Radar Maneuvering Target Tracking Based on LSTM Network

The nonlinear maneuvering target tracking problem is a state estimation problem in the case of system model mutation. The traditional multiple models method based on model switching has the practical problem of model mismatch, and the statistical accuracy is also limited. In this paper, a tracking scheme based on recurrent neural network structure is proposed. The implementation of this scheme is to extract conditional probability relations from a large number of training data through LSTM network, and apply it to continuous observation data, and finally get the state estimation results. Simulation results show that, compared with other common methods, this method can obtain more stable and accurate estimation effect in a shorter time, and is more anti-sensitive to target maneuvering.