Real-time radar tracking system with deep learning

Real-time data tracking plays an essential role in the flight test processes of an aircraft. The data flowing from the aircraft to the ground control center must be real-time and uninterrupted. However, sometimes ground control systems can cause disconnection with the aircraft, making it difficult to track them. This thesis firstly gives a brief survey of real-time aircraft tracking systems and then proposes a deep learning-based, real-time 3D prediction of the next location named DeepAT for uninterrupted real-time data tracking. Our DeepAT model uses an Encoder-Decoder GRU model to predict the next location of the aircraft. Thus, in case of any disconnection, the tracking of the aircraft can be sustainable. In the experiments, real flight test sensor data collected with the telemetry system are used. Experimental analyzes are performed for two structurally different aircraft, one of which is a highly maneuverable fixed-wing propeller aircraft and the other an Unmanned Air Vehicle (UAV). The efficiency and superiority of the proposed method is demonstrated by comparing it with state-of-the-art methods in terms of Mean Absolute Error (MAE) and Mean Squared Error (MSE) metrics. The results show that our proposed method outperforms the state-of-the-art and gives better prediction of the next location of aircraft.