Transformer-based tracking Network for Maneuvering Targets

For strong maneuvering targets, the drastic change of target motion models makes the tracking methods hard to adapt and provide accurate state estimation promptly. To solve the state estimation problem of strong maneuvering targets, we propose a transformer-based tracking network, named TrTNet. The TrTNet model learns a new residual mapping from the observation trajectory to the real trajectory. Through the parallel manner, the TrTNet model extracts single state features of the observation trajectory at all time steps simultaneously. The correlation information between states can be explored and utilized to enhance state features of single time step, benefiting to learn the transition law between states and generate high-quality tracking model. Simulation results show that the proposed TrTNet model achieves better performance in maneuvering target tracking than other competitive approaches.