Differential Evolution for Optimizing Parameter Estimation in Practical D2D Channels

Inspired by the theory of natural evolution, Differential Evolution (DE) functions constitute an optimization tool in the field of evolutionary algorithms. Here, we propose employing DE to estimate physically acceptable fading parameters in device-to-device (D2D) communication channels. We examine four real-world D2D propagation channel measurements obtained under various conditions: indoor, outdoor, line-of-sight (LOS), and non-LOS. Four popular fading models,
$\kappa-\mu, \eta-\mu, \kappa-\mu$
/inverse gamma, and
$\eta-\mu$
/inverse gamma are used to characterize these links. Two fitness functions, Kullback-Leibler divergence (KLD) and mean squared error (MSE), are utilized for evaluation. We also compare the DE with another evolutionary algorithm, namely the genetic algorithm (GA). Notably, our results demonstrate that while both algorithms deliver excellent estimation performances, DE emerges as significantly faster and more robust compared to GA. Regarding fitness performances, the algorithm, when paired with KLD, outperforms the pairing with MSE, as assessed through the minimization of the Akaike information criterion.