Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-Agent Corridor Generating Algorithm

Arseniy Pertzovsky, Roni Stern, Roie Zivan, Ariel Felner

TL;DR

MAPF requires conflict-free routes for $n$ agents to fixed goals on a graph. The authors introduce MACGA, which builds corridors along optimal single-agent routes and evacuates blocking agents, and MACGA+PIBT, which integrates PIBT for improved runtime and solution quality. The methods are polynomial-time, incomplete solvers with a reachability guarantee, and they achieve higher success rates on dense and maze-like MAPF benchmarks than several baselines, with MACGA+PIBT often reducing makespans. This work advances scalable, suboptimal MAPF planning in dense environments and provides a practical framework for hybrid rule-based and planner-based coordination.

Abstract

In this paper, we propose the Multi-Agent Corridor Generating Algorithm (MACGA) for solving the Multi-agent Pathfinding (MAPF) problem, where a group of agents need to find non-colliding paths to their target locations. Existing approaches struggle to solve dense MAPF instances. In MACGA, the agents build \emph{corridors}, which are sequences of connected vertices, from current locations towards agents' goals, and evacuate other agents out of the corridors to avoid collisions and deadlocks. We also present the MACGA+PIBT algorithm, which integrates the well-known rule-based PIBT algorithm into MACGA to improve runtime and solution quality. The proposed algorithms run in polynomial time and have a reachability property, i.e., every agent is guaranteed to reach its goal location at some point. We demonstrate experimentally that MACGA and MACGA+PIBT outperform baseline algorithms in terms of success rate, runtime, and makespan across diverse MAPF benchmark grids.

Multi-Agent Corridor Generating Algorithm

TL;DR

MAPF requires conflict-free routes for agents to fixed goals on a graph. The authors introduce MACGA, which builds corridors along optimal single-agent routes and evacuates blocking agents, and MACGA+PIBT, which integrates PIBT for improved runtime and solution quality. The methods are polynomial-time, incomplete solvers with a reachability guarantee, and they achieve higher success rates on dense and maze-like MAPF benchmarks than several baselines, with MACGA+PIBT often reducing makespans. This work advances scalable, suboptimal MAPF planning in dense environments and provides a practical framework for hybrid rule-based and planner-based coordination.

Abstract

In this paper, we propose the Multi-Agent Corridor Generating Algorithm (MACGA) for solving the Multi-agent Pathfinding (MAPF) problem, where a group of agents need to find non-colliding paths to their target locations. Existing approaches struggle to solve dense MAPF instances. In MACGA, the agents build \emph{corridors}, which are sequences of connected vertices, from current locations towards agents' goals, and evacuate other agents out of the corridors to avoid collisions and deadlocks. We also present the MACGA+PIBT algorithm, which integrates the well-known rule-based PIBT algorithm into MACGA to improve runtime and solution quality. The proposed algorithms run in polynomial time and have a reachability property, i.e., every agent is guaranteed to reach its goal location at some point. We demonstrate experimentally that MACGA and MACGA+PIBT outperform baseline algorithms in terms of success rate, runtime, and makespan across diverse MAPF benchmark grids.

Paper Structure

This paper contains 14 sections, 3 theorems, 7 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background
  3. MAPF algorithms
  4. The Corridor Generating Algorithm (CGA)
  5. Multi-Agent CGA (MACGA)
  6. Choosing a Corridor
  7. Evacuating the Corridor
  8. Passing through the Corridor
  9. Temporary Targets
  10. Pseudocode
  11. Theoretical Properties
  12. Discussion
  13. Experimental Results
  14. Conclusion

Key Result

Lemma 1

In MACGA, if the first agent in order $a_1$ is occupying a non-SV and the number of unoccupied vertices in a graph $G$ is greater than or equal to the length of the longest corridor in $G$, then the FindEVs procedure will successfully find EVs for all agents from any corridor for agent $a_1$.

Figures (7)

  • Figure 1: SVSs for different graphs. Black cells are obstacles; red cells marked by "SV" are the SVs; green cells are non-SVs.
  • Figure 2: An example of MACGA execution.
  • Figure 3: An example of assigning a temporary goal.
  • Figure 4: MAPF Grids: (a) empty-32-32, (b) random-32-32-10, (c) random-32-32-20, (d) room-32-32-4, (e) maze-32-32-2, (f) maze-32-32-4
  • Figure 5: Success Rate
  • ...and 2 more figures

Theorems & Definitions (5)

  • Definition 1: Separating Vertex
  • Definition 2: Corridor
  • Lemma 1
  • Theorem 1: Completeness for $a_1$
  • Theorem 2: Reachability of MACGA