Lesson 13 of 21 in this unit
Collect useful robot demonstrations.
Define observation/action timing, resets and whole-trajectory splits.
By the end: Define observation/action timing, resets and whole-trajectory splits.
Prerequisites and path options
Why learn this? Inspect provenance and action timing before treating a demonstration as trustworthy training evidence.
Start with something concrete.
A teaching robot moves a puck. You record its location, issue one command and observe the result. Reversing that order silently teaches the wrong mapping.
Try the interactive modelWhy split complete trajectories rather than random neighboring frames?
Change one thing. Watch what follows.
Loading this browser experiment…
Understand what changed.
A shorter explanation, with the original worked examples available below. Both explain the same experiment and its limits.
The question
A teaching robot moves a puck. You record its location, issue one command and observe the result. Reversing that order silently teaches the wrong mapping.
Try it and inspect what changed
Record five rightward commands and a correction; inspect before/action/after values.
Record two trajectories from the same reset, then a third with a changed drag. Which change adds dynamics coverage, and which mainly repeats a familiar situation? Repetition can measure variability without adding a new task.
The calculation in words
At drag 0.2, a 1 m/s command lasting 0.1 s moves 0.08 m. The position before the command is the input; the command is its supervised target. The next position is useful for replay validation, not an input available before execution.
Keep the conclusion within the evidence
Keep a data card with operator, timestamp convention, action units, resets, goal and collection conditions. A successful path is not enough: include recovery states and deliberately inspect gaps. Splitting by whole episodes prevents adjacent states from crossing the training/development boundary. Our training and development starts differ, but remain authored and public.
Demonstration provenance is a declared record, not authenticated authorship. The simulator assumes permanent ideal pusher contact. It cannot represent a missed push or slipping contact.
You can explore every lesson directly. Completing this page, submitting a project report and demonstrating understanding on a separate case are different records. Optional code is available in the robot-learning studio; it does not increase guided lesson completion requirements.
Open the full worked examples, calculations and limitations
This is the same reference used in Technical mode. Expand it whenever you need the exact numbers, assumptions or further explanation.
Start with the task
A teaching robot moves a puck. You record its location, issue one command and observe the result. Reversing that order silently teaches the wrong mapping.
Mechanism and units
At drag 0.2, a 1 m/s command lasting 0.1 s moves 0.08 m. The position before the command is the input; the command is its supervised target. The next position is useful for replay validation, not an input available before execution.
Keep a data card with operator, timestamp convention, action units, resets, goal and collection conditions. A successful path is not enough: include recovery states and deliberately inspect gaps. Splitting by whole episodes prevents adjacent states from crossing the training/development boundary. Our training and development starts differ, but remain authored and public.
Work through the evidence
Record two trajectories from the same reset, then a third with a changed drag. Which change adds dynamics coverage, and which mainly repeats a familiar situation? Repetition can measure variability without adding a new task.
Record five rightward commands and a correction; inspect before/action/after values.
What this experiment can establish
Demonstration provenance is a declared record, not authenticated authorship. The simulator assumes permanent ideal pusher contact. It cannot represent a missed push or slipping contact.
Keep collection, fitting, frozen execution and interpretation as separate records. Save the configuration and complete result, including failures, in the robot-learning project. Report submission checks required evidence and text presence; reasoning and implementation quality need independent review.
A recorded command is +1 m/s for 0.1 s at drag 0.4. What x displacement should its receipt contain?
Take this idea with you.
Demonstration provenance is a declared record, not authenticated authorship. The simulator assumes permanent ideal pusher contact. It cannot represent a missed push or slipping contact.
Explore related topics and the topic map
How this topic connects
Explore this topic on the interactive mapUse these connections to review a foundation or choose where to go next. You can open any topic without completing the others.
What it builds on
- Geometry and control
Prevent coordinate-frame mistakes when combining robot positions, sensor readings and targets.
Define observation/action timing, resets and whole-trajectory splits.
What it enables
- Robot-learning project Apply in a project
Combine demonstration provenance, policy comparisons and complete-task evidence in the robot-learning project; keep embodiment limits and review status explicit.