SPLINE ESTIMATION OF PATHS USING BEARINGS-ONLY TRACKING DATA

In many applications, bearings are measured to a moving object with th e goal of estimating the object's course of movement. If movement is a ppropriately modeled as a smooth deterministic curve in the plane, the n a cubic spline is a reasonable representation of the curve. Maximum likelihood estimators are presented for parameters of regression splin es, assuming that observation errors follow a Von Mises distribution. Location estimates are obtainable even when data are sparse. Path esti mation error, number and placement of knots, and outlier detection are discussed. Examples, including both simulated paths and observations from a wildlife radio-tracking study, are presented.