This paper deals with the design and implementation of an algorithm fo
r track formation and maintenance in a multisensor Air Traffic Surveil
lance scenario. The major contribution of the present work is the deve
lopment of the combined likelihood function that enables the replaceme
nt of the Kalman filter (KF) with the much more versatile interacting
multiple model (IMM) estimator which, as a self adjusting variable-ban
dwidth state estimator, accounts for the various motion modes of the a
ircraft. This likelihood function defines the objective function used
in the measurement to track assignment algorithm. Also, this algorithm
incorporates both skin and beacon returns, i.e., it fuses the primary
and secondary radar data. Data from two FAA radars are used to evalua
te the performance of this algorithm. The use of the IMM estimator yie
lds considerable noise reduction during uniform motion, while maintain
ing accuracy of the state estimates during maneuver. Overall, the mean
square prediction error (to the next observation time) is reduced by
a factor of three over the KF. The usefulness of the tracker presented
here is also demonstrated on a noncooperative target.