Here are some comments on your book, from the team that taught the course,
i.e. Bo, myself and the TA Magnus Jansson. Notice that our course is the
advanced (second) control course for the undergraduates. Some of the comments
below should therefore probably be filtered accordingly.
Generally we found that the book was too hard for our students, but that by
having the book we managed to teach them a lot of important stuff that was not
included in the previous course.
In terms of technical content etc, the book is extremely good. However, we
feelt it perhaps has some pedagogical weaknesses. Sometimes there is no clear
line in terms of when and where results are presented. For example, the first
chapters introduce and use facts that relies on the coming chapters. All
throughout the book there are lots of forward references. Just to mention one,
on page 24 there is a discussion on handling of model uncertainty where LQG is
mentioned before it has been formally introduced. This is probably okey in a
graduate course where the students have already heard of LQG, but for our
students it must have been a bit annoying. Another example is RGA, which is
described at least in 4 different places.
Another thing that we find very strange is your definition of the
control error. In almost all other books it is defined as r-y, but you are
using y-r. In our opinion all possible advantages with this definition pale
before the disadvantage that in a block diagram you now find -e instead of e,
which is weird from an implementation point of view .
Below you also find a list of some scattered detailed remarks. We make no
claims that it is complete, since most of our reading has not been in proof
reading mode...
Finally, bear in mind that when one gives comments on something one is often
over-critical. We have learnt a lot from your very ambitious book.
Best regards,
Alf
=========================================================================
p48. just before eq 2.56: to give to realizable => to give the
p50. rtilde = 1 should be rtilde=R
p69. feedback rule:L=G2G1
p70. fig.3.3 missing minus sign (negative feedback)
p 84. paragraph starting "The idea of using..." seen => seem
p91. eq after eq 3.67 : Acl=... should be equal to -I (not +I)
p94. sing values after eq. 3.73 should not be a function of s, but the
absolute value of the frequency function
p96. between eq 3.76 and eq 3.77 : an "xs" to much
p108. before eq 3.100 : fond => found
p117. line 7 : function => functions , and just before eq 4.23 "is a called a"
p125. eq 4.45 : a lot of missing transposes and it should be A^2
p133. top row: One => On
p137. Exercise 4.3: How many poles...
Exercise 4.5 the zeros of a..
p138. Exercise 4.7: b12 => b21 to conform with standard notation
p141. Exercise 4.11: Do you really mean G1,G2 and not L1, L2?
In the second case I could not find out wether the RHP zero is in G2 or
in the controller. (I do agree that the pole should be in G1 !)
p 143. eq 4.85: GQ => QG
p153. after eq 4.116 : minimizing the square => minimizing the sum of the
squares...
p183. after eq 5.38 : w_p => w_T
p186 third line from bottom: y_1 => y_2
p188 third line from bottom: G_d => G_d d
p190. line 7: the the
p197. Rule 1 should be 1/Gd
p199 fig5.17: last M_5 should be M_6
p213. Chapter 6.1, end of 3rd line, "where we WERE able..."
p215. line after (6.5): has has
p217. Remark 7th line: plants => plant
p228. Chapter 6.9.1 1st line: remove either "a" or s in "plants"
p247. Chapter 6.11.2, 4th line: "noT TO expect"
p251: top row : Plot that => Plot the
p257 middle theses => these
p260. bottom line: w_a => w_A
p269 midlle: plant => plants
p285 before eq 7.89: can BE written as
p293 eq 8.2 : should shift places of N_22 and N_11
p295 top row: perturbations => perturbation
first row after fig 8.2: N_22 should be N_11
p301 line 4 : delete "we"
p307 eq 8.44 : missing \infty index
p311. 7lines from bottom : the same eigenvalues of(?) the matrix...
p322 3rd line from bottom: out => our
p330 STEP A : D => \delta_P
p331. bottom line : missing minus sign
p353. point 4.: be => by
p360 The Kalman filter minimizes: E xx^T and thus in particular E x^T x
p361. same remark as above
p362 Exercise 9.2: in the matrix Kr => K_r
What do you mean by closed loop poles, the observer poles cancels (in
theory)
p363. before the matlab code: Example 1.6 => 2.5
p 364. sensitivity function AT THE INPUT
p365 line 6 : Example 8.2 => 9.2
p366 line 5 from below: labelled => labeled
p367. bottom line : maximum => minimum
p371. 8th line from bottom: missing trace and the conjugate should be
hermitian transpose.
p372. line 7: missing trace and the conjugate should be hermitian transpose.
p374. 2nd line: (9.9)-(9.9)
p384. line 5: (7.67) => (8.69)
missing definition of R in (9.16) R=I+DD^T
p403. line5: contrpl => control
p409. middle : the one which has a SMALL(?) sigma(G(0))..
extra "which" at the end of the sentence
p410. line 7 from bottom: missing transpose on G^\dagger
p411. line 8: missing transpose
p413 line 6: missing element in matrix (=10), "large RGA-elementS"
p414. line 2: controller
p416. end of page : Is it really good to have integral actions in both loops?
I think I have read somewhere that one should avoid that because it may lead
to stability problems.
p417. line 9 from bottom: advantage OF
p419. remark 2 : are tuned ONE at a time
line 10 from bottom: missing word "decentralized ??? each ...
p485. (12.14): \Lambda instead of \Gamma ??
p485. Figure 12.12: Missing y in Frequency
p486. line 13: hey => they
p488. design step 5: T_o (second time T_0?) is called T_{ref} in Chap 9.4.3
p504. eq A.50 is only true if sigma(A) > 1
p505. middle : missing square on x_2
p512. RGA = G.*pinv(G.') (that is, a transpose and not complex conjugate
transpose)
p518. before A.110 : between tWo
p522. eq. A.128 : absolute value on e_i (even if the norm is OK)
- ------- End of Forwarded Message
------- End of Forwarded Message