Advanced Data Association Algorithms to Address Emerging Threats

At the core of the challenge facing MDA today is one of adversary air and missile threats being able to fly non-ballistic, highly maneuvering hypersonic trajectories all while having the capability of close formation flight of multiple missile threats. This problem results in challenges related to detection of the threat targets as well as the resulting filtering and data association problem. There is a tight coupling between the filtering and data association aspect of the problem. At its core, associating new sensor measurements to existing tracks requires (1) finding all possible associations, (2) scoring these possible associations, and (3) selecting the best associations. The number of possible associations can be reduced by both accurate prediction of the target state to the current time of the sensor measurements or use of signature measurements from wide band radar or EO sensors for association. For this effort, we are proposing a two-pronged solution to address the advanced data association problem for emerging threats. The first is to learn target destination, trajectory, and maneuver tactics to augment physics-based models to predict target state for association to measurements; and the second is to learn possible association between signature measurements and existing tracks. Approved for Public Release | 20-MDA-10398 (2 Mar 20)