Data & collection
MCAP
MCAP is an open container format for timestamped robotics and multimodal log messages. It can store heterogeneous serialised messages with schemas, channels, attachments, metadata, optional compression, and indexes, but it does not define whether the recorded sensors, clocks, or robot actions are correct.
Also known as: MCAP log format, MCAP file, MCAP files
Updated
What MCAP stores
The MCAP specification describes records for schemas, channels, timestamped messages, chunks, attachments, metadata and summary indexes. A channel names a topic and its message encoding; a message includes log and publish times plus the serialised payload. The payload may use formats such as protobuf, JSON or another registered encoding.
What MCAP does not decide
MCAP is a container, not a robotics training schema. It does not decide which observation pairs with an action, where an episode starts, which coordinate frame a pose uses, or whether two device clocks agree. A recording can be readable and indexed while still containing a bad calibration, a dropped stream or a timing error.
A useful MCAP package should identify the recorder, schemas, topic meanings, clock sources, calibration, units, episode boundaries and any conversion applied later. The MCAP CLI can report file summaries, list channels and schemas, inspect indexed messages, and recover some truncated files, but those tools do not replace a domain-specific quality review.
MCAP and training exports
A robotics team may keep MCAP as the source recording and convert selected topics into a training format such as LeRobot. The conversion should retain a manifest that maps source topics to output fields and records resampling, filtering, redaction, dropped messages and version changes.
Sources
Related terms
Data & collection
Robot training data
Robot training data is recorded experience used to train, fine-tune, or adapt models for robot perception, prediction, planning, or control. It can include sensor observations, robot state, actions, task instructions, rewards or outcomes, demonstrations, failures, and embodiment metadata. Not every dataset contains every field, but their timing and physical meaning must be clear.
Data & collection
Data synchronisation
Data synchronisation is the process of placing sensor, state, action, annotation, and outcome records on a common timeline so samples that describe the same physical instant or transition can be matched. It requires trustworthy timestamps or trigger relationships and an explicit rule for handling streams with different rates, delays, dropped samples, and clock offsets.
Data & collection
Sensor calibration
Sensor calibration is the estimation and documentation of parameters that map raw sensor readings into physically meaningful values and known spatial relationships. In robotics it can include scale, bias, distortion, intrinsic camera parameters, sensor-to-sensor or sensor-to-robot transforms, and timing offsets. Calibration does not remove all noise or drift.
Hardware & control
Coordinate frame
A coordinate frame is a defined origin and set of oriented axes used to express positions, orientations, motions, forces, or other spatial quantities. A value has no complete geometric meaning until its frame and convention are known. Transformations relate measurements expressed in frames such as world, robot base, camera, end effector, object, or sensor.