Continual Multi-Robot Learning from Black-Box Visual Place Recognition Models

Kenta Tsukahara, Kanji Tanaka, Daiki Iwata, Jonathan Tay Yu Liang, Wuhao Xie

2025

Abstract

In the context of visual place recognition (VPR), continual learning (CL) techniques offer significant potential for avoiding catastrophic forgetting when learning new places. However, existing CL methods often focus on knowledge transfer from a known model to a new one, overlooking the existence of unknown black-box models. This study explores a novel multi-robot CL approach that enables knowledge transfer from black-box VPR models (teachers), such as those of local robots encountered by traveler robots (students) in unknown environments. Specifically, we introduce Membership Inference Attack (MIA), a privacy attack applicable to black-box models, and leverage it to reconstruct pseudo training sets, which serve as the transferable knowledge between robots. Furthermore, we address the low sampling efficiency of MIA by leveraging prior insights from the literature on place class prediction distributions and unseen-class detection. Finally, we analyze both the individual and combined effects of these techniques.

Paper Citation

in Harvard Style

Tsukahara K., Tanaka K., Iwata D., Liang J. and Xie W. (2025). Continual Multi-Robot Learning from Black-Box Visual Place Recognition Models. In Proceedings of the 22nd International Conference on Informatics in Control, Automation and Robotics - Volume 1: ICINCO; ISBN 978-989-758-770-2, SciTePress, pages 126-135. DOI: 10.5220/0013771900003982

in Bibtex Style

@conference{icinco25,
author={Kenta Tsukahara and Kanji Tanaka and Daiki Iwata and Jonathan Liang and Wuhao Xie},
title={Continual Multi-Robot Learning from Black-Box Visual Place Recognition Models},
booktitle={Proceedings of the 22nd International Conference on Informatics in Control, Automation and Robotics - Volume 1: ICINCO},
year={2025},
pages={126-135},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013771900003982},
isbn={978-989-758-770-2},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 22nd International Conference on Informatics in Control, Automation and Robotics - Volume 1: ICINCO
TI - Continual Multi-Robot Learning from Black-Box Visual Place Recognition Models
SN - 978-989-758-770-2
AU - Tsukahara K.
AU - Tanaka K.
AU - Iwata D.
AU - Liang J.
AU - Xie W.
PY - 2025
SP - 126
EP - 135
DO - 10.5220/0013771900003982
PB - SciTePress