Research on Trajectory Tracking Algorithm Based on LSTM-UKF

Abstract

Aiming at the problems of excessive error and inability to track the traditional target tracking algorithm in the absence of observations, a trajectory tracking model combined with Long Short-Term Memory (LSTM) is designed. Combining the LSTM network model with the Unscented Kalman Filter (UKF), using the autonomous learning and memory characteristics of the LSTM network, provide the UKF algorithm with the predicted value of the observations, and optimize the UKF algorithm for the target object in the absence of the observations. Tracking effect. Finally, the verification and analysis are carried out for three different sports conditions. The simulation results show that the LSTM-UKF algorithm model still has a good tracking effect even in the absence of observations.