Literature by the same author
plus at Google Scholar

Bibliografische Daten exportieren
 

Spatially Grouped Curriculum Learning for Multi-Agent Path Finding

Title data

Phan, Thomy ; Koenig, Sven:
Spatially Grouped Curriculum Learning for Multi-Agent Path Finding.
In: Proceedings of the AAAI Conference on Artificial Intelligence. Vol. 40 (2026) Issue 35 . - pp. 29642-29650.
ISSN 2159-5399
DOI: https://doi.org/10.1609/aaai.v40i35.40208

Official URL: Volltext

Project information

Project title:
Project's official title
Project's id
AI Research Institute for Advances in Optimization
2112533
Causal Foundations for Decision Making and Learning
2321786

Project financing: National Science Foundation
Amazon Robotics
Donald Bren Foundation

Abstract in another language

Multi-agent path finding (MAPF) is the challenging problem of finding conflict-free paths with minimal costs for multiple agents. While traditional MAPF solvers are centralized using heuristic search, reinforcement learning (RL) is becoming increasingly popular due to its potential to learn decentralized and generalizing policies. RL-based MAPF must cope with spatial coordination, which is often addressed by combining independent training with ad hoc measures like replanning and communication. Such ad hoc measures often complicate the approach and require knowledge beyond the actual accessible information in RL, such as the full map occupation or broadcast communication channels, which limits generalizability, effectiveness, and sample efficiency. In this paper, we propose Partitioned Attention-based Reverse Curricula for Enhanced Learning (PARCEL), considering a bounding region for each agent. PARCEL trains all agents with overlapping regions jointly via self-attention to avoid potential conflicts. By employing a reverse curriculum, where the bounding regions grow as the policies improve, all agents will eventually merge into a single coordinated group. We evaluate PARCEL in two simple coordination tasks and four MAPF benchmark maps. Compared with state-of-the-art RL-based MAPF methods, PARCEL demonstrates better effectiveness and sample efficiency without ad hoc measures.

Further data

Item Type: Article in a journal
Refereed: Yes
Institutions of the University: Faculties
Faculties > Faculty of Mathematics, Physics und Computer Science
Faculties > Faculty of Mathematics, Physics und Computer Science > Department of Computer Science
Faculties > Faculty of Mathematics, Physics und Computer Science > Department of Computer Science > Junior Professor Artificial Intelligence and Machine Learning
Faculties > Faculty of Mathematics, Physics und Computer Science > Department of Computer Science > Junior Professor Artificial Intelligence and Machine Learning > Junior Professor Artificial Intelligence and Machine Learning - Juniorprof. Dr. Thomy Phan
Research Institutions > Central research institutes > Research Center for AI in Science and Society
Result of work at the UBT: Yes
DDC Subjects: 000 Computer Science, information, general works > 004 Computer science
Date Deposited: 04 May 2026 06:57
Last Modified: 30 Jul 2026 06:21
URI: https://eref.uni-bayreuth.de/id/eprint/96963