Naval Undersea Warfare Center Division Newport , Rhode Island THE a ~ FILTER IN DRAG WITH DATA ASSOCIATION

The well-known a(3 filter achieves its best performance when measurement outliers are rare. Two modifications of the a(3 filter are presented for improving its performance when outliers are commonplace. The a f3 J-l filter uses a drag coefficient to ensure that the target motion model has a finite asymptotic velocity variance. Although the af3J-l filter is less sensitive to outliers than the a (3 filter , it does not directly address the problem of large spurious measurements. The a (3 J-lCY filter is a probabilistically based, nonlinear target estimation filter that is much more robust against outliers than the a(3 filter. To facilitate understanding of the filter optimization problem and to enable computation of filter outputs crucial to the post-processing task of associating collections of tracking filters, it is recommended that the a(3 filter , together with all the important earlier work that has made it useful in applications, be reinterpreted in the Kalman filter paradigm. ADMINISTRATIVE INFORMATION This memorandum was prepared under the sponsorship of the Naval Sea Systems Command (NAVSEA). Funding was from the Submarine Tactical Passive Advanced Development Project (ASTO, Robert E. Zarnich) Program Element 63504N, NUWC Job Order No. A19606. The principal investigator was David Pistacchio (Code 2123). The technical reviewer for this memorandum was Tod E. Luginbuhl.