A GM-PHD Filter with Estimation of Probability of Detection and Survival for Individual Targets
R.A. Thivanka Perera, Mingi Jeong, Alberto Quattrini Li, Paolo Stegagno
Abstract
This paper proposes a modification of the Gaussian mixture probability hypothesis density (GM-PHD) filter to compute online the probability of detection (P_{D})(P_{D}) and probability of survival (P_{S})(P_{S}) of targets. This eliminates the need for predetermined and/or constant P_{D}P_{D} and P_{S}P_{S} values, that may degrade the estimation. The proposed filter estimates the P_{D}P_{D} and P_{S}P_{S} values for each individual target based on newly introduced parameters, which are updated during the measurement update process. The effectiveness of the proposed filter was validated through an in-lab experiment using four unmanned ground robots with varying P_{D}P_{D} values and a real-world lidar-based obstacle tracking system implemented on an Automated Surface Vehicle operating in a lake with real-time boat traffic. The results of the experiments demonstrate that the proposed filter outperforms the standard PHD filter with incorrect P_{D}P_{D} and P_{S}P_{S} values. These findings highlight the potential benefits of the proposed filter in improving target tracking performance in complex environments.
BibTeX
@inproceedings{iros2023_agmphdfilterwith,
title = {A GM-PHD Filter with Estimation of Probability of Detection and Survival for Individual Targets},
author = {R.A. Thivanka Perera and Mingi Jeong and Alberto Quattrini Li and Paolo Stegagno},
booktitle = {IROS 2023},
year = {2023}
}