Should I Replan? Learning to Spot the Right Time in Robust MAPF Execution

2026-04-28

AI Summary

This work addresses the challenge of execution-time delays in multi-agent path finding (MAPF), which often disrupt synchronization and compromise plan validity. While existing robust approaches ensure safety, they incur excessive waiting costs, and naive replanning may be ineffective or even increase overhead. To overcome these limitations, this study proposes the first learning-based mechanism for deciding when to replan: it constructs state features using an action dependency graph (ADG) and employs a fully connected feedforward neural network to predict the expected benefit of replanning, thereby intelligently triggering replanning only when advantageous. Evaluated on a newly curated large-scale dataset comprising 12,000 experimental instances, the method reduces the impact of potential delays by up to 94.6%, significantly outperforming baseline strategies that either always or never replan.

Abstract

During the execution of Multi-Agent Path Finding (MAPF) plans in real-life applications, the MAPF assumption that the fleet's movement is perfectly synchronized does not apply. Since one or more of the agents may become delayed due to internal or external factors, it is often necessary to use a robust execution method to avoid collisions caused by desynchronization. Robust execution methods - such as the Action Dependency Graph (ADG) - synchronize the execution of risky actions, but often at the expense of increased plan execution cost, because it may require some agents to wait for the delayed agents. In such cases, the execution's cost can be reduced while still preserving safety by finding a new plan either by rescheduling (reordering the agents at crossroads) or the more general replanning capable of finding new paths. However, these operations may be costly, and the new plan may not even lead to lower execution cost than the original plan: for example, the two plans may be the exact same. Therefore, we estimate the benefit that can be achieved by single replanning in scenarios with delayed agents given an immediate state of the execution with a fully connected feed-forward neural network. The input to the neural network is a set of newly designed ADG-based features describing the robust execution's state and the impact of potential delays, and the output is an estimated benefit achievable by replanning. We train and test the network on a new labeled dataset containing 12,000 experiments, and we show that our proposed method is capable of reducing the impact of delays by up to 94.6% of the achievable reduction.

Problem

• Multi-Agent Path Finding
• Robust Execution
• Replanning
• Action Dependency Graph
• Execution Cost

Innovation

• Robust MAPF Execution
• Replanning Decision
• Action Dependency Graph (ADG)
• Neural Network Prediction
• Delay-aware Path Planning