Demonstrates real-world robotic navigation policy training and deployment in under 120 minutes using only a consumer laptop and no human intervention.
March 30, 2026
Original Paper
120 Minutes and a Laptop: Minimalist Image-goal Navigation via Unsupervised Exploration and Offline RL
arXiv · 2603.26441
The Takeaway
It breaks the paradigm that effective robotics requires massive GPU clusters and weeks of pretraining. This 'minimalist' approach democratizes real-world robotics research by proving that offline RL and unsupervised exploration can work with extreme hardware constraints.
From the abstract
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