Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2026
|
| Sujets: | |
| Accès en ligne: | https://arxiv.org/abs/2603.26441 |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866911548273852416 |
|---|---|
| author | Liu, Xiaoming Zhang, Borong Li, Qingbiao Morad, Steven |
| author_facet | Liu, Xiaoming Zhang, Borong Li, Qingbiao Morad, Steven |
| contents | The prevailing paradigm for image-goal visual navigation often assumes access to large-scale datasets, substantial pretraining, and significant computational resources. In this work, we challenge this assumption. We show that we can collect a dataset, train an in-domain policy, and deploy it to the real world (1) in less than 120 minutes, (2) on a consumer laptop, (3) without any human intervention. Our method, MINav, formulates image-goal navigation as an offline goal-conditioned reinforcement learning problem, combining unsupervised data collection with hindsight goal relabeling and offline policy learning. Experiments in simulation and the real world show that MINav improves exploration efficiency, outperforms zero-shot navigation baselines in target environments, and scales favorably with dataset size. These results suggest that effective real-world robotic learning can be achieved with high computational efficiency, lowering the barrier to rapid policy prototyping and deployment. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_26441 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | 120 Minutes and a Laptop: Minimalist Image-goal Navigation via Unsupervised Exploration and Offline RL Liu, Xiaoming Zhang, Borong Li, Qingbiao Morad, Steven Robotics The prevailing paradigm for image-goal visual navigation often assumes access to large-scale datasets, substantial pretraining, and significant computational resources. In this work, we challenge this assumption. We show that we can collect a dataset, train an in-domain policy, and deploy it to the real world (1) in less than 120 minutes, (2) on a consumer laptop, (3) without any human intervention. Our method, MINav, formulates image-goal navigation as an offline goal-conditioned reinforcement learning problem, combining unsupervised data collection with hindsight goal relabeling and offline policy learning. Experiments in simulation and the real world show that MINav improves exploration efficiency, outperforms zero-shot navigation baselines in target environments, and scales favorably with dataset size. These results suggest that effective real-world robotic learning can be achieved with high computational efficiency, lowering the barrier to rapid policy prototyping and deployment. |
| title | 120 Minutes and a Laptop: Minimalist Image-goal Navigation via Unsupervised Exploration and Offline RL |
| topic | Robotics |
| url | https://arxiv.org/abs/2603.26441 |