2024 · Stanford · Intermediate
Mobile ALOHA: Low-Cost Whole-Body Teleoperation
Mobile ALOHA collects whole-body, bimanual mobile manipulation demonstrations with a low-cost teleoperation system.
Direct answer
What does Mobile ALOHA: Low-Cost Whole-Body Teleoperation contribute?
Mobile ALOHA collects whole-body, bimanual mobile manipulation demonstrations with a low-cost teleoperation system.
Background
The system extends tabletop imitation learning into mobile household tasks that require moving, reaching, and coordinating both arms. Co-training with existing datasets improves success on several long-horizon tasks.
Problem
The work addresses a central constraint in Teleoperation: building systems that learn useful representations or actions while remaining general enough to transfer beyond a single demonstration or environment.
Core idea
Mobile ALOHA collects whole-body, bimanual mobile manipulation demonstrations with a low-cost teleoperation system.
Architecture and method
The system extends tabletop imitation learning into mobile household tasks that require moving, reaching, and coordinating both arms. Co-training with existing datasets improves success on several long-horizon tasks.
- Low-cost mobile teleoperation platform
- Whole-body bimanual task demonstrations
- Co-training recipe with existing datasets
Results and impact
It made the data-collection problem concrete: useful home robotics needs whole-body demonstrations, not only tabletop arm data.
Prerequisites
- Imitation learning
- Teleoperation
- Mobile manipulation
Recommended reading order
Read the explanation above, review the related topic pages, then use the primary-source links below to inspect the abstract, figures, experiments, and released implementation.
Primary sources
External links are provided after the context needed to evaluate the work.
Follow-up research
Related papers and concepts
Open X-Embodiment and RT-X
Open X-Embodiment combines robot datasets across institutions and trains policies that transfer across embodiments.
OpenVLA: An Open-Source Vision-Language-Action Model
OpenVLA is an open 7B-parameter VLA trained on the Open X-Embodiment dataset.
DROID: Distributed Robot Interaction Dataset
DROID provides diverse in-the-wild robot manipulation demonstrations across many scenes, tasks, and collectors.
Ego4D: Unscripted First-Person Video Dataset
Ego4D is a large first-person video dataset for studying what people see, remember, manipulate, and anticipate.
Common questions
Frequently asked questions
What is the main idea of Mobile ALOHA: Low-Cost Whole-Body Teleoperation?
Mobile ALOHA collects whole-body, bimanual mobile manipulation demonstrations with a low-cost teleoperation system.
Why is Mobile ALOHA: Low-Cost Whole-Body Teleoperation important?
It made the data-collection problem concrete: useful home robotics needs whole-body demonstrations, not only tabletop arm data.
What should I learn before reading Mobile ALOHA: Low-Cost Whole-Body Teleoperation?
Recommended prerequisites are Imitation learning, Teleoperation, Mobile manipulation.