Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Loosely Synchronized Rule-Based Planning for Multi-Agent Path Finding with Asynchronous Actions

Shuai Zhou, Shizhe Zhao, Zhongqiang Ren

TL;DR

This work tackles MAPF with asynchronous actions (MAPF-AA) by introducing Loosely Synchronized Rule-based Planning (LSRP), a scalable planner that fuses Loosely Synchronized Search with PIBT and leverages a caching mechanism to handle varied action durations. By planning actions for the earliest-arriving agents and using a recursive ASY-PUSH procedure with clash-handling rules, LSRP achieves order-of-magnitude improvements in agent scalability (up to ~1000 agents) at the cost of modest increases in makespan (~25%). The paper also extends LSRP with a swap operation (LSRP-SWAP) to mitigate deadlocks and improve completeness in solvable instances, with empirical results showing strong scalability across multiple maps and instances. Overall, LSRP/SWAP demonstrate practical viability for large-scale MAPF-AA in settings where asynchronous durations are significant, enabling faster planning for numerous agents at the expense of optimality and with potential for anytime refinements in future work.

Abstract

Multi-Agent Path Finding (MAPF) seeks collision-free paths for multiple agents from their respective starting locations to their respective goal locations while minimizing path costs. Although many MAPF algorithms were developed and can handle up to thousands of agents, they usually rely on the assumption that each action of the agent takes a time unit, and the actions of all agents are synchronized in a sense that the actions of agents start at the same discrete time step, which may limit their use in practice. Only a few algorithms were developed to address asynchronous actions, and they all lie on one end of the spectrum, focusing on finding optimal solutions with limited scalability. This paper develops new planners that lie on the other end of the spectrum, trading off solution quality for scalability, by finding an unbounded sub-optimal solution for many agents. Our method leverages both search methods (LSS) in handling asynchronous actions and rule-based planning methods (PIBT) for MAPF. We analyze the properties of our method and test it against several baselines with up to 1000 agents in various maps. Given a runtime limit, our method can handle an order of magnitude more agents than the baselines with about 25% longer makespan.

Loosely Synchronized Rule-Based Planning for Multi-Agent Path Finding with Asynchronous Actions

TL;DR

This work tackles MAPF with asynchronous actions (MAPF-AA) by introducing Loosely Synchronized Rule-based Planning (LSRP), a scalable planner that fuses Loosely Synchronized Search with PIBT and leverages a caching mechanism to handle varied action durations. By planning actions for the earliest-arriving agents and using a recursive ASY-PUSH procedure with clash-handling rules, LSRP achieves order-of-magnitude improvements in agent scalability (up to ~1000 agents) at the cost of modest increases in makespan (~25%). The paper also extends LSRP with a swap operation (LSRP-SWAP) to mitigate deadlocks and improve completeness in solvable instances, with empirical results showing strong scalability across multiple maps and instances. Overall, LSRP/SWAP demonstrate practical viability for large-scale MAPF-AA in settings where asynchronous durations are significant, enabling faster planning for numerous agents at the expense of optimality and with potential for anytime refinements in future work.

Abstract

Multi-Agent Path Finding (MAPF) seeks collision-free paths for multiple agents from their respective starting locations to their respective goal locations while minimizing path costs. Although many MAPF algorithms were developed and can handle up to thousands of agents, they usually rely on the assumption that each action of the agent takes a time unit, and the actions of all agents are synchronized in a sense that the actions of agents start at the same discrete time step, which may limit their use in practice. Only a few algorithms were developed to address asynchronous actions, and they all lie on one end of the spectrum, focusing on finding optimal solutions with limited scalability. This paper develops new planners that lie on the other end of the spectrum, trading off solution quality for scalability, by finding an unbounded sub-optimal solution for many agents. Our method leverages both search methods (LSS) in handling asynchronous actions and rule-based planning methods (PIBT) for MAPF. We analyze the properties of our method and test it against several baselines with up to 1000 agents in various maps. Given a runtime limit, our method can handle an order of magnitude more agents than the baselines with about 25% longer makespan.

Paper Structure

This paper contains 23 sections, 2 theorems, 8 figures, 5 algorithms.

Table of Contents

  1. Introduction
  2. Problem Definition
  3. Preliminaries
  4. Priority Inheritance with Backtracking
  5. Loosely Synchronized Search
  6. Other Related Approaches
  7. Method
  8. Notation and State Definition
  9. Algorithm Overview
  10. Recursive Asynchronous Push
  11. Cache Future Actions
  12. Toy Example
  13. Relationship to PIBT and Causal PIBT
  14. Analysis
  15. Extension with Swap Operation
  16. ...and 8 more sections

Key Result

Theorem 1

In a c-graph, in LSRP, when $i_* \in I_{curr}$, let $v^*$ denote the nearest vertex from $v^{i_*}_g$ among all vertices in $C$, then $i_*$ can reach $v^*$ within time $N \cdot D_{max}$.

Figures (8)

  • Figure 1: (a) shows the occupation of vertices when an agent traverses edge $(v_1,v_2)$ between times $(t_1,t_2)$ as shown by the black arrow. Round brackets represent open intervals, while square brackets represent closed intervals. The black vertical lines mean the vertices are occupied during the time range. (b) By duration conflict, the agents cannot move, while in conventional MAPF stern2019multi, the agents can move together clockwise or counter-clockwise.
  • Figure 2: A toy example illustrating LSRP.
  • Figure 3: An example of LSRP never terminating with only $\rm ASY\text{-}PUSH$. The blue and yellow agents will repeatedly push each other but will never successfully swap their locations to reach their goals.
  • Figure 4: Success rate and runtime results
  • Figure 5: SoC and makespan ratios of LSRP
  • ...and 3 more figures

Theorems & Definitions (3)

  • Remark 1
  • Theorem 1
  • Theorem 2