Research on Prediction Method of UAV Heat Seeking Navigation Control Based on GRU Networks

Abstract

AbstractThis paper establishes the thermal updraft prediction model based on GRU (Gated Recurrent Unit) network and complete the network training. GRU network is a simplified evolution version of LSTM network. The thermal updraft predictor based on Gru network is designed, and the network structure parameters and network training parameters are given, focusing on the differences with LSTM network. The experiment of this paper is given, which mainly compares the prediction accuracy and running speed of two kinds of predictors under the straight-line flight mode of UAV between LSTM network and GRU network. The experimental results show that in the task of thermal updraft prediction, GRU network can obtain the prediction accuracy similar to LSTM network, but it can reduce a lot of time. Compared with LSTM network, GRU network is more suitable for engineering application.KeywordsUAVThermal updraft predictionLSTMGRU