A Hybrid Passive-to-Active SONAR Tracking Architecture Using Constant Acceleration EKF-TOMHT in Cluttered Environments

Abstract

Subsurface maneuvering target tracking under shallow-water hydroacoustic clutter presents severe operational challenges due to target dynamic lag, range unobservability during passive stealth surveillance, and combinatorial computational explosion during multi-hypothesis data association. This paper proposes an adaptive passive-to-active target tracking architecture that integrates a six-dimensional Constant-Acceleration (6D-CA) motion model driven by continuous white jerk noise with a track-oriented multiple hypothesis tracking extended Kalman filter engine. To resolve dynamic tracking lag during high-acceleration tactical evasive maneuvers, the target state space explicitly incorporates Cartesian acceleration states. Raw acoustic hydrophone signals digitized at 100 Hz undergo digital beamforming and CFAR thresholding to feed validated measurement centroids to the tracking filter at a sampling interval of Δt=1.0 s (1.0 Hz). Mode transition from passive stealth to active sonar interrogation is governed dynamically by an automated position covariance trace threshold. Monte Carlo simulations under uniform spatial Poisson clutter demonstrate that the proposed 6D CA-TOMHT framework maintains a mean position RMSE during the bearing-only passive phase (t∈[0,15] s), which drops dramatically to a steady-state active position RMSE of 7.08 m upon active sonar activation, restricting the maximum transient maneuver error to . With score-based hypothesis pruning and structural branch capping, the algorithm maintains an average active pool dimension of 4.97 branches and achieves a mean processing time, confirming real-time execution capabilities.