Robotics · Intermediate · 15-30 hours
Robot Learning
Methods that allow robots to acquire behavior from demonstrations, rewards, interaction, or generated experience.
Direct answer
What is Robot Learning?
Methods that allow robots to acquire behavior from demonstrations, rewards, interaction, or generated experience.
Definition and scope
Methods that allow robots to acquire behavior from demonstrations, rewards, interaction, or generated experience.
Robot learning uses imitation learning, reinforcement learning, offline datasets, self-supervision, and model-based planning.
Why it matters
Hand-programming every situation does not scale to open, changing environments.
How it works
Robot learning uses imitation learning, reinforcement learning, offline datasets, self-supervision, and model-based planning.
Beginner learning path
Start with behavior cloning, then study reinforcement learning and dataset design.
Recommended next topics
Primary sources
Key papers
RT-1: Robotics Transformer for Real-World Control at Scale
RT-1 trains one transformer policy on a large multi-task dataset of real robot demonstrations.
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.
Mobile ALOHA: Low-Cost Whole-Body Teleoperation
Mobile ALOHA collects whole-body, bimanual mobile manipulation demonstrations with a low-cost teleoperation system.
pi0: A Vision-Language-Action Flow Model for General Robot Control
pi0 is a generalist robot policy trained on broad robot data to follow language instructions across dexterous tasks.
NVIDIA Isaac GR00T-Dreams
GR00T-Dreams uses world foundation models to generate synthetic robot trajectories from a single image and instruction.
Xiaomi-Robotics-1: Scaling VLA Models with 100K Hours of Real-World Trajectories
Xiaomi-Robotics-1 studies how VLA-style robot policies scale when pre-trained on over 100,000 hours of real-world manipulation trajectories.
FLUX 3 x mimic: Video-Action Models for Robotics
FLUX-mimic uses the FLUX 3 multimodal backbone as a dynamics-aware foundation for robot action prediction.
Research ecosystem
Organizations working in this area
Organization
Physical Intelligence
Generalist robot foundation models
View profile →Organization
Covariant
AI robotics for warehouse manipulation
View profile →Organization
Toyota Research Institute
Robot learning, manipulation, human-centered AI
View profile →Organization
Skild AI
General-purpose robotic brain trained across tasks, embodiments, and human videos
View profile →Organization
Figure AI
General-purpose humanoid robots and onboard VLA systems
View profile →Organization
Apptronik
General-purpose humanoid robots and embodied AI with Google DeepMind
View profile →Organization
Xiaomi Robotics
Robot foundation models, UMI-style real-world data, world models, and cross-embodiment manipulation
View profile →Organization
mimic robotics
Industrial robot learning, dexterous manipulation, and video-action models for production tasks
View profile →Common questions
Frequently asked questions
What is Robot Learning?
Methods that allow robots to acquire behavior from demonstrations, rewards, interaction, or generated experience.
Why does Robot Learning matter for Physical AI?
Hand-programming every situation does not scale to open, changing environments.
How should a beginner learn Robot Learning?
Start with behavior cloning, then study reinforcement learning and dataset design.