Lesson 18 of 21 in this unit
Separate task planning from motor control.
Use explicit skill preconditions, termination receipts and bounded recovery.
By the end: Use explicit skill preconditions, termination receipts and bounded recovery.
Prerequisites and path options
Why learn this? Keep high-level task choices separate from motor commands and test each interface using observable receipts.
Start with something concrete.
A robot can issue a push command before it has reached the object. A plausible high-level plan is not evidence that its prerequisites were achieved.
Try the interactive modelWhen reach fails, may the next push be treated as completed?
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 robot can issue a push command before it has reached the object. A plausible high-level plan is not evidence that its prerequisites were achieved.
Try it and inspect what changed
Inject a failed reach and inspect both retries and the unresolved handoff.
Compare normal and failed traces. Count attempted skills separately from completed tasks. Explain why retrying a deterministic failed reach may not help until the cause changes. Propose one additional observation or fallback, then state a work budget and a condition that ends recovery.
The calculation in words
A reach receipt requires tool-object distance at most 0.04 m. Push requires an accepted reach and ends only when object-goal distance is at most 0.05 m. Retraction requires a tool-object separation of at least 0.2 m. These thresholds belong to the skill contracts rather than the policy’s confidence.
Keep the conclusion within the evidence
The example supervisor composes reach, push and retract. Failed reach allows at most two attempts and then an unresolved handoff. It never upgrades an unexecuted push to success. This visible trace is a programmed contract example; the separate workbench trains the motor imitation policy.
This is not learned task decomposition or language planning. The planar permanent-contact plant deliberately omits the contact-establishment mechanics; the hierarchy trace exposes that missing boundary rather than pretending the motor experiment demonstrates it.
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 robot can issue a push command before it has reached the object. A plausible high-level plan is not evidence that its prerequisites were achieved.
Mechanism and units
A reach receipt requires tool-object distance at most 0.04 m. Push requires an accepted reach and ends only when object-goal distance is at most 0.05 m. Retraction requires a tool-object separation of at least 0.2 m. These thresholds belong to the skill contracts rather than the policy’s confidence.
The example supervisor composes reach, push and retract. Failed reach allows at most two attempts and then an unresolved handoff. It never upgrades an unexecuted push to success. This visible trace is a programmed contract example; the separate workbench trains the motor imitation policy.
Work through the evidence
Compare normal and failed traces. Count attempted skills separately from completed tasks. Explain why retrying a deterministic failed reach may not help until the cause changes. Propose one additional observation or fallback, then state a work budget and a condition that ends recovery.
Inject a failed reach and inspect both retries and the unresolved handoff.
What this experiment can establish
This is not learned task decomposition or language planning. The planar permanent-contact plant deliberately omits the contact-establishment mechanics; the hierarchy trace exposes that missing boundary rather than pretending the motor experiment demonstrates it.
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 reach remains outside its contact threshold after two attempts. What receipt can follow?
Take this idea with you.
This is not learned task decomposition or language planning. The planar permanent-contact plant deliberately omits the contact-establishment mechanics; the hierarchy trace exposes that missing boundary rather than pretending the motor experiment demonstrates it.
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.
Use explicit skill preconditions, termination receipts and bounded recovery.
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.