Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Hierarchical Trajectory (Re)Planning for a Large Scale Swarm

Lishuo Pan, Yutong Wang, Nora Ayanian

TL;DR

The paper introduces a real-time lifelong hierarchical framework for coordinating large swarms in cluttered spaces by partitioning the workspace into convex cells and executing cell-local MAPF alongside a high-level inter-cell routing layer. A novel MCF-based (MCF/OD) and One-shot MCF routing scheme regulates cell influx and detours paths to reduce congestion while maintaining real-time performance. Within each cell, an anytime MAPF/C planner and a distributed trajectory optimizer produce deadlock-free, collision-free trajectories, with a failure-tolerant relaxation when safety corridors change abruptly. The approach demonstrates substantial speedups over centralized planning and robust performance relative to decentralized methods, validated in simulations with up to 142 robots and physical experiments with 24 Crazyflie quadrotors. The work advances practical large-scale swarm coordination by merging centralized routing quality with decentralized, real-time replanning capabilities, suitable for on-demand delivery and other dynamic multi-robot tasks.

Abstract

We consider the trajectory replanning problem for a large-scale swarm in a cluttered environment. Our path planner replans for robots by utilizing a hierarchical approach, dividing the workspace, and computing collision-free paths for robots within each cell in parallel. Distributed trajectory optimization generates a deadlock-free trajectory for efficient execution and maintains the control feasibility even when the optimization fails. Our hierarchical approach combines the benefits of both centralized and decentralized methods, achieving a high task success rate while providing real-time replanning capability. Compared to decentralized approaches, our approach effectively avoids deadlocks and collisions, significantly increasing the task success rate. We demonstrate the real-time performance of our algorithm with up to 142 robots in simulation, and a representative 24 physical Crazyflie nano-quadrotor experiment.

Hierarchical Trajectory (Re)Planning for a Large Scale Swarm

TL;DR

The paper introduces a real-time lifelong hierarchical framework for coordinating large swarms in cluttered spaces by partitioning the workspace into convex cells and executing cell-local MAPF alongside a high-level inter-cell routing layer. A novel MCF-based (MCF/OD) and One-shot MCF routing scheme regulates cell influx and detours paths to reduce congestion while maintaining real-time performance. Within each cell, an anytime MAPF/C planner and a distributed trajectory optimizer produce deadlock-free, collision-free trajectories, with a failure-tolerant relaxation when safety corridors change abruptly. The approach demonstrates substantial speedups over centralized planning and robust performance relative to decentralized methods, validated in simulations with up to 142 robots and physical experiments with 24 Crazyflie quadrotors. The work advances practical large-scale swarm coordination by merging centralized routing quality with decentralized, real-time replanning capabilities, suitable for on-demand delivery and other dynamic multi-robot tasks.

Abstract

We consider the trajectory replanning problem for a large-scale swarm in a cluttered environment. Our path planner replans for robots by utilizing a hierarchical approach, dividing the workspace, and computing collision-free paths for robots within each cell in parallel. Distributed trajectory optimization generates a deadlock-free trajectory for efficient execution and maintains the control feasibility even when the optimization fails. Our hierarchical approach combines the benefits of both centralized and decentralized methods, achieving a high task success rate while providing real-time replanning capability. Compared to decentralized approaches, our approach effectively avoids deadlocks and collisions, significantly increasing the task success rate. We demonstrate the real-time performance of our algorithm with up to 142 robots in simulation, and a representative 24 physical Crazyflie nano-quadrotor experiment.

Paper Structure

This paper contains 31 sections, 2 theorems, 5 equations, 14 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Preliminaries
  4. Trajectory Parameterization
  5. Multi-agent Pathfinding (MAPF) on Euclidean Graph
  6. MAPF for Embodied Agents and Conflict Annotation
  7. Problem Formulation
  8. Geometric Partitioning
  9. Roadmap Generation
  10. Graph Partitioning and Spatial Linear Separation
  11. Local Goal Generation
  12. Multi-commodity Flow with Optimal Detour
  13. High-level Planning Formulation
  14. Cell-based MAPF Planner
  15. Local Goal Assignment
  16. ...and 16 more sections

Key Result

Theorem 1

MCF/OD is complete on a locally finite graph.

Figures (14)

  • Figure 1: Long exposure of $24$ quadrotors navigating a cluttered environment.
  • Figure 2: The user inputs a map, start, and goal locations. Our approach generates a geometric partition, distributes robots among cells (hierarchical planner), and coordinates them within each cell in parallel.
  • Figure 3: (a) inter-robot collision configuration $\mathcal{C}_{col}$ and buffered hyperplanes. (b)-(d) vertex-vertex, edge-edge, and edge-vertex generalized conflicts across cells.
  • Figure 4: Geometric partitioning with $Q\!\!=\!\!5$, before spatial linear separation (a), and after (b). Each cell is a convex polytope in the workspace.
  • Figure 5: Robots (cones) fix their paths within the buffer zone and start replanning after crossing the hyperplane.
  • ...and 9 more figures

Theorems & Definitions (4)

  • Theorem 1
  • proof
  • Theorem 2
  • proof