EmbodiedAgent: A Scalable Hierarchical Approach to Overcome Practical Challenge in Multi-Robot Control

Hanwen Wan, Yifei Chen, Zeyu Wei, Dongrui Li, Zexin Lin, Donghao Wu, Jiu Cheng, Yuxiang Zhang, Xiaoqiang Ji · 2025

Read EmbodiedAgent: A Scalable Hierarchical Approach to Overcome Practical Challenge in Multi-Robot Control by Hanwen Wan, Yifei Chen, Zeyu Wei, Dongrui Li, Zexin Lin, Donghao Wu, Jiu Cheng, Yuxiang Zhang, Xiaoqiang Ji (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-10030.txt.

Description

Hanwen Wan and co-authors build a tiered boss for teams of unlike robots, called EmbodiedAgent. Big sight-text models can plan robot jobs, but they dream up steps the gear can't do. Their boss splits each job into skills the robots hold, checks each step against the room, and keeps a neat memory of what has been done. A bank of past plans helps it stay grounded.

That bank, MultiPlan+, holds more than eighteen thousand marked plans across a hundred scenes. Part of the set is made of jobs that can't be done, to teach the boss to say no. A new score card grades plans by machine counts plus expert reads from a model. The boss topped the card at about seventy-two in a hundred, past the best past models. A real office run showed it can steer unlike robots through a long job.

Catalog record

CREATORHanwen Wan, Yifei Chen, Zeyu Wei, Dongrui Li, Zexin Lin, Donghao Wu, Jiu Cheng, Yuxiang Zhang, Xiaoqiang Ji
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE28 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-10030

Browse the full library · About the Tower of Babel