Targeted Parallelization of Conflict-Based Search for Multi-Robot Path Planning

TL;DR

This paper tackles the computational bottleneck of multi-robot path planning by proposing targeted parallelization of bounded-suboptimal conflict-based search. It introduces DP-ECBS, a decentralized parallelization for strongly interacting robots on small, dense maps, and PB-ECBS, a parallel bypass and conflict-counting approach for large, sparse instances, both preserving bounded-suboptimality and completeness. Empirical results show DP-ECBS yields 2×–4× speedups and higher success rates on dense problems, while PB-ECBS enables solving thousands of robots on large maps within seconds with modest suboptimality (≈1.0–1.1). The work demonstrates practical improvements in MRPP scalability and provides a path for adaptive, parallel algorithms in real-world, multi-robot systems.

Abstract

Multi-Robot Path Planning (MRPP) on graphs, equivalently known as Multi-Agent Path Finding (MAPF), is a well-established NP-hard problem with critically important applications. As serial computation in (near)-optimally solving MRPP approaches the computation efficiency limit, parallelization offers a promising route to push the limit further, especially in handling hard or large MRPP instances. In this study, we initiated a \emph{targeted} parallelization effort to boost the performance of conflict-based search for MRPP. Specifically, when instances are relatively small but robots are densely packed with strong interactions, we apply a decentralized parallel algorithm that concurrently explores multiple branches that leads to markedly enhanced solution discovery. On the other hand, when instances are large with sparse robot-robot interactions, we prioritize node expansion and conflict resolution. Our innovative multi-threaded approach to parallelizing bounded-suboptimal conflict search-based algorithms demonstrates significant improvements over baseline serial methods in success rate or runtime. Our contribution further pushes the understanding of MRPP and charts a promising path for elevating solution quality and computational efficiency through parallel algorithmic strategies.

Figures (8)

• Figure 1: Types of MRPP instances and their illustrative search tree structure. (a) An example with strong robot-robot coupling. (b) Solving strongly coupled instances demands traversing numerous branches through the search tree, as many branches lead to dead ends. (c) An example with weak robot-robot coupling. (d) Solving weakly coupled instances generally does not involve extensive branch exploration, as many branches lead to good quality feasible solutions.
• Figure 2: Experimental results comparing DP-ECBS, ECBS, EECBS, and LaCAM on map orz201d. Metrics include computation time, success rate, and SOC optimality.
• Figure 3: Experimental results comparing DP-ECBS, ECBS, EECBS, and LaCAM on map random-32-32-20. Metrics include computation time, success rate, and SOC optimality.
• Figure 4: Experimental results comparing DP-ECBS, ECBS, EECBS, and LaCAM on rearranging robots on a $60 \times 60$ obstacle-free map (an example instance of which is shown in Fig. \ref{['fig:example_rearrangement']}. Metrics include computation time, success rate, and SOC optimality.
• Figure 5: An example instance of dense multi-robot rearrangement.

Theorems & Definitions (2)

• Theorem III.1
• proof