Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Decentralized Unlabeled Multi-agent Pathfinding Via Target And Priority Swapping (With Supplementary)

Stepan Dergachev, Konstantin Yakovlev

TL;DR

This work addresses anonymous multi-agent pathfinding (AMAPF) in a fully decentralized setting where agents independently decide targets and motions using only local communication. Building on TSWAP, the authors introduce TP-SWAP, a target-priority swapping framework that maintains consistent goal assignments via a persistent TP table and local information exchange, with TP-UPDATE ensuring progress toward completion. Theoretical results prove completeness and a decreasing potential function guarantees termination, while extensive experiments on MovingAI maps show TP-SWAP outperforms naive decentralization and approaches centralized baselines in flowtime, with makespan benefits that improve with agent density. The approach advances scalable, complete decentralized AMAPF, offering practical implications for distributed robotic swarms and autonomous systems in uncertain communication environments.

Abstract

In this paper we study a challenging variant of the multi-agent pathfinding problem (MAPF), when a set of agents must reach a set of goal locations, but it does not matter which agent reaches a specific goal - Anonymous MAPF (AMAPF). Current optimal and suboptimal AMAPF solvers rely on the existence of a centralized controller which is in charge of both target assignment and pathfinding. We extend the state of the art and present the first AMAPF solver capable of solving the problem at hand in a fully decentralized fashion, when each agent makes decisions individually and relies only on the local communication with the others. The core of our method is a priority and target swapping procedure tailored to produce consistent goal assignments (i.e. making sure that no two agents are heading towards the same goal). Coupled with an established rule-based path planning, we end up with a TP-SWAP, an efficient and flexible approach to solve decentralized AMAPF. On the theoretical side, we prove that TP-SWAP is complete (i.e. TP-SWAP guarantees that each target will be reached by some agent). Empirically, we evaluate TP-SWAP across a wide range of setups and compare it to both centralized and decentralized baselines. Indeed, TP-SWAP outperforms the fully-decentralized competitor and can even outperform the semi-decentralized one (i.e. the one relying on the initial consistent goal assignment) in terms of flowtime (a widespread cost objective in MAPF

Decentralized Unlabeled Multi-agent Pathfinding Via Target And Priority Swapping (With Supplementary)

TL;DR

This work addresses anonymous multi-agent pathfinding (AMAPF) in a fully decentralized setting where agents independently decide targets and motions using only local communication. Building on TSWAP, the authors introduce TP-SWAP, a target-priority swapping framework that maintains consistent goal assignments via a persistent TP table and local information exchange, with TP-UPDATE ensuring progress toward completion. Theoretical results prove completeness and a decreasing potential function guarantees termination, while extensive experiments on MovingAI maps show TP-SWAP outperforms naive decentralization and approaches centralized baselines in flowtime, with makespan benefits that improve with agent density. The approach advances scalable, complete decentralized AMAPF, offering practical implications for distributed robotic swarms and autonomous systems in uncertain communication environments.

Abstract

In this paper we study a challenging variant of the multi-agent pathfinding problem (MAPF), when a set of agents must reach a set of goal locations, but it does not matter which agent reaches a specific goal - Anonymous MAPF (AMAPF). Current optimal and suboptimal AMAPF solvers rely on the existence of a centralized controller which is in charge of both target assignment and pathfinding. We extend the state of the art and present the first AMAPF solver capable of solving the problem at hand in a fully decentralized fashion, when each agent makes decisions individually and relies only on the local communication with the others. The core of our method is a priority and target swapping procedure tailored to produce consistent goal assignments (i.e. making sure that no two agents are heading towards the same goal). Coupled with an established rule-based path planning, we end up with a TP-SWAP, an efficient and flexible approach to solve decentralized AMAPF. On the theoretical side, we prove that TP-SWAP is complete (i.e. TP-SWAP guarantees that each target will be reached by some agent). Empirically, we evaluate TP-SWAP across a wide range of setups and compare it to both centralized and decentralized baselines. Indeed, TP-SWAP outperforms the fully-decentralized competitor and can even outperform the semi-decentralized one (i.e. the one relying on the initial consistent goal assignment) in terms of flowtime (a widespread cost objective in MAPF
Paper Structure (30 sections, 1 theorem, 1 equation, 8 figures, 4 tables, 2 algorithms)

This paper contains 30 sections, 1 theorem, 1 equation, 8 figures, 4 tables, 2 algorithms.

Table of Contents

  1. Related Works
  2. Problem Statement
  3. Centralized AMAPF
  4. Decentralized AMAPF
  5. Methods
  6. TSWAP
  7. Initial Goal Assignment
  8. Moving towards the goals with target swapping
  9. Decentralized TSWAP with Consistent Target Assignment
  10. Naive Fully-decentralized TSWAP
  11. TP-SWAP: Target-Priority Swapping For Decentralized AMAPF
  12. TP-UPDATE: Procedure to resolve conflicts in TP-table
  13. Running Example
  14. Theoretical Analysis
  15. TP-UPDATE Correctness
  16. ...and 15 more sections

Key Result

Theorem 1

There exists a finite time $t$ at which all goals will be achieved by the agents utilizing Alg. alg:main_tpswap.

Figures (8)

  • Figure 1: An example of a decentralized AMAPF instance with a consistent (left side) and an inconsistent (right side) initial assignments. Solid circles depict agents. Red squares are the goals. The colored area around each agent is its communication zone (it is able to communicate with the others only if they reside inside this zone).
  • Figure 2: Examples of conflict and deadlock resolutions in TSWAP: (a) Illustration of the target-swapping mechanism when an agent occupying its target location blocks the path for another agent. (b) Illustration of the deadlock resolution mechanism, where a sequence of agents forms a loop, causing them to block each other's paths.
  • Figure 3: An example of solving a decentralized AMAPF instance. Agents are depicted as disks (with the number showing their current priority). The dashed lines illustrate the agents' communication ranges. The red cells represent the goals that the agents need to reach.
  • Figure 4: Maps that are used in the experiments.
  • Figure 5: Average flowtime, makespan and standard deviations (shaded ahead) of the evaluated AMAPF solvers.
  • ...and 3 more figures

Theorems & Definitions (2)

  • Theorem 1
  • proof