Enhanced Target Movement Prediction

A new method based on neural networks and advanced computational intelligence techniques is presented in this proposal for predicting target movement. Due to lack of learning capabilities, current approaches cannot incorporated target behavior information into decision process to generate a reliable prediction. The new method uses two cascaded neural networks to learn the behavior pattern of targets from their historical movement. The future movement of targets is then predicted by these two neural networks based on their recent movement and actions of correlated targets. To measure the reliability of the prediction, a reasoning network is finally used to generate a belief interval for each prediction.