An intelligent hierarchical recognition method for long-term orbital maneuvering intention of non-cooperative satellites

The difficult problem of recognizing the long-term non-cooperative orbital maneuvering intentions of gaming satellite is investigates in this paper. By converting the intention recognition task into a classification problem of multi-dimensional data sequences, a deep learning-based hierarchical recognition method with trajectory partitioning mechanism is proposed. The method begins by decomposing target trajectory through a partitioning module. It converting long-term orbit maneuvering trajectory into manageable units. An intention prediction model is then built based on the bidirectional long short-term memory network. Optimized for data feature extraction and sampling weight allocation layers, the model can accurately fit the mapping relation from data set to intention set. Finally, the method incorporates the fitting relations in trajectory data, intention units and long-term intentions, achieving the goal of transforming orbit data input to long-term intention label output. The proposed method overcomes the limitation that traditional approaches cannot recognize the long-term non-cooperative intentions of multiple maneuvering satellite. Experimental results show that the prediction model utilized enhances average recognition accuracy by 1.95% compared to conventional models. It can also maintain over 90% recognition success rate for the long-term non-cooperative intentions of close gaming targets with certain observational errors.