Track Editors

Daniel Harabor, Monash University
Sven Koenig, University of California, Irvine

Overview

Multi-Agent Path Finding (MAPF) is the abstract combinatorial problem of computing collision-free movement plans for a team of cooperative agents. The ability to solve instances of MAPF, efficiently and effectively, is a key enabler for many current and emerging industrial applications. These include warehouse logistics, train planning, pipe routing, robotic manufacturing, and many others. 

MAPF is studied by researchers and practitioners from across all areas of Artificial Intelligence, including Planning, Discrete Optimisation, Heuristic Search, Machine Learning, and Robotics. Topics of interest include but are not limited to: 

• Theoretical foundations and complexity analyses.
  
• Symbolic approaches for solving MAPF, including search algorithms, compilation and reduction methods as well as reactive and rule-based techniques.
  
• Learning approaches for solving MAPF, including supervised, unsupervised, and reinforcement techniques.
  
• Generalizations of MAPF, including agent kino-dynamics, limited communication, and delivery deadlines.
  
• Lifelong MAPF, including online planning, task allocation, execution monitoring, and explainability.
  
• Execution considerations, including action failures, agent delays, dynamic map changes, and other sources of uncertainty.
  
• Thoughtful critiques, meta-analyses, and surveys of the subject area.
  
• Empirical analyses and MAPF benchmarks.
• Real-world applications.

Call for Submissions

We invited researchers and practitioners to submit novel, original, and significant works, on all aspects of MAPF, to the JAIR special track. 

Submissions must be original, meaning that they have not previously appeared in the archival proceedings of any scholarly conference or journal. Papers that have only appeared at workshops or only as extended abstracts of no more than 2 pages are considered original. Substantially extended and improved versions of archival conference papers may also be considered original, provided that authors include appropriate citations to the conference paper and explain how their new submission extends and improves upon the earlier work.

Novelty and significance can be established by the artificial intelligence techniques themselves, their analysis, their experimental evaluations (including via comparisons to existing techniques), and by their application to important industrial settings. Novelty and significance can also be demonstrated via thoughtful critiques of the area. Surveys and meta-analyses in particular should enrich the body of scholarly work being discussed. 

Status

The track is closed for new submissions. Accepted submissions will be added to this page on publication.

Contents

Articles in the special track will be listed below upon publication.