Real-Time LaCAM

2025-04-08

AI Summary

Real-time multi-agent path finding (MAPF) faces two key challenges: excessive computational overhead of full-path planning and the lack of formal completeness guarantees in existing online methods—leading to deadlocks or livelocks. To address these, this paper proposes the first provably complete incremental real-time MAPF framework, built upon LaCAM (2023). It integrates incremental path refinement, millisecond-level truncated scheduling, and iterative constraint enforcement to ensure feasibility and completeness within each decision step. Experiments demonstrate that, under high congestion, the framework achieves success rates comparable to full-horizon LaCAM while maintaining sub-millisecond response times. Moreover, it natively supports integration with learned policies. The core contribution lies in establishing the first theoretical completeness guarantee for real-time MAPF, coupled with a scalable, implementation-ready paradigm that bridges formal verification and practical deployment.

Technology Category

• Multiagent Systems: Multiagent Planning
• Planning, Routing, and Scheduling: Learning for Planning and Scheduling
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

The vast majority of Multi-Agent Path Finding (MAPF) methods with completeness guarantees require planning full horizon paths. However, planning full horizon paths can take too long and be impractical in real-world applications. Instead, real-time planning and execution, which only allows the planner a finite amount of time before executing and replanning, is more practical for real world multi-agent systems. Several methods utilize real-time planning schemes but none are provably complete, which leads to livelock or deadlock. Our main contribution is to show the first Real-Time MAPF method with provable completeness guarantees. We do this by leveraging LaCAM (Okumura 2023) in an incremental fashion. Our results show how we can iteratively plan for congested environments with a cutoff time of milliseconds while still maintaining the same success rate as full horizon LaCAM. We also show how it can be used with a single-step learned MAPF policy. The proposed Real-Time LaCAM also provides us with a general mechanism for using iterative constraints for completeness in future real-time MAPF algorithms.

Problem

• Real-time MAPF with completeness guarantees for practical multi-agent systems
• Avoiding livelock or deadlock in real-time multi-agent path planning
• Incremental LaCAM for iterative planning in congested environments

Innovation

• Real-Time LaCAM with provable completeness guarantees
• Incremental planning for congested environments
• Single-step learned MAPF policy integration