Anonymous Multi-Agent Path Finding with Individual Deadlines

Gilad Fine, Dor Atzmon, Noa Agmon

Research output: Contribution to journalConference articlepeer-review

Abstract

Anonymous Multi-Agent Path Finding (AMAPF) is the problem of planning conflict-free paths to a set of target locations for a group of agents, where each agent is not associated with a specific target. This paper studies AMAPF with Individual Deadlines (AMAPFwID), where each target must be reached before a specific deadline, and the agent stays there to perform some long-term mission, for instance securing an asset, responding to an emergency, or performing a maintenance job. We examine three types of behavior of agents when reaching a target: (a) disappear on target, (b) stay on target, and (c) move after the deadline. The latter is only possible if the agent is replaced by another agent. We refer to this replacement as hot swapping. We propose a solution to AMAPFwID with each type of such behavior, based on a reduction to Network Flow. We test all solutions experimentally and show cases where hot swapping is beneficial. Finally, we also provide a solution to the case where agents disappear on targets that maximizes the number of targets reached and discuss other aspects of the problem.

Original languageEnglish
Pages (from-to)869-877
Number of pages9
JournalProceedings of the International Joint Conference on Autonomous Agents and Multiagent Systems, AAMAS
Volume2023-May
StatePublished - 1 Jan 2023
Externally publishedYes
Event22nd International Conference on Autonomous Agents and Multiagent Systems, AAMAS 2023 - London, United Kingdom
Duration: 29 May 20232 Jun 2023

Keywords

  • MAPF
  • Multiagent Systems
  • Network Flow
  • Path Planning

ASJC Scopus subject areas

  • Artificial Intelligence
  • Software
  • Control and Systems Engineering

Fingerprint

Dive into the research topics of 'Anonymous Multi-Agent Path Finding with Individual Deadlines'. Together they form a unique fingerprint.

Cite this