Robot SysID Lab

Course delivery status

This page is the delivery ledger for the 19 K/L lesson IDs. It separates the reading or concept page from usable numerical evidence and from teaching media. A published page is not automatically a runnable experiment or a video lecture. The shared rules are in the lesson delivery policy.

Evidence and media matrix

Available means that the linked artifact is checked in and usable for the named question. Planned means the question is useful but its run or render is not complete. Not needed means a static derivation or compact figure carries the question more clearly; the reason is stated in the row.

ID Page class Usable evidence Video status and named question
K0 Concept + cited lab L0 mismatch report and replayed mismatch clip Available · 1: l0-mismatch.mp4, reused for “what does a bad model look like?”
K1 Reading Signal and parameter explanation on page Not needed: the boundary table and exercise are shorter and clearer as static text
K2 Concept Labelled pendulum diagram and analytic decomposition; explicitly illustrative Not needed: a static derivation and figure answer the units/force-balance exercise; no experiment evidence is implied
K3 Concept + cited L1 L1 friction report and L1 boundary evidence Available · 3 reused: l1-boundary.mp4, l1-phase.mp4, l1-friction.mp4 for command delay and resisting torque
K4 Concept + L0-E evidence L0-E report and sensitivity/coverage figures Available · 1 reused: l0-excitation.mp4 for the excitation comparison
K5 Concept + L0 evidence L0 report, fit path and multi-start metrics Available · 1 reused: l0-fit-walk.mp4 for how an optimizer path changes the model
K6 Concept + held-out evidence L0 report, fit/held-out split and residuals Available · 1 reused: l0-heldout.mp4 for a prediction on a withheld motion
K7 Concept + cited friction evidence L1 friction report, with effective-parameter limits Not needed: the report figure and comparison table carry the compensation question; no extra clip is required
K8 Concept Control/transfer protocol and answer guide on page Not needed: the bounded protocol is a static design decision; downstream robot evidence is planned with L4
L0 Runnable L0 fixed report, metrics and local runner Available · 5: mismatch, fit lands, fit walk, robust starts, held-out
L0-E Runnable L0-E frozen report, sensitivity/coverage/loss/multistart/held-out outputs Available · 1: l0-excitation.mp4 for weak excitation and comparison
L1 Runnable L1 report plus friction report Available · 4: boundary, phase, held-out, friction
L1-O Design draft Contract and planned observation-noise comparison on page; no checked run Planned: observation noise has a named residual question, but no rendered clip or report yet
L1-S Design draft Contract and planned saturation comparison on page; no checked run Planned: saturation would benefit from a clip once its local run exists
L2 Design draft Coupled-leg contract; simulator/report pending Planned: coupling and parameter compensation need a real leg run before a clip is made
L3 Design draft Contact contract; simulator/report pending Planned: contact onset and residuals need a real run
L4 Design draft Whole-robot/control-comparison contract; report pending Planned: component/global route and bounded controller comparison need checked evidence
L5 Design draft Parameter-sharing contract; report pending Planned: comparing independent and shared parameters needs the actual scaling run before a clip can be made
L6 Design draft Structural-mismatch and cross-engine contract; reports pending Planned: two mismatch cases require actual backend runs before media can be scoped

The ten independent current clips are short, silent teaching aids; subtitles and written explanations carry the language-specific narrative. Reused clips are counted in each lesson only when their question and boundary are stated. L1-O, L1-S, and L2–L6 remain design drafts; independent learner acceptance for L0/L1 and the full course is still pending.

Current delivered path

K0, K1, L0 and L1 are readable today. L0 and L1 have reproducible local CPU runners and checked reports; L1 includes the isolated friction extension. L0-E is now a runnable CPU evidence package. K2 is a concept lesson with a derivation and analytic teaching figure, explicitly separate from experiment evidence. K3–K8 cite the evidence that answers their questions or state why a clip is not needed. The remaining L pages are contracts awaiting their numerical paths.

The website provides paired English and Chinese pages. Marimo apps remain local adapters; a browser/WASM surface is optional and is never inferred from page publication. Independent learner acceptance still requires an actual reader to identify the initial mismatch, explain held-out performance, change one setting, and state the assumptions.