L2R-Nav: A Large Language Model-Enhanced Framework for Robotic Navigation

  • Xiaoze Wu
  • , Qingfeng Li
  • , Chen Chen
  • , Xinlei Zhang
  • , Haochen Zhao
  • , Jianwei Niu*
  • *Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Robot navigation in dynamic, unfamiliar environments poses a significant challenge, as it traditionally relies on static maps, which are inadequate for the ever-changing scenarios encountered in daily life. This paper introduces L2R-Nav, an innovative, end-to-end intelligent robot navigation framework. It harnesses large language model technology combined with reinforcement learning to facilitate navigation tasks based on user instructions. L2R-Nav integrates the sophisticated cognitive abilities of large language models with the training of a local navigation model, employing a novel probabilistic graph approach. This integration is aimed at pioneering new methodologies in robot interaction and navigation. The robustness and effectiveness of the L2R-Nav framework are demonstrated through extensive empirical evaluations in a variety of environments, underscoring its potential as a significant advancement in the field of robotic navigation.

Original languageEnglish
Title of host publicationKnowledge Science, Engineering and Management - 17th International Conference, KSEM 2024, Proceedings
EditorsCungeng Cao, Huajun Chen, Liang Zhao, Junaid Arshad, Yonghao Wang, Taufiq Asyhari
PublisherSpringer Science and Business Media Deutschland GmbH
Pages73-84
Number of pages12
ISBN (Print)9789819755004
DOIs
StatePublished - 2024
Event17th International Conference on Knowledge Science, Engineering and Management, KSEM 2024 - Birmingham, United Kingdom
Duration: 16 Aug 202418 Aug 2024

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume14887 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference17th International Conference on Knowledge Science, Engineering and Management, KSEM 2024
Country/TerritoryUnited Kingdom
CityBirmingham
Period16/08/2418/08/24

Keywords

  • End-to-end Navigation
  • Probability Graph
  • Reinforcement Learning
  • Robot Interaction

Fingerprint

Dive into the research topics of 'L2R-Nav: A Large Language Model-Enhanced Framework for Robotic Navigation'. Together they form a unique fingerprint.

Cite this