Multisensor--Multitarget Bearing--Only Sensor Registration

TL;DR

This paper introduces a novel multisensor approach for modeling and estimating offset biases in bearing-only sensors, improving multitarget tracking accuracy by compensating for sensor biases and deriving theoretical bounds.

Contribution

It proposes a maximum likelihood bias estimator for bearing-only sensors, handling time-varying targets and sensor biases, with derivation of the Cramér-Rao Lower Bound for accuracy assessment.

Findings

The algorithm effectively estimates sensor biases in simulated scenarios.

The method improves tracking accuracy with bias compensation.

Simulation results demonstrate robustness even with false alarms and missed detections.

Abstract

Bearing--only estimation is one of the fundamental and challenging problems in target tracking. As in the case of radar tracking, the presence of offset or position biases can exacerbate the challenges in bearing--only estimation. Modeling various sensor biases is not a trivial task and not much has been done in the literature specifically for bearing--only tracking. This paper addresses the modeling of offset biases in bearing--only sensors and the ensuing multitarget tracking with bias compensation. Bias estimation is handled at the fusion node to which individual sensors report their local tracks in the form of associated measurement reports (AMR) or angle-only tracks. The modeling is based on a multisensor approach that can effectively handle a time--varying number of targets in the surveillance region. The proposed algorithm leads to a maximum likelihood bias estimator. The corresponding Cram\'er--Rao Lower Bound to quantify the theoretical accuracy that can be achieved by the proposed method or any other algorithm is also derived. Finally, simulation results on different distributed tracking scenarios are presented to demonstrate the capabilities of the proposed approach. In order to show that the proposed method can work even with false alarms and missed detections, simulation results on a centralized tracking scenario where the local sensors send all their measurements (not AMRs or local tracks) are also presented.