A Real-Time Trajectory Prediction Method of Small-Scale Quadrotors Based on GPS Data and Neural Network

Abstract

This paper proposes a real-time trajectory prediction method for quadrotors based on a bidirectional gated recurrent unit model. Historical trajectory data of ten types of quadrotors were obtained. The bidirectional gated recurrent units were constructed and utilized to learn the historic data. The prediction results were compared with the traditional gated recurrent unit method to test its prediction performance. The efficiency of the proposed algorithm was investigated by comparing the training loss and training time. The results over the testing datasets showed that the proposed model produced better prediction results than the baseline models for all scenarios of the testing datasets. It was also found that the proposed model can converge to a stable state faster than the traditional gated recurrent unit model. Moreover, various types of training samples were applied and compared. With the same randomly selected test datasets, the performance of the prediction model can be improved by selecting the historical trajectory samples of the quadrotors close to the weight or volume of the target quadrotor for training. In addition, the performance of stable trajectory samples is significantly better than that with unstable trajectory segments with a frequent change of speed and direction with large angles.

Short-Term 4D Trajectory Prediction for UAV Based on Spatio-Temporal Trajectory Clustering

“…Huang [34] proposed an effective error calculation and mitigation method to establish a more accurate UAV trajectory prediction model by evaluating the performance of trajectory prediction accuracy. Yang [8] proposed a real-time trajectory prediction method for quadrotors based on a bidirectional gated recurrent unit model. The model produced better prediction results than the baseline models for all scenarios of the testing datasets.…”

“…The input of each prediction model is set as the trajectory parameter vector of two adjacent time spans. The prediction result is the trajectory parameter vector of the next single step [8], which is: The prediction mode (bootstrap) based on the prediction data as the model input is adopted to predict the trajectory parameters after k time spans in the future, which is: Equation ( 6) represents the iterative process of prediction. The process uses the adjacent timestamp trajectory parameter vectors Pt-2, Pt-1 to predict the trajectory parameter Pt+k of the t+k timestamp.…”

“…The way to capture the complex motion patterns of UAVs becomes the key when predicting the trajectories. Based on machine learning methods (e.g., neural networks), UAV trajectory data collected from airborne global positioning system (GPS) sensors have been applied to solve the problem in recent years [8]. However, few studies attempt to classify and understand the behavior characteristics of UAVs to build more targeted prediction models.…”

A classification method for urban functional regions based on the transfer rate of empty cars

“…PA(1) to PA(4) represent the exact area of the four triangles. Equation (2) represents the distance between two points, which are i and j. Lat i represents the latitude value of point i, and Long i represents the longitude value of point i. Equations ( 11)- (14) represent PA (1), PA (2), PA (3), and PA(4), respectively. The PA shown in Equation ( 15) is the final area of the hexagon.…”

“…We use half of the product of the longest diagonal and the shortest diagonal to approximate the area of the hexagon. It is worth FIGURE 3 The structure of AGRU mentioning that the latitude and longitude coordinates should be standardized before participating in the computation process, and the method of standardization is Z-score standardization. We define the area of a hexagon obtained by this method as the approximate standardized area for convenience.…”