Bounds on performance for multiple target tracking

A theoretical lower bound on mean-square-estimation error is derived for tracking in dense multiple-target environments. This family of bounds is computationally tractable, because it does not require computing the optimal estimate. Computational complexity can be traded for tightness of the lower bounds by varying the number of hypotheses considered. The theory can be used to study the fundamental limitations of tracking performance, as an alternative to pursuing an endless quest for better algorithms. However, for some applications, the use of the theory could show that there is substantial room for algorithmic improvement.<
>