Models & learning
Robot learning
Robot learning is the use of data and machine-learning methods to acquire or improve robot perception, prediction, planning or control. A robot may learn from demonstrations, rewards, direct interaction, simulation or previously collected datasets. The field includes reinforcement learning and imitation learning, but neither term is a synonym for robot learning as a whole.
Updated
What a robot can learn
Robot learning can target one component or an end-to-end policy. A system may learn visual features, object dynamics, grasp scores, a predictive world model, an action policy or a recovery behaviour. The JMLR review of robot learning for manipulation organises methods around the structure they exploit in robot tasks, rather than treating every learned component as the same algorithm.
The supervision also varies. Imitation learning fits behaviour from demonstrations. Reinforcement learning improves a policy from rewards obtained through interaction. Other methods learn representations or dynamics without directly producing actions.
How it differs from classical control
Robot learning does not imply that every part of the control stack is learned. A learned vision policy may send targets to a conventional feedback controller, while an analytical planner may use a learned dynamics or contact model. The Robot Learning tutorial presents analytical and learned methods as parts of one robotics stack.
For humanoids, conventional safety limits, state estimation and fast balance controllers can remain outside a learned high-level policy. The useful boundary depends on control frequency, available data, model error and the consequences of a bad action.
What the training record must preserve
Training data needs to keep observations, robot state, actions, task context and outcomes aligned in time. It should also identify the embodiment, action units, coordinate frames and controller mode. A large dataset cannot teach a reliable observation-to-action mapping when those fields are ambiguous or shifted relative to one another.
Sources
Related terms
Models & learning
Policy
A policy is the decision rule that maps a robot’s current observations or estimated state, and sometimes a task instruction, to an action or probability distribution over actions. It can be hand-designed or learned from demonstrations, rewards or both. In humanoid robotics, its outputs may be joint targets, torques, end-effector changes or higher-level skills.
Models & learning
Reinforcement learning
Reinforcement learning is a method in which an agent learns a policy by interacting with an environment and optimising cumulative reward. In humanoid robotics, actions change the robot and world, while observations, rewards and episode endings provide experience for improving balance, locomotion or manipulation behaviour.
Models & learning
Imitation learning
Imitation learning is a family of methods that learns a policy from examples of expert behaviour rather than specifying every control rule by hand. In robotics, demonstrations pair observations or states with actions, trajectories or inferred objectives. Behaviour cloning is one imitation-learning method; interactive and inverse approaches address different supervision and distribution-shift problems.
Data & collection
Demonstration
A demonstration is a recorded example of how an intended task or behaviour is performed, usually represented as a time-aligned sequence of observations, states and actions. For humanoid robot learning, demonstrations may come from teleoperation, kinaesthetic guidance, motion capture or autonomous experts and provide targets for imitation.
Models & learning
Robot foundation model
A robot foundation model is a broadly pretrained model intended to provide a reusable starting point for multiple robot tasks, environments or embodiments. It learns from diverse robotics and sometimes web or human data, then acts directly or is adapted with target-domain data. The term describes a training and reuse strategy, not one fixed architecture.