AI Algorithm for predicting and optimizing trajectory of UAV swarm

Abstract

This thesis explores the application of Artificial Intelligence (AI) techniques for generating the trajectories of fleets of Unmanned Aerial Vehicles (UAVs). The two main challenges addressed include accurately predicting the paths of UAVs and efficiently avoiding collisions between them, which we discuss in the two paragraphs below, respectively. In all the previous studies that predicted the path, a Feedforward Neural Network (FFNN) with a single hidden layer and standard activation functions like Sigmoid, Tanh, and ReLU was used. These activation functions resulted in high errors (Mean Squared Error or MSE and Root MSE or RMSE) in the predicted path. In this work, we apply a non-standard set of activation functions, including Swish and Elliott, and also propose our new activation function, AdaptoSwelliGauss. AdaptoSwelliGauss is a sophisticated fusion of Swish and Elliott activations, seamlessly integrated with a scaled and shifted Gaussian component. This dynamic combination is specifically designed to excel in capturing the complexities of UAV trajectory prediction. The accuracy obtained with our new activation function is better by three to four orders of magnitude as compared to the standard activation functions.