Drone Localization Estimation Using Wireless Sensor Network and Deep-Learning Techniques

This paper presents a powerful hybrid localization model of drones to be used in GPS-denied areas through the combination of Long Short-Term Memory (LSTM) networks with an Extended Kalman Filter (EKF). The system network is based on a Wireless Sensor Network (WSN) in which sensors are arranged in equilateral triangle triples to enable the combination of the Angle-of-Arrival (AoA) and Time-Difference-of-Arrival (TDoA) measurements. The traditional EKF only model may not be good with complex residual dynamics as well as motion uncertainties but the LSTM component is specifically used to model these nonlinearities which gives much better state estimation compared to what conventional kinematic models can offer. The work presents a strict theoretical framework of triangular anchor geometry using Dilution of Precision (DOP) analysis and applying the nearest triple geometry to approximate localization using triangulation techniques when a drone goes into sensor field. Wide Monte Carlo simulation and statistical analysis with confidence intervals prove that the framework is very precise, that is, it can attain a distance error of less than 0.48 meters. Moreover, the suggested approach has shown a 42 percent reduction in Root Mean Square Error (RMSE) as compared to conventional EKF benchmarks. The contribution of each component is verified in detailed ablation studies, which prove that the system has the real-time computational performance required to deploy a drone in practice.