RoLocMe: A Robust Multi-Agent Source Localization System with Learning-Based mAP Estimation

Abstract

This paper addresses the source localization problem by introducing RoLocMe, a multi-agent reinforcement learning system that integrates SkipNet - a skip-connection-based RSS estimation model - with parallel Q-learning. SkipNet predicts RSS propagation of the entire search region, enabling agents to explore efficiently. The agents leverage dueling DQN, value decomposition, and λ-returns to learn cooperative policies. RoLocMe converges faster and achieves at least 20% higher success rates than existing methods in dense and sparse reward settings. A drop-one ablation study confirms each component’s importance and RoLocMe’s effectiveness for larger teams.

Cite

Text

Le et al. "RoLocMe: A Robust Multi-Agent Source Localization System with Learning-Based mAP Estimation." International Joint Conference on Artificial Intelligence, 2025. doi:10.24963/IJCAI.2025/18

Markdown

[Le et al. "RoLocMe: A Robust Multi-Agent Source Localization System with Learning-Based mAP Estimation." International Joint Conference on Artificial Intelligence, 2025.](https://mlanthology.org/ijcai/2025/le2025ijcai-rolocme/) doi:10.24963/IJCAI.2025/18

BibTeX

@inproceedings{le2025ijcai-rolocme,
  title     = {{RoLocMe: A Robust Multi-Agent Source Localization System with Learning-Based mAP Estimation}},
  author    = {Le, Thanh Dat and Ye, Lyuzhou and Huang, Yan},
  booktitle = {International Joint Conference on Artificial Intelligence},
  year      = {2025},
  pages     = {152-160},
  doi       = {10.24963/IJCAI.2025/18},
  url       = {https://mlanthology.org/ijcai/2025/le2025ijcai-rolocme/}
}