Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Encoding Reusable Multi-Robot Planning Strategies as Abstract Hypergraphs

Khen Elimelech, James Motes, Marco Morales, Nancy M. Amato, Moshe Y. Vardi, Lydia E. Kavraki

TL;DR

This paper tackles the scalability of Multi-Robot Task Planning (MR-TP) by combining the Decomposable State Space Hypergraph (DaSH) representation with learning-by-abstraction to create reusable planning strategies. It extends single-robot abstraction to multi-robot settings through Abstract Hypergraphs (AH), which abstract away explicit robot identities while preserving strategic structure. The approach consists of two steps: abstracting solutions into AHs and then grounding and refining them to solve new problems using DaSH, enabling rapid adaptation to changes in robot counts, reachability, and task constraints. This framework promises faster, lifetime-long planning by reusing generalized strategies across diverse MR-TP problems, thereby improving scalability and responsiveness in real-world multi-robot systems.

Abstract

Multi-Robot Task Planning (MR-TP) is the search for a discrete-action plan a team of robots should take to complete a task. The complexity of such problems scales exponentially with the number of robots and task complexity, making them challenging for online solution. To accelerate MR-TP over a system's lifetime, this work looks at combining two recent advances: (i) Decomposable State Space Hypergraph (DaSH), a novel hypergraph-based framework to efficiently model and solve MR-TP problems; and \mbox{(ii) learning-by-abstraction,} a technique that enables automatic extraction of generalizable planning strategies from individual planning experiences for later reuse. Specifically, we wish to extend this strategy-learning technique, originally designed for single-robot planning, to benefit multi-robot planning using hypergraph-based MR-TP.

Encoding Reusable Multi-Robot Planning Strategies as Abstract Hypergraphs

TL;DR

This paper tackles the scalability of Multi-Robot Task Planning (MR-TP) by combining the Decomposable State Space Hypergraph (DaSH) representation with learning-by-abstraction to create reusable planning strategies. It extends single-robot abstraction to multi-robot settings through Abstract Hypergraphs (AH), which abstract away explicit robot identities while preserving strategic structure. The approach consists of two steps: abstracting solutions into AHs and then grounding and refining them to solve new problems using DaSH, enabling rapid adaptation to changes in robot counts, reachability, and task constraints. This framework promises faster, lifetime-long planning by reusing generalized strategies across diverse MR-TP problems, thereby improving scalability and responsiveness in real-world multi-robot systems.

Abstract

Multi-Robot Task Planning (MR-TP) is the search for a discrete-action plan a team of robots should take to complete a task. The complexity of such problems scales exponentially with the number of robots and task complexity, making them challenging for online solution. To accelerate MR-TP over a system's lifetime, this work looks at combining two recent advances: (i) Decomposable State Space Hypergraph (DaSH), a novel hypergraph-based framework to efficiently model and solve MR-TP problems; and \mbox{(ii) learning-by-abstraction,} a technique that enables automatic extraction of generalizable planning strategies from individual planning experiences for later reuse. Specifically, we wish to extend this strategy-learning technique, originally designed for single-robot planning, to benefit multi-robot planning using hypergraph-based MR-TP.
Paper Structure (6 sections, 3 figures)

This paper contains 6 sections, 3 figures.

Table of Contents

  1. Introduction
  2. Problem Definition
  3. Approach
  4. Step I: solution abstraction
  5. Step II: generalization and reuse
  6. Acknowledgements

Figures (3)

  • Figure 1: (a) Multi-Robot Task Planning problem: robots should re-stack the boxes from their current position on the right pedestal, into a desired position on the left pedestal; we may assume both robots can reach both pedastals. (b) Solution hypergraph generated using DaSH, encoding a feasible action plan to complete the task: the blue robot first picks up the top box (represented by hyperarc 1); while it places it at the goal location, the red robot picks up the middle box (hyperarcs 2 and 3, respectively); then, the blue robot picks up the bottom box (hyperarc 4) and adds it to the goal stack (hyperarc 5); finally, the red robot places the box it is holding on top of the goal stack (hyperarc 6). (c) Abstract hypergraph, representing the generalizable solution strategy: all robot entities have been removed and explicit labels of the box entities have been stripped; the abstract hyperarcs (dashed) encode the progression of entity compositions.
  • Figure 2: (a) New planning problem: objects, locations of start and goal, robot position and reachability changed; moving boxes now requires handoffs. (b) Reconstructing the abstract hypergraph from \ref{['fig:orig-problem']}c to match this new problem. (c) Refining the reconstructed hypergraph into a complete solution: each abstract hyperarc represents a MR-TP sub-problem and is replaced with its solution hypergraph, generated using DaSH; handoff actions in the final solution are highlighted with dashed hyperarcs.
  • Figure 3: (a) New planning problem: one robot is no longer operable. (b) Reconstructing the abstract hypergraph from \ref{['fig:orig-problem']}c to match this new problem. (c) Refining the reconstructed hypergraph into a complete solution.