An Intelligent Multi-Drone Navigation-Based Trajectory Prediction and Classification Framework Using a Hybrid Recurrent Neural Network Model

Unmanned Aerial Vehicles (UAVs), commonly referred to as drones, have drawn attention from various fields. Navigation is a core module of the drone guidance system, along with control and navigation methods, which are vital subsystems of the drones. Numerous drone navigation technologies have been proposed, comprising GPS, reference frames, and models. Additionally, computer vision models have been effectively verified and improved in real-world applications. One of the most noticeable advances is the application of computer vision in drones, which provides the technology for visual navigation and obstacle avoidance, enabling unmanned control. Recent developments and the advances of Deep Learning (DL)-based solutions have also made the navigation of autonomous vehicles more feasible. This study presents an Intelligent Multi-Drone Navigation-Based Trajectory Prediction and Classification System (IMDN-TPCS) model, in which data pre-processing is performed using min-max normalization. This is followed by the employment of a Bidirectional Gated Recurrent Unit with an attention mechanism (BiGRU-Attn) method for an accurate trajectory prediction. Finally, the Deep Belief Network (DBN) method is utilized for classification. The comparative study of the IMDN-TPCS method demonstrated a superior accuracy of 98.80% compared to other models on the UAV Autonomous Navigation dataset.