Font Size: a A A

Research On Path Prediction Algorithm Of UAV Based On Recurrent Neural Network

Posted on:2021-05-25Degree:MasterType:Thesis
Country:ChinaCandidate:L Y LiFull Text:PDF
GTID:2392330629452717Subject:Software engineering
Abstract/Summary:PDF Full Text Request
The wide application of unmanned aerial vehicle(uav)has been an important topic in military and civil fields in recent years.Since the uav came into being,it has played an important role in modern war and various fields due to its low cost,convenient use and easy operation.With the rapid development of uav technology,the research on the application system of uav is also going on.In the current world,the military powers led by the United States have successively established a comprehensive uav system to cope with the enhancement of their military strength.Path prediction of uav is an important part of it.How to accurately predict the path of uav within a certain time window has become a research hotspot in the field of intelligent control.Based on the motion process of uav as the research object,this paper studies the path prediction algorithm in uav flight by combining the recurrent neural network with heuristic threshold,and proposes some algorithms for path prediction.In this paper,the modeling of uav flight process is studied.Generally speaking,the flight process of uav is regarded as a system with dynamic change of continuous variables.Some operational means of differential calculus are used to deal with the trajectory,and the differential equation of non-homogeneous variable coefficient is used to derive the differential expression satisfying the current path.However,the fitting effect of differential equation is very complicated.The uav is affected by many factors in the actual flight process,and sometimes it cannot effectively fit the real trajectory.It is easier to mine the internal characteristics of data by means of modeling,and the effectiveness of the method is proved by experiments.In this paper,according to the characteristics of data sequences in terms of timing,neural network is introduced into the cycle path prediction to above,using the circulation in the neural network time series the above advantages.The preprocessing of uav motion track points according to time and to track the above points of design and features are extracted,the circulation of the neural network training samples,and neural network to carry out a series of improvement to the traditional cycle,it is concluded that,in accordance with the sample from the training models.At the same time,the training model is used to obtain the prediction points and predict the position information of the position points in a certain time window in the future.Simulation experiments are conducted to verify the effectiveness of recurrent neural network for path prediction.The experiments prove that the recurrent neural network has a good effect on path prediction of uav.In this paper,aiming at the deficiencies of the established threshold in the path prediction,the heuristic threshold method is adopted to judge the data,and the corresponding logistic regression training sample is obtained.The training sample set was trained to obtain the corresponding model.Based on the model,the size sequence of the weight of the path points used for processing on the flight path was obtained,and the points on the path prediction were effectively screened to effectively avoid the problem of overfitting on the path prediction.The theoretical feasibility is verified by simulation experiments and good results are obtained.Combined with all the aspects discussed above,this paper proposes a path prediction algorithm for uav,which can effectively predict the path of uav under a certain time window.The effectiveness of the algorithm is proved by simulation experiments and good experimental results are obtained.At the end of the paper,the contents and results of the research are summarized,and show what work can be done in future.
Keywords/Search Tags:Unmanned Aerial Vehicle, Recurrent Neural Networks, Heuristics, Thresholds, Trajectory Prediction
PDF Full Text Request
Related items