A Hybrid Neural Network-Based IE and IMM Architecture for Target Tracking

In order to enable a tracking system to work stably in the environment with fast maneuver and rapidly changing noise, a new hybrid architecture combining interacting multiple model (IMM) and neural network-based input estimate (IE) together is presented in this paper. In this architecture, IMM provides estimation of covariance of measurement noise to neural network-based IE, while IE enables the system to work effectively when the targets lead fast and complex maneuver, both of the outputs of IMM and NNIE will be fused in fusion module. In order to verify the effectiveness of this architecture, several simulations were leaded and the results prove it can work stably with rapidly changing noise and fast maneuver.