On Merging Agents in Multi-Agent Pathfinding Algorithms

Abstract

In Multi-Agent Pathfinding (MAPF), the task is to find non-colliding paths for a set of agents. This paper focuses on search-based MAPF algorithms from the Conflict-Based Framework, which is introduced here. A common technique in such algorithms is to merge a group of dependent agents into a meta-agent and plan non-colliding paths for the meta-agent using a low-level MAPF sub-solver. We analyze the patterns that emerge when agents are merged in an arbitrary order. We then introduce policies for choosing which agents or meta-agents to merge to achieve improved efficiency in three algorithms: Independence Detection (ID) and Improved Conflict-Based Search (ICBS), which are optimal, and Priority-Based Search (PBS), which is a fast suboptimal algorithm. Experimental results show a significant improvement in efficiency

Greedy Priority-Based Search for Suboptimal Multi-Agent Path Finding

“…We run an experimental evaluation in order to (1) support our design choices for TR and IC within GBPS, (2) evaluate the impact of different combinations of GBPS improvements, and (3) compare the success rates and SOCs of GPBS with all the improvements with PP, PBS, state-of-theart bounded-suboptimal algorithm, namely Explicit Estimation Conflict-Based Search (EECBS) (Li, Ruml, and Koenig 2021), and state-of-the-art suboptimal algorithms, namely PBS with the merging technique (PBS w/m) (Boyarski et al 2022) and Large Neighborhood Search (MAPF-LNS2) (Li et al 2022).…”

“…Several improvements have been proposed for PP, such as assigning priorities to agents according to the distances from their start to their goal locations (Berg and Overmars 2005) and restarting PP with randomly-assigned priorities upon failure (Bennewitz, Burgard, and Thrun 2001). However, few techniques have been developed to improve the effectiveness of PBS (Boyarski et al 2022). We close this gap by proposing several improvements for PBS by adopting a range of techniques and concepts from optimal and bounded-suboptimal search algorithms.…”

“…Empirically, we compare GPBS with other prioritized algorithms, namely PP and PBS. We also compare it with all our improvements to the state-of-the-art boundedsuboptimal algorithm, namely Explicit Estimation Conflict-Based Search (EECBS) (Li, Ruml, and Koenig 2021), and suboptimal algorithms, namely PBS with the merging technique (PBS w/m) (Boyarski et al 2022) and Large Neighborhood Search (MAPF-LNS2) (Li et al 2022). Results over 6 maps with up to 2,000 agents show that GPBS with all our improvements has a higher success rate in all cases than these state-of-the-art algorithms for a 1-minute runtime limit.…”