ANN-Based Formulation of Path Loss Prediction for Radio Wave Propagation for Indoor Agriculture and Sensor Networks

Abstract

The wireless sensor network has proven to be a useful instrument for providing farmers with accurate data on the condition of their crops. Although the FSPL, 2-Ray, COST235, and linear path loss regression curve fit model (LRCFM) give an explanation for the propagation of 2.4 GHz radio waves through vegetation, several substantial inconsistencies were discovered when applied to field experiments with plants greenhouses. This study uses artificial neural networks (ANNs) to make a prediction model that can be used to look at how tree growth affects path loss across a wide range of transceiver heights and operating parameters. The artificial neural networks were created using the experimental data. The neural network is trained using as input parameters the height and distance of the antennas of the transmitting and receiving nodes and, as a desired parameter, the amount of path losses (PL). Using the network weights, a new PL prediction formula was created. This formula predicts the amount of path losses more accurately, and the mean absolute relative deviation (AAPD) between our formula and the FSPL+COST235, 2Ray+COST235, FSPL, 2Ray, and LRCFM correlations is 0.36%, 17%, 55.5%, 42.7%, 81.11%, and 5.454%, respectively.