Optimization

Abstract

The transition to 5G and beyond has necessitated highly granular radio frequency (RF) planning. Parametric propagation models, such as the Cost-231 Hata or the ECC-33 model, rely heavily on accurate calibration for specific geographical environments. Traditional manual tuning-based on Least Squares Estimation (LSE) often suffers from local optima entrapment and inability to handle non-linear constraints. Meta- heuristic and direct- search methods are widely employed, yet systematic comparative studies that jointly evaluate convergence dynamics, computational efficiency, and parameter- estimation precision are scarce. This paper investigates the application of meta-heuristic algorithms, specifically the Particle Swarm Optimisation (PSO), Genetic Algorithms (GA), Pattern Search (PS), and Simulated Annealing (SA)in tuning these models. We analyze these methods across three critical dimensions: computational efficiency, speed of convergence, and root-mean-square error (RMSE) precision. Our findings indicate that while PSO offers superior convergence speed, PS provides a more refined balance between precision and computational overhead in complex, highdensity urban environments.