Optimizing UAV Trajectories with Multi-Layer Artificial Neural Networks

Abstract

As Unmanned Aerial Vehicles (UAVs) are considered an essential part in many applications in life due to their cost-effectiveness and flexibility, they are facing many challenges. One of these challenges is predicting and optimizing their flight paths in dynamic environments. Although the traditional methods are reliable, but their effectiveness is lacking, which needs advanced methods to overcome the challenges. This study explored using Artificial Neural Networks (ANNs) to improve UAV trajectory prediction and optimization, focusing on flight time, UAV speed, and altitude. A high-level neural network written in Python was used to model multi-hidden layers of ANN. For this study, two datasets were divided into training and testing sets in 80%-20% and 70%-30% ratios, respectively. A 10-fold cross-validation was conducted to provide a more generalized view of the model’s performance. Statistical metrics were used to evaluate the performance of predictive model, includes Coefficient of Determination (R²), Mean Absolute Error (MAE), and Root Mean Square Error (RMSE). The results show that with an R² of 99.45% and MAE of 0.158, the model showed strong performance in distance prediction, though altitude predictions lagged with an R² of 53.95% and MAE of 15.2