EIGENROOMLAB WORKSHEET 32ANSWER KEY · FOR INSTRUCTORSState feedback & observer
Recover a hidden velocity
Goal
Measure how quickly a Luenberger observer corrects a wrong initial estimate, and what a faster observer costs.
Setup
Lab eigenroom.com/lab/state-feedback?lesson=observer-recovery · short link eigenroom.com/observer-recovery
Open it on a phone or a computer; there is nothing to log in to or install. The link loads run A's settings:
- Controller ωn
- 2 rad/s
- Controller damping ζ
- 0.7
- Observer speed ratio
- 1
- Feedback state
- Observer estimate
- Target position
- 0.2 m
- Initial position
- 0 m
- Initial estimate error
- 0.3 m
- Position noise amplitude
- 0 m
- Force limit
- 10 N
- APreset, exactly as the link opens it.
- BChange only: Observer speed ratio → 3.
- CChange only: Feedback state → True state.
- DChange only: Initial estimate error → 0 m. (ext.)
1. Predict
Only position is measured. Can the observer correct an initially wrong position estimate and infer velocity?
The innovation y−C x̂ corrects a model prediction. L places the eigenvalues of A−LC; faster error dynamics remove this initial mismatch sooner but can produce a transient peak in estimated velocity. The observer receives the applied, clipped force, not an unavailable requested force. Separation of controller and observer poles assumes a linear, unsaturated exact model.
2. Observe
| Quantity | Where to read it | A | B | C | D ext. |
|---|---|---|---|---|---|
| Position estimation error at t = 0.2 s (m) | Readout at 0.2 s | -0.17708 m | -0.0022004 m | -0.17708 m | 0 m |
| Position estimation error at t = 0.5 s (m) | Readout at 0.5 s | -0.060512 m | 0.035632 m | -0.060512 m | 0 m |
| Peak |estimated velocity|, 0–1 s (m/s) | Plot, with the replay cursor | 0.109 m/s | 0.4125 m/s | 0.05149 m/s | 0.1834 m/s |
| Peak |applied force|, 0–1 s (N) | Plot, with the replay cursor | 0.8 N | 1.625 N | 0.8 N | 0.8 N |
| Position at t = 1 s (m) | Readout at 1 s | 0.18397 m | 0.24809 m | 0.14514 m | 0.14514 m |
Values computed by State-feedback, tank and RLC engine 1.0.0 at the settings above. A replay readout can differ in the last digit if the slider is not exactly at the stated time; values read from a plot are approximate.
3. Explain
How much faster does the estimation error shrink in run B?
At 0.2 s the error is -0.0022004 m in run B against -0.17708 m in run A: the observer poles are three times faster, so the initial mismatch decays sooner.
What does the faster observer cost?
In the first second its estimated velocity peaks at 0.4125 m/s, against 0.109 m/s, and the applied force peaks at 1.625 N instead of 0.8 N: correcting the initial error quickly produces a transient that the controller acts on.
Run C feeds back the true state. Why is its estimation error the same as run A's?
The estimation error obeys e′ = (A − LC)e, which does not depend on the control input when the observer receives the applied force; at 0.5 s both show -0.060512 m. The control differs: the position at 1 s is 0.14514 m against 0.18397 m.
Extension (optional)
Set the initial estimate error to 0 with run A's other settings (run D). Does observer feedback still differ from true-state feedback?
No: with a correct initial estimate and no noise the estimate equals the true state, and the position at 1 s is 0.14514 m, as in run C.