Robot recording guide
Reviewed · Zagolo
For robot dynamics calibration, useful recordings pair issued commands with the robot’s measured joint response, preserve their timing, and identify the configuration used for each run. The right signals and motions depend on the parameters you want to investigate.
You do not need to prepare a data package before talking with Zagolo. Start with your robot, simulator, and the motion that does not match. Existing logs can stay in their current format. This guide supports the recording plan we prepare with your engineer if new runs are needed.
Record commands and response together.
Capture issued joint commands alongside measured positions and torques. Include velocities when available. Identify the control mode and distinguish measured torque from commanded or estimated effort. A position target and a measured actuator torque describe different parts of the control loop; keep their meanings explicit.
Record joint names, ordering, units and sign conventions. A signal without its units or mapping can look like a dynamics error when it is actually a recording or replay mismatch.
Preserve the original timing.
Include timestamps for commands and measurements, the clock source, and the sampling rate. Flag dropped samples, clock resets and any known offset between streams. Preserve raw timestamps rather than silently shifting, resampling or smoothing the recordings.
Timing matters when evaluating actuator delay and trajectory tracking. If data from two devices use different clocks, record how they were synchronized. We agree how to align the streams before comparing hardware and simulation.
Document the robot, model and controller.
- Robot identity, joint mapping, payload and starting state for each run.
- Model file/version and simulator version, including relevant solver or timestep settings.
- Firmware, control mode, controller settings and any limits active during recording.
- Run notes for contacts, saturation, faults, interruptions and configuration changes.
Keep the configuration consistent across the before-and-after comparison. If something changes, identify it explicitly so the report does not attribute that change to parameter fitting.
Agree the excitation before operating the robot.
Your engineer operates the hardware within your team’s operating limits. We review the proposed motions and recording script together. The motion should expose the parameters being fitted; repeating one movement at one speed may leave friction, damping or other effects difficult to distinguish.
This guide is a data reference, not a robot operating procedure. The exact sampling rate, motion range, speed and torque limits are robot-specific and must be agreed before new runs.
Reserve complete test runs before fitting.
Identify calibration and final-evaluation runs separately. Reserve complete runs for testing before fitting begins, rather than selecting favorable samples afterward. Freeze the robot, firmware, payload and comparison protocol across the evaluation.
Keep identifiers with every recording. The calibration and evaluation method explains how we compare original and fitted models on the same reserved runs and report remaining mismatch.
Keep raw data and explain transformations.
Preserve unfiltered recordings and run notes. Agree the export format and required fields before preparing a new export. If you filter, smooth or resample data, retain the original and record the transformation, its settings and the reason for it.
Before sending confidential logs, agree the transfer method, access, retention and deletion arrangements with us. A brief description is sufficient for first contact. See data and privacy.
Technical background
These upstream references describe simulator-specific system identification tools. They provide background; compatibility with your particular robot, controller and workflow is confirmed during setup review.