PlatonicNav: Unveiling Semantic Correspondence in Navigation with Platonic Topological Maps

This is the official repository for the paper:

PlatonicNav: Unveiling Semantic Correspondence in Navigation with Platonic Topological Maps

Junlin Long^1*, Zeyu Zhang^2*†, Xu Deng^3*, Yiran Wang^1*
Yue Yang², Luke Borgnolo², Maxwell Twelftree², Yang Zhao^4‡

¹USYD   ²Maincode   ³UNSW   ⁴La Trobe

^*Equal contribution. ^†Project lead. ^‡Corresponding author: y.zhao2@latrobe.edu.au.

Paper | Website | Code

Abstract

Embodied visual navigation, where an agent perceives a complex environment and acts to reach a goal from raw sensory input, underpins a wide range of applications such as household service robotics, assistive robotics, and large-scale autonomous exploration. However, recent attempts to unify vision-and-language navigation (VLN) and object goal navigation (ObjNav) remain at the level of architectural fusion, mixed-task training, and large vision-language pretraining, without examining whether independently trained vision and language encoders may already share a common semantic structure. Moreover, even object-centric topological maps still ground language goals through explicit cross-modal supervision such as CLIP or large vision-language models, leaving open whether such grounding is possible from a purely vision-built map. To address these challenges, we extend the Platonic Representation Hypothesis to embodied navigation and recast vision-only ObjNav, cross-modal ObjNav, and VLN as three different interfaces to the same object-centric semantic manifold. We further introduce PlatonicNav, a training-free framework whose Platonic Topological Map fuses geometric and semantic node distances from a self-supervised visual encoder, and grounds language goals via blind matching without any paired vision-language data. Extensive experiments on simulation benchmarks including HM3D-IIN, OVON, and R2R-CE on MP3D, together with deployment on Unitree Go2, which demonstrate that PlatonicNav generalizes across tasks, modalities, and embodiments without explicit cross-modal training.

Overview

PlatonicNav includes three asset-level reference pipelines:

• OVON: category-level blind matching.
• HM3D-IIN: GT mask + IoU goal grounding for PTM ablation.
• R2R-CE: natural-language instruction goal extraction followed by blind matching.

Installation

git clone https://github.com/AIGeeksGroup/PlatonicNav.git
cd PlatonicNav

conda create -n platonicnav-public python=3.10
conda activate platonicnav-public

pip install -e .

For full simulator-level experiments, see the environment files under envs/.

Quick Start

python scripts/run_ovon_bm.py --config configs/ovon_bm.yaml
python scripts/run_iin_gtmask.py --config configs/iin_gtmask.yaml
python scripts/run_r2rce_bm.py --config configs/r2rce_bm.yaml

Run tests:

PYTHONPATH=src pytest -q

Repository Structure

src/platonicnav/
├── mapping/      # asset loading, graph I/O, DINOv3-style cull, PTM weights
├── grounding/    # blind matching, GTmask IoU, instruction goal extraction
├── planning/     # candidate-goal Dijkstra
├── control/      # frozen ObjectReact-style controller handoff
└── pipeline/     # OVON, HM3D-IIN, and R2R-CE pipeline runners

Citation

@article{platonicnav2026,
  title={PlatonicNav: Unveiling Semantic Correspondence in Navigation with Platonic Topological Maps},
  author={Long, Junlin and Zhang, Zeyu and Deng, Xu and Wang, Yiran and Yang, Yue and Borgnolo, Luke and Twelftree, Maxwell and Zhao, Yang},
  journal={arXiv preprint},
  year={2026}
}

License

This project is released under the license specified in LICENSE.

Acknowledgments

PlatonicNav builds on ideas and tooling from ObjectReact, Habitat, SAM2, DINOv3, ETPNav, and open-vocabulary embodied navigation benchmarks.