Target positioning and tracking in degenerate geometry

This paper analyzes the effects of special target-sensor geometries, particularly, some degenerate cases (near singular observation matrix), on the performance of target positioning and tracking. A scenario of practical significance is when two or more sensors form nearly parallel line of sight (LOS) vectors to targets. Examples include netted radars that happen to line up their observations in azimuth. Such encounter geometry is degenerate for conventional methods. In this paper, we first analyze conventional estimates in such degenerate geometry and then describe three adapted methods, namely, a geometric 1D solution for 2D scenarios, a reduced-order least squares method, and a subspace least squares method. Simulation results are presented to compare the methods. The analysis provides the basis for active sensor management performance prediction and optimization via sensor placement/ assignment.