AI-Based 3D Simulation for Drone Flight Dynamics

Abstract

Unmanned aerial vehicles (drones) have provided new potential in areas like surveillance, transportation, construction, and agriculture. Simulating drone dynamics is vital in these domains, as it allows researchers to test drones in complex or risky circumstances. However, evaluating drone behavior is complicated because to the various elements involved. Traditional models based on Newtonian and fluid dynamics use parameters including force, gravity, propeller characteristics, and air density. While these models can replicate a generic drone, they are not realistic for replicating the dynamics of a specific drone due to the complex nature of the parameters. An AI-based technique gives a simpler way to model drone dynamics compared to these older methods. This approach leverages advanced AI models trained on massive datasets from real-world flight events. The datasets cover a range of flight maneuvers, including figure-eight, circular, and lazy-eight patterns, illustrating several sorts of drone motions. Several methods were utilized to develop the models, including multi-output regression, support vector machines (SVM), neural networks (NN), and convolutional neural networks (CNN). The CNN model achieved the highest accuracy at 78%. To validate the models, anticipated drone shifts were compared with realworld flight data. Future work will focus on further refining the CNN-based model and integrating it with a virtual reality environment for improved simulation.