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.
Also known as: expert demonstration, robot demonstration, demo
Updated
What a demonstration contains
A demonstration expresses an example of desired performance, not a learning algorithm. The robot learning from demonstration survey describes examples as state–action sequences recorded while a teacher performs the desired behaviour. A learning system may copy those actions directly, infer a plan or use the data to shape another objective.
Demonstrations can be gathered by driving the robot through teleoperation, physically guiding a compliant mechanism, retargeting captured human motion or running an existing controller. The recording is also a trajectory, but the words emphasise different properties: “trajectory” identifies the temporal sequence, while “demonstration” says that the sequence is presented as an example to learn from. The label alone does not guarantee success or expertise, so outcome and quality metadata matter.
Useful data for humanoid learning
A humanoid demonstration should keep observations, actions and robot state on a shared clock. Depending on the task, useful fields include head and wrist video, joint positions, commanded targets, force or tactile readings, task language, calibration and the final outcome. Missing timing or coordinate-frame information can make otherwise plausible motion impossible to reproduce.
The DROID dataset illustrates this structure with robot demonstrations containing multiple camera views, depth, calibration and language annotations. Across robots, the meaning of an action can still differ. Open X-Embodiment pooled trajectories from many embodiments, but its common representation still had to account for different sensors and action spaces. Humanoid data therefore needs both behavioural diversity and precise embodiment metadata.
Sources
Related terms
Data & collection
Trajectory
A trajectory is a time-ordered sequence of states or observations, actions and, where applicable, rewards generated as an agent or robot evolves. A complete episode or policy rollout often yields a trajectory, but the terms are not universally identical: trajectories may be partial, while episodes have dataset- or environment-defined boundaries.
Data & collection
Teleoperation
Teleoperation is real-time human control of a robot from a remote or mediated interface. For humanoid training-data collection, the operator’s inputs are mapped to robot motions while cameras, proprioception, commands and outcomes are recorded, producing embodied demonstrations in the robot’s own observation and action spaces.
Models & learning
Behaviour cloning
Behaviour cloning is a form of imitation learning that fits a policy to expert observation–action pairs as a supervised prediction problem. For humanoid robots, the training examples typically align camera or proprioceptive observations with commands recorded during demonstrations, so the learned policy can reproduce similar behaviour without an explicit reward model.
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.