End-to-End GPS Tracker Based on Switchable Fuzzy Normalization Codec for Assistive Drone Application

As one of the consumer electronics, drones have received more and more attention in applications such as dynamic monitoring, transportation of goods, and unmanned logistics. In these applications, it is necessary to obtain the accurate position of the drone and can track and navigate. The drones usually work outdoors, and GPS signals are the most important information to obtain their position. Therefore, GPS-based positioning is a key research issue for drone applications. On the other hand, this puts forward higher requirements for GPS-based mobile target tracking and positioning technology. The GPS tracker can obtain real-time dynamics location through state estimation. Classical estimation methods require a system model with Gaussian white noise, whereas GPS data usually contains pink noise, making an exact match to the actual system challenging. This study uses a new end-to-end GPS location tracker implemented through a data-driven mechanism incorporating a codec to catch complex nonlinear dynamics. Further, the switchable fuzzy normalization is loaded in the codec, using three different normalization algorithms, such as z-score, to adaptively process the input data, to realize the measurement data’s normalization and adaptive correction and extract the dynamic features of GPS. We experimentally conclude that compared with the classical deep learning model, the method proposed in this paper reduces the RMSE by an average of 7.95% and 20.6%, respectively, compared with the optimal model, avoids the modelling process of the system, can efficiently overcome the chromatic noise and dynamics in GPS observation, and outperforms the trajectory estimation performance of classical filtering methods.