| --- |
| datasets: |
| - ZichenYan/AION-dataset-files |
| license: mit |
| pipeline_tag: robotics |
| tags: |
| - reinforcement-learning |
| - drones |
| - aerial-navigation |
| - object-goal-navigation |
| --- |
| |
| # AION: Aerial Indoor Object-Goal Navigation Using Dual-Policy Reinforcement Learning |
|
|
| [](https://www.youtube.com/watch?v=TgsUm6bb7zg) |
|
|
| AION is an end-to-end dual-policy reinforcement learning (RL) framework that decouples exploration and goal-reaching behaviors into two specialized policies for vision-based aerial ObjectNav without relying on external localization or global maps. |
|
|
| ## Files |
| | Checkpoint | Description | |
| |------------|-------------| |
| | `AION-g.dat` | Goal-reaching model | |
| | `AION-e.dat` | Exploration model | |
|
|
| ## Links |
| - 📄 Paper: [arXiv](https://arxiv.org/abs/2601.15614) |
| - 💻 Code: [GitHub](https://github.com/Zichen-Yan/AION) |
|
|
| ## Citation |
| ```bibtex |
| @article{yan2026aion, |
| title={AION: Aerial Indoor Object-Goal Navigation Using Dual-Policy Reinforcement Learning}, |
| author={Yan, Zichen and Hou, Yuchen and Wang, Shenao and Gao, Yichao and Huang, Rui and Zhao, Lin}, |
| journal={arXiv preprint arXiv:2601.15614}, |
| year={2026} |
| } |
| ``` |
