Optimization of Wireless Sensor Network Node Placement in Oil Palm Plantations using Particle Swarm Optimization for Improved Path Loss Prediction

This study proposes a data-driven framework for optimizing Wireless Sensor Network (WSN) node placement in oil palm plantations on Sebatik Island, Indonesia, by integrating site-specific path loss modeling with a modified Particle Swarm Optimization (PSO) algorithm. Field measurements across 120 transmitter-receiver pairs at 433 MHz revealed that conventional log-distance models poorly predict signal attenuation in dense vegetation (R² < 0.35, RMSE > 8 dB), while calibrated quadratic and cubic polynomial models achieved high accuracy (R² up to 0.9857, RMSE as low as 1.12 dB). These empirical models were embedded into the PSO fitness function to optimize spatial deployment of 20 nodes over a 500 m × 800 m area. The optimized layout achieved 94.7% coverage, 98% connectivity, 42% energy savings over random placement, and 95.6% Packet Delivery Ratio (PDR). Validation against independent field data confirmed robust prediction accuracy (RMSE = 4.3 dB), significantly outperforming generic models like ITU-R. This approach demonstrates that vegetation-aware, empirically calibrated modeling combined with metaheuristic optimization substantially enhances WSN performance in tropical agro-forestry environments, offering a scalable solution for smart agriculture in remote, ecologically complex regions.