No-model Tracking by Data-driven Method using BP Networks

In the model-driven tracking approach, the tracking performance is mainly based on the system model. But the accurate model is difficult to obtain, while data-driven tracking approach can't depend on the accuracy of system model and it can tracking by the measurement data. This paper use BP net to develop a data-driven tracking. The BP net is trained by system measurement and prediction get by Kalman filter and the tracking as the output data. The simulation has shown that the data-driven tracking by BP net can obtain good performance even when the moving target is different.