基于 DBSCAN 聚类和 LSTM 网络的装甲车辆集群轨迹预测方法

It is difficult to accurately predict the movement trajectory of armored vehicle cluster due to the complexity of armored vehicle motion states, the uncertainty of battlefield situations, and the tactical confusion and deception. This paper proposes a trajectory prediction method for armored vehicle cluster based on density-based spatial clustering of applications with noise (DBSCAN) clustering algorithm and long short-term memory (LSTM) network. Firstly, a kinematic model of armored vehicles is established based on the states of armored vehicles driving on slopes, turning and interacting with other vehicles. And then the trajectory characteristics such as maneuver features, environmental features and vehicle-to-vehicle interaction features are selected, and the trajectory of an individual armored vehicle is predicted using a dual-layer LSTM network. Finally, the DBSCAN algorithm is utilized to segment the multiple single-vehicle predicted trajectories, calculate the similarities among them, and cluster them to obtain a representative trajectory for the cluster, as the predicted trajectory for the armored vehicle cluster. Simulated results demonstrate that the proposed method can effectively predict the trajectories of armored vehicle clusters.