Instruction-Augmented Long-Horizon Planning: Embedding Grounding Mechanisms in Embodied Mobile Manipulation

Fangyuan Wang, Shipeng Lyu, Peng Zhou, Anqing Duan, Guodong Guo, David Navarro-Alarcon · 2025

Read Instruction-Augmented Long-Horizon Planning: Embedding Grounding Mechanisms in Embodied Mobile Manipulation by Fangyuan Wang, Shipeng Lyu, Peng Zhou, Anqing Duan, Guodong Guo, David Navarro-Alarcon (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-2503-08084.txt.

Description

Fangyuan Wang and co-authors present a planner that lets language models direct a mobile robot through long, multi-step jobs, in a 2025 paper. Humanoid robots still struggle to chain many skills in messy real rooms. Planners built on large language models often guess at the scene from words alone. Their system, IALP, instead feeds the model live sensor data and grounds each step in what the robot can actually do.

The system turns a spoken instruction into a formal planning problem, solves it, and loops back with fresh feedback. In real trials, each job chained seven distinct manipulation skills, and the system finished over eighty percent of them on average. The authors offer the planner as a high-level brain that other robot bodies can use. The work targets homes and warehouses where tasks are long and the rooms are never tidy.

Catalog record

CREATORFangyuan Wang, Shipeng Lyu, Peng Zhou, Anqing Duan, Guodong Guo, David Navarro-Alarcon
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE48 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-08084

Browse the full library · About the Tower of Babel