Analytic combinatorics in multiple object tracking
Roy L. Streit · 2017
The method of analytic combinatorics (AC) is a unified approach to multiple object tracking that encodes joint probability distributions into probability generating functionals (PGFLs). PGFLs characterize distributions exactly. A high level view of the tracking applications of PGFLs is outlined in this paper. Assignment models in well-known filters are modeled as products of PGFLs. MHT and multiBernoulli PGFLs are compared. Track extraction and the “notched” filter of the (reduced) Palm process are discussed. Bounded complexity approximate particle filter weights are found by saddle point methods applied to the Cauchy integral form of the derivatives.