.png){kind=icon}
Return
Approximate message passing algorithm for decentralised task assignment and scheduling
.png){kind=icon}
PRE
AI
DOI：10.1080/00207721.2025.2602076.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This paper proposes a decentralised algorithm, called task assignment and scheduling via approximate message passing (TAS-AMP), to address the task assignment and scheduling (TAS) problem in multi-agent systems. Approximate message passing (AMP) is a distributed algorithm developed for vehicle routing problems and is based on belief propagation in graphical models. Leveraging the framework of AMP while addressing its convergence limitations, TAS-AMP rapidly generates near-optimal task assignments and execution schedules by iteratively exchanging local messages among agents. To improve message convergence and solution quality, TAS-AMP introduces two key mechanisms: a pruning process and a conflict resolution phase. The pruning process refines each agent's schedule from the previous iteration using updated message values. This prevents unnecessary schedule reinitialization and thereby improves message convergence and solution stability. The conflict resolution phase reduces unassigned tasks and removes redundant assignments, ensuring a conflict-free solution. An ablation study and convergence analysis were conducted on various TAS-AMP configurations to validate the effectiveness of these mechanisms. Furthermore, numerical comparisons across diverse TAS instances demonstrated that TAS-AMP achieves enhanced solution quality and computational efficiency even under high reward heterogeneity.
Keywords:
Task assignment and scheduling
belief propagation
approximate message passing
multi-agent system
decentralised task planning
Journal
I
IF:
4.6
Papers:
1.1K
Citations:
7.3K