Deploying Ten Thousand Robots: Scalable Imitation Learning for Lifelong Multi-Agent Path Finding

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Jiang, He, Wang, Yutong, Veerapaneni, Rishi, Duhan, Tanishq, Sartoretti, Guillaume, Li, Jiaoyang
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866912380433203200
author Jiang, He
Wang, Yutong
Veerapaneni, Rishi
Duhan, Tanishq
Sartoretti, Guillaume
Li, Jiaoyang
author_facet Jiang, He
Wang, Yutong
Veerapaneni, Rishi
Duhan, Tanishq
Sartoretti, Guillaume
Li, Jiaoyang
contents Lifelong Multi-Agent Path Finding (LMAPF) repeatedly finds collision-free paths for multiple agents that are continually assigned new goals when they reach current ones. Recently, this field has embraced learning-based methods, which reactively generate single-step actions based on individual local observations. However, it is still challenging for them to match the performance of the best search-based algorithms, especially in large-scale settings. This work proposes an imitation-learning-based LMAPF solver that introduces a novel communication module as well as systematic single-step collision resolution and global guidance techniques. Our proposed solver, Scalable Imitation Learning for LMAPF (SILLM), inherits the fast reasoning speed of learning-based methods and the high solution quality of search-based methods with the help of modern GPUs. Across six large-scale maps with up to 10,000 agents and varying obstacle structures, SILLM surpasses the best learning- and search-based baselines, achieving average throughput improvements of 137.7% and 16.0%, respectively. Furthermore, SILLM also beats the winning solution of the 2023 League of Robot Runners, an international LMAPF competition. Finally, we validated SILLM with 10 real robots and 100 virtual robots in a mock warehouse environment.
format Preprint
id arxiv_https___arxiv_org_abs_2410_21415
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Deploying Ten Thousand Robots: Scalable Imitation Learning for Lifelong Multi-Agent Path Finding
Jiang, He
Wang, Yutong
Veerapaneni, Rishi
Duhan, Tanishq
Sartoretti, Guillaume
Li, Jiaoyang
Multiagent Systems
Artificial Intelligence
Machine Learning
Robotics
Lifelong Multi-Agent Path Finding (LMAPF) repeatedly finds collision-free paths for multiple agents that are continually assigned new goals when they reach current ones. Recently, this field has embraced learning-based methods, which reactively generate single-step actions based on individual local observations. However, it is still challenging for them to match the performance of the best search-based algorithms, especially in large-scale settings. This work proposes an imitation-learning-based LMAPF solver that introduces a novel communication module as well as systematic single-step collision resolution and global guidance techniques. Our proposed solver, Scalable Imitation Learning for LMAPF (SILLM), inherits the fast reasoning speed of learning-based methods and the high solution quality of search-based methods with the help of modern GPUs. Across six large-scale maps with up to 10,000 agents and varying obstacle structures, SILLM surpasses the best learning- and search-based baselines, achieving average throughput improvements of 137.7% and 16.0%, respectively. Furthermore, SILLM also beats the winning solution of the 2023 League of Robot Runners, an international LMAPF competition. Finally, we validated SILLM with 10 real robots and 100 virtual robots in a mock warehouse environment.
title Deploying Ten Thousand Robots: Scalable Imitation Learning for Lifelong Multi-Agent Path Finding
topic Multiagent Systems
Artificial Intelligence
Machine Learning
Robotics
url https://arxiv.org/abs/2410.21415