H-infinity mixed-synthesis #151
Conversation
|
@roryyorke There's a small conflict with master that it would be good to fix. Will try to do a more thorough review in the coming days... |
|
Will have a look at the conflict. |
New functions augw and mixsyn modelled after Matlab and Octave's function of the same names. python-control doesn't currently support input group naming, so that is not implemented.
Both examples taken from Skogestad and Postlethwaite's Multivariable Feedback Control.
control/robust.py
Outdated
|
|
||
|
|
||
| def _size_as_needed(w,wname,n): | ||
| """_size_as_needed(w,wname,n) -> w2 |
There was a problem hiding this comment.
Nonstandard docstring. I don't think we use this type of docstring (fcn(args) -> output) anyplace else. In addition, parameters, returns, etc are documented in a different format. Would prefer to use a consistent style, as done in hinfsyn in this file.
There was a problem hiding this comment.
Will fix here and elsewhere (augw, _matrix, ...). I see I also used this in xferfcn.py in 310f580 in _dc_gain_cont.
control/robust.py
Outdated
|
|
||
|
|
||
| def augw(g,w1=None,w2=None,w3=None): | ||
| """augw(g,w1=None,w2=None,w3=None) -> p |
control/robust.py
Outdated
| cl: closed system mapping evaluation inputs to evaluation outputs; if p is the augmented plant, with | ||
| [z] = [p11 p12] [w], then cl is the system from w->z with u=-k*y. StateSpace object. | ||
| [y] [p21 g] [u] | ||
| info: namedtuple with fields, in order, |
There was a problem hiding this comment.
This is the first use of a namedtuple in python-control. Is there a more standard way to return this sort of information in numpy or scipy? If so, perhaps we use that format?
There was a problem hiding this comment.
namedtuple has been around since Python 2.6, so it's hardly non-standard ;). I couldn't find namedtuple use in numpy, but it is used in scipy (stats/stats.py, for example).
I can change this to a plain tuple; if we do keep the namedtuple, I'd like to remove the leading underscore from the name.
control/statesp.py
Outdated
| @@ -54,7 +54,7 @@ | |||
| import math | |||
| import numpy as np | |||
| from numpy import all, angle, any, array, asarray, concatenate, cos, delete, \ | |||
| dot, empty, exp, eye, matrix, ones, poly, poly1d, roots, shape, sin, \ | |||
| dot, empty, exp, eye, matrix, ones, pi, poly, poly1d, roots, shape, sin, \ | |||
There was a problem hiding this comment.
We should avoid using numpy.pi since this can generate errors when numpy is incorporated as a mock function. See PR 171, commit b881b7. Replace with math.pi.
There was a problem hiding this comment.
unqualified pi no longer used following your aa6e0df anyway. Will remove.
control/statesp.py
Outdated
| def _matrix(a): | ||
| """_matrix(a) -> numpy.matrix | ||
| a - passed to numpy.matrix | ||
| Wrapper around numpy.matrix; unlike that function, _matrix([]) will be 0x0 |
There was a problem hiding this comment.
Not clear to my why this is useful? Can you say a bit here (or in a comment in the code)?
There was a problem hiding this comment.
I'll expand the commentary; it's just a helper to make it easier to get 0x0 matrices, which are needed for empty state space systems.
control/robust.py: - doc-strings conform to python-control conventions - use plain tuple instead of namedtuple for auxiliary output in mixsyn control/statesp.py: - don't import pi from numpy - doc-string for _matrix expanded, conforms to conventions control/tests/robust_test.py - change test to accept plain tuple
|
I thought I'd check the docstrings via sphinx, but when I run I get the same error on master. I see This is my conda environment (2.7-based, and AFAIK up-to-date; don't know why I don't have Scipy 1.0): |
1 similar comment
|
@roryyorke I had the same problem with sphinx on my local machine. For some reason it seems to work on readthedocs.org. I think the issue may be with the way that version string is being reported (somehow comes in as a float). We should probably chase this down, but perhaps via a separate issue. I'll try to make sure things look OK on readthedocs when I get a chance (probably tonight or this weekend). |
|
Sent @roryyorke a PR that creates the documentation on readthedocs. |
This adds methods "augw" and "mixsyn", with interfaces similar to the Matlab and Octave functions of the same name. They are for so-called mixed-sensitivity design, in which the sensitivity, complementary sensitivity, and "input sensitivity" (K*S, or reference to input mapping) can be weighted and stacked for H-infinity synthesis.
As part of this, I've fixed what I think is a bug, namely that empty StateSpace objects could not be created. Such empty objects are convenient in the augw implementation.
I've added two examples for mixsyn, one SISO and one MIMO.
augw is the major part of the new code. It may be that there is some better way to implement it, but I have tried to write it in a way that no uncontrollable or unobservable modes are introduced, and so that is still (somewhat) clear.