AFFTC has a compelling need for the development of an advanced Kalman filtering technology to enable real-time tracking of multiple targets in highly dynamic environments. The key to achieve this goal is to augment the State-of-the-art (SOA) Kalman filter (KF) with an adaptive learning capability such that it can accommodate the tracking complexity imposed by the turbulent environment, dynamic 2-D motion detection and quantification. The on-line learning capability will also eliminate the need for time-consuming manual selection of tracking parameters. Etonnet proposes to develop an innovative Hybrid Neural Network Augmented Kalman Filter (HNN-KF) technique that will utilize an Etonnet proprietary Radial Basis Function Neural Network (RBFNN) algorithm and integrate it into KF to provide high-speed self-learning capability for adaptive feedback to the KF to minimize the tracking error. Etonnet will develop a HNN-KF software tool and demonstrate multiple targets tracking capability.
Keywords: Kalman Filter, Neural Network, On-Line Adaptive Learning, Multiple Target Tracking
