BiLSTM-Attention: An Air Target Tactical Intention Recognition Model

Abstract

In the traditional air target tactical intention recognition process, reasoning analysis is only based on a single moment, but in the actual battlefield, the target tactical intention is realized by a series of actions, so the target state presents the characteristics of dynamic and temporal changes. In order to solve this problem, the bi-directional(Bidirectional) propagation mechanism and attention(Attention) mechanism are introduced on the basis of long-term and short-term memory network (LSTM), and a target tactical intention recognition model based on BiLSTM-Attention is proposed. The hierarchical method is used to construct the feature set of air combat intention and encode it into time series feature. The decision maker’s experience is encapsulated into a label. The deep information in the feature vector of air combat intention is learned by BiLSTM neural network, and the network weight is adaptively assigned by attention mechanism. Finally, the air combat feature information with different weights is put into Softmax function layer for intention recognition. Through the comparison with the traditional air tactical target intention recognition model and the analysis of ablation experimental, the proposed model effectively improves the recognition efficiency of air target tactical intention. It has important theoretical significance and reference value for auxiliary combat system.

Keywords

• |intention recognition|attention mechanism|bi-directional long-term and short-term memory network|air target|temporal feature