Covariance-based assignment can fail when a large tracking innovation has more than one plausible cause. A target maneuver should increase target-side process uncertainty, whereas a degraded sensor should reduce only the value of assignments that use that physical sensor. This paper presents a source-aware Bayesian estimation-to-assignment framework for heterogeneous unmanned aerial vehicle (UAV) tracking. Three binary factors represent maneuver uncertainty, range degradation, and bearing degradation, and their combinations generate eight joint hypotheses in an interacting multiple-model extended Kalman filter (IMM-EKF). The joint posterior is converted into factor marginals and fused state moments. An uncertainty-owner interface then maps target-side and sensor-side evidence to different covariance terms before constrained assignment. Paired Monte Carlo experiments with 100 paired holdout seeds show that source-aware estimation supplies the main tracking gain, while explicit sensor-risk action gives a smaller scenario-dependent improvement and direct maneuver gating adds no reliable benefit under fixed UAV trajectories. The framework therefore clarifies not only which source is plausible, but also where that information should act in the closed loop.
 

sensor-to-target assignment,sensor degradation,maneuvering target,interacting multiple model,heterogeneous multi-UAV tracking