Rocket Sensor Data Prediction Based on LSTM with Attention Mechanism
Guorui Liao, Wei Yang, Kun Wang, Ruimeng He, Tao You · 2021
Accurately predicting the sensor data of the rocket in the future for a period of time from historical rocket sensor data is of great significance to make corresponding control adjustments in time and prevent major danger. Traditional linear models are difficult to solve multi-variable or multi-input problems, while neural networks such as LSTM are skilled at handling multivariable problems. This feature makes it helpful to solve time series prediction problems. This paper proposes a method for rocket sensor data prediction, which uses the attention mechanism and recurrent neural network to predict. It is verified that the average absolute error of this method for rocket sensor data prediction is only 1.38%, and it has strong applicability and accuracy in predicting the rocket attitude.