Analysing and Predicting the maneuver behaviors of aircraft with Box-Jenkins ARIMA model

The maneuver behavior analysis and prediction tend to be a complex research problem of situation awareness in aerial combat. Some of the recent studies on maneuver decisionmaking in deep learning have reported that few milliseconds operating time in advanced is a challenging issue. Most of these studies are based on supervised learning, the network learns values not only from large sets of data, but also the data of several future time steps to get the deep model trained. However, insufficient data will affect the decision performance in actual combat environment. This study aims at utilizing the Box-Jenkins autoregressive integrated moving average (ARIMA) model for maneuver behaviors' parameters prediction of aerial vehicles on radar observation data with less amount of data requirements. The minimum value of AIC/BIC score is obtained through the series correlation of ACF and PACF after the optimal ARIMA model parameters’ selection. Residuals diagnostic is performed to verify the model for efficient parameter prediction. Experiment results indicate that the method could offer an effective approach to analyse and predict the maneuver behaviors on radar data with a high accuracy.