<title>MHT tracking for crossing sonar targets</title>

Abstract

Sometimes radar targets cross and become unresolved; this is a concern, but with a reasonable track depth and an appropriate merged-measurement model the concern is considerably mitigated. Sonar targets, however, can become merged (in the same beam) for considerably longer, particularly with bearing-only measurements. In such cases the crossing times can be 100 scans long, and no reasonable depth exists for an multi-frame tracker that can "see" both ends of the merged period. Further, there is a demonstrable tendency for estimated targets to repel each other as they are being tracked. In this paper we explore the hypothesis-oriented multi-hypothesis tracker (HO-MHT), an MHT approach that uses the new "rollout" optimization insight and the to give an appropriate and cost-effective means to rank hypotheses, and also the PMHT tracker that operates on batches of scans with linear computational complexity in most quantities. We show results in terms of estimation error (RMSE), consistency (NEES) and computational effort in both linear and beam-space tracking scenarios.

Further analysis of the track repulsion effect in automatic tracking

“…Equations (2)(3)(4)(5) provide predicted bounds on target position and velocity at time t. Assume that a measurement is available at time t, based on the following linear measurement model:…”

“…Recently, it has been observed that an opposite effect exists with trackers that utilize hard data association [3], which we denote as the track repulsion effect; this effect leads to track swapping in difficult target-swapping scenarios. In [4], multi-hypothesis tracking (MHT) technology is applied to the crossing-target problem.…”

An equivalence-class approach to multiple-hypothesis tracking

“…We denote by n Z k1 the smaller measurement from the n th sensor at time k t , and the larger by n Z k1 . Equations (11)(12)14) still apply. Equations (8)(9)(10)13) are replaced by the following.…”

“…It has been shown that the PDAF exhibits track coalescence [12][13], while the GNNF and MHT exhibit track repulsion [14][15][16]. Several approaches have been investigated to mitigate the impact of track coalescence in soft data association filters, see e.g.…”

An Efficient Implementation of a Batch-Oriented, Multitarget, Multidimensional Assignment Tracking Algorithm with Application to Passive Sonar

“…During the crossing, if the contacts are not of approximately equal signal-to-noise ratio (SNR), detections will be assigned to the higher SNR contact, and the lower SNR track will be lost. For contacts that have equal probability of detection (P D ), Willett et al [1] have recently shown that the estimated tracks generated with a standard assignment model exhibit a repulsion behavior during the crossing that violates the expected behavior of these large contacts, i.e., that they are known to travel at constant course and speed and cannot make sudden changes because of their size. This repulsion behavior causes problems downstream for algorithms that require state estimates to be consistent with contact trajectory.…”