Bearing-Only Active Sensing Under Merged Measurements
Luke Calkins, Philip D. Baldoni, James McMahon, Corbin Wilhelmi, Michael M. Zavlanos
Abstract
In this letter we propose an algorithm to actively track multiple moving targets using a bearing-only sensor in the presence of merged measurements. Merged measurements arise from sensor resolution constraints and therefore targets that are close in relative bearing to the sensor get reported as a single group measurement. We employ a merged measurement model in a nonlinear joint probabilistic data association filter for tracking multiple targets through merging events. We also propose an online adaptive planning algorithm that maneuvers the sensor in order to increase tracking performance. We introduce a novel method based on forward value iteration that incorporates the merged measurement information into the planning strategy. The resulting trajectory is biased away from situations where merged measurements occur, as this leads to more uncertainty in the target state estimates. We demonstrate our algorithm both in simulation as well as onboard real unmanned ground vehicles. This is the first time bearing-only tracking with merged measurements has been accomplished with a mobile sensor in practice. Furthermore, to the best of our knowledge, this is the first time merged measurement data association information has been utilized to effectively plan for such situations.
BibTeX
@inproceedings{ral2021_bearingonlyactiv,
title = {Bearing-Only Active Sensing Under Merged Measurements},
author = {Luke Calkins and Philip D. Baldoni and James McMahon and Corbin Wilhelmi and Michael M. Zavlanos},
booktitle = {RA-L 2021},
year = {2021}
}