ESER- Machine Learning Model for Predicting RSSI of FSO Communications in Forest Environments

This study investigates the application of machine learning (ML) algorithms to predict the received signal strength indicator (RSSI) in free space optical (FSO) communication systems operating in forest areas. FSO technology offers high data transmission rates and bandwidths, but faces challenges due to atmospheric conditions, especially in dynamic environments such as forests. Conventional empirical models often fail to accurately estimate RSSI due to the complex, non-linear nature of these environments. To address this problem, we collected a comprehensive dataset over a 12-month period, covering various atmospheric parameters such as temperature, humidity and wind speed. We evaluated several ML models - K-Nearest Neighbors (KNN), Decision Trees (DT), Random Forest (RF), Gradient Boosting Regression (GBR) and Artificial Neural Networks (ANN). In addition, we have developed a new ML model that combines RF and GBR methods and provides the best results, called “Enhanced Stacked Ensemble Regressor” (ESER) - to determine its effectiveness in predicting RSSI. The results show that the ESER model has the lowest root mean square error (RMSE) and a high coefficient of determination (R2), demonstrating its robustness and accuracy. This study highlights the potential of ML techniques to improve the reliability of FSO links in challenging environments. It highlights directions for future work, including real-time prediction and adaptive FSO systems.