We introduce the Probabilistic Multi-Hypothesis Tracking (PMHT) algorithm f
or particle tracking in high-energy physics detectors. This algorithm has b
een developed recently for tracking multiple targets in clutter, and it is
based on maximum likelihood estimation with help of the EM algorithm. The r
esulting algorithm basically consists of running several iterated and coupl
ed Kalman filters and smoothers in parallel. It is similar to the Elastic A
rms algorithm, but it possesses the additional feature of being able to tak
e process noise into account, as for instance multiple Coulomb scattering.
Herein, we review its basic properties and derive a generalized version of
the algorithm by including a deterministic annealing scheme. Further develo
pments of the algorithm in order to improve the performance are also discus
sed. In particular, we propose to modify the hit-to-track assignment probab
ilities in order to obtain competition between hits in the same detector la
yer. Finally, we present results of an implementation of the algorithm on s
imulated tracks from the ATLAS Inner Detector Transition Radiation Tracker
(TRT). (C) 1999 Elsevier Science B.V. All rights reserved.