319 General Concept of Controllability and Observability

320 A system is said to be completely state controllable at time t0 if it is possible by means of control
321 (input) vector u(t) to transfer the system from any initial state x( t0 ) to any other desired state x( t f
322 ) in a finite interval of time  t f  t0  .

323 Note that an uncontrollable system is a system that is physically disconnected from input.
324 A system is said to be completely observable if its initial state can be determined from the
325 knowledge of the input u(t) and the output y(t) over a finite interval of time. In other words, it
326 follows that a system is observable if and only if the initial state can be determined from the
327 input/output records for a finite interval.
328 There are four possible states for the system
329 (i) The system is controllable and observable
330 (ii) The system is controllable and unobservable
331 (iii) The system is uncontrollable and observable
332 (iv) The system is uncontrollable and unobservable
333 The state controllability tests of State Space Models are based on two main methods.
334 The Kalman’s test for controllability: this method is applied to any form of matrix A whether A is
335 canonical form or otherwise.
336 The Gilbert’s test for controllability: this method is based on transforming the matrix A into the
337 diagonal canonical form in case of distinct eigenvalues or into Jordan canonical form in case of
338 repeated eigenvalues, after that, the test is used to determine the state controllability of the system.
339 The concepts of controllability and observability were introduced by Kalman. They are important
340 in the design of control systems in state space. The mathematical tests of controllability and
341 observability developed by Kalman are acclaimed to be elegant and admirable but lack in
342 providing a physical insight into the problems involved. The approach introduced by Gilbert forms
343 the basis of the alternative test for controllability and observability. The approach uses canonical
344 state model and provides better physical insight into the problem.
345 Test for complete state controllability (Kalman’s test).
346 A general nth order multiple input and multiple-out linear time-invariant control system with an
347 m-vector input u and p-vector output y
348 x  Ax  Bu
349 y  Cx

350 Is completely controllable if and only if the rank of the composite matrix (otherwise called
351 controllable matrix)

CONTROL ENGINEERING III EEE 512 ENGR. DR. O. J. TOLA

352 Q   B AB A2 B ..... An 1 B  is n

353 Hence,
354 Find the controllability matrix Q of the size n X (n X m)
355 Find the rank of the controllability matrix Q
356 The necessary and sufficient condition for the system to be complete state controllable is that the
357 rank of the controllability matrix must be equal to n
358 If the rank of the controllability matrix is less than (n), then the system is not completely state
359 controllable. The matrix Q has a rank (n) means the determinant of order (n X n) of any sub
360 matrix of Q has nonzero value.
361 Example
362 Determine the controllability of the systems described by the state equations

 x1   2 1   x1  1
363 (i)  x    1 2   x   1 u
 2   2  
 x1   2 0   x1  1
364 (ii)  x   1 1   x   1 u
 2    2  

365 Test for complete state observability (Kalman’s test).

366 A general nth order multiple input and multiple-out linear time-invariant control system with an
367 m-vector input u and p-vector output y
368 x  Ax  Bu
369 y  Cx

370 Is completely observable if and only if the rank of the composite matrix (otherwise called
371 observable matrix)

372 S  C T AT C T .....( AT ) n 1 C T  is n

373 Example
374 Determine the controllability of the systems described by the state equations

 x1   0 1 0   x1  0 
375
 x    0 0 1  x2   0  u (t )
 2 
 x3   6  11  6   x3   2 

376 y (t )  1 0 0 x  t 

CONTROL ENGINEERING III EEE 512 ENGR. DR. O. J. TOLA

377 SOLUTION

 0 1 0 0 
378 A   0 0 1 , B   0 
 6  11  6   2 

0 1 0  0   0 
379

AB   0 0 1 0    2 
 6  11  6   2   12 

 2
380 A B   12
2

 50 

381 Q   B : AB : A2 B 

0 0 2 
382 Q  0 2  12 

 2  12 50 

383 Q 0

384 It is full rank,

385 Therefore, the system is controllable
386 Observability

Q  C T : AT C T : AT C T 
2
387
 

1  0 0  6 
388 C   0  ,
T
A  1 0  11
T

 0   0 1  6 

0 0
389 A C  1  ,
T T
A C  0 
T2 T

 0  1 

1 0 0
390 Q  0 1 0 
0 0 1

CONTROL ENGINEERING III EEE 512 ENGR. DR. O. J. TOLA

391 Rank of the matrix is 3.
392 Therefore, the system is observable.
393 Example
394 A single input single output system is given as

 1 0 0  1 
395 x (t )   0 2 0  x(t )  1  u


 0 0  3   0 

396 y  1 0 2 x  t 

397 Test for controllability and observability

398 Solution

 1 0 0  1 
399 A   0 2 0  , B  1 
 0 0  3  0 

400 C  1 0 2

 1 0  1   1 
0
401 AB   0 2 0  1    2 
 0 0  3  0  0 

1 
402 A B   4 
2

0 

1 1 1
403 Q   B : AB : A B   1
2
2 4 
0 0 0 

404 The system is uncontrollable

405 Observability

Q  C T : AT C T : AT C T 
2
406
 

CONTROL ENGINEERING III EEE 512 ENGR. DR. O. J. TOLA

  1 0 0  1   1 
407 A C   0
T T
2 0  0    0 
 0 0  3  2   6 

1 
408 A C  0 
T2 T

18

1  1 1 
409 Q  0 0 0 
 2  6 18

410 The given system is uncontrollable and unobservable

411

412

413

414

415

416

417

418

419

420 Gilbert Test for Complete State Controllability

421 Let us assume that the eigenvalues 1 , 2 ,..... n of matrix A are all distinct.

422 Find the transformation matrix T

423 T=V and the transformation matrix T is the Vander Monde Matrix of A and is given as

1 1 1 1 
   n 
424 
T V  2
1 2 3

1 22 32 ... n2 

 n 1 n 1 n 1 n 1 
1 2 3 ...n 

425 Find the transformed matrices  At , Bt , Ct , Dt  of the diagonal canonical form as given below

CONTROL ENGINEERING III EEE 512 ENGR. DR. O. J. TOLA

426 At  T 1 AT  Diagonal Matrix

427 Bt  T 1 B

428 Ct  CT

429 Dt  D

430 The system with distinct eigenvalues is completely state controllable if and only if no row of
431 Bt  T 1 B has all zero elements.

432 If the elements of one row of the (n X m) matrix Bt are all zero, then the corresponding state
433 variable cannot be controlled.
434 Example
435 Consider the system with state equations

 x1   0 1 0   x1  0 
436
 x    0 0 1  x2   0  u (t )
 2 
 x3   6  11  6   x3  1 

437 y (t )  1 0 0 x  t 

438 Estimate the state controllability by Gilbert’s Test

439 Solution
440 It is necessary to express A in the canonical form. Find eigenvalues of A

 1 0
441  I  A  0  1  0
6 11   6

442  3  6 2  11  6  0

443    1   2    3  0
444 1  1,

445 2  2,

446 3  3

CONTROL ENGINEERING III EEE 512 ENGR. DR. O. J. TOLA

 1 1 1   1 1 1
  
447 T   1 2 3   1  2  3 
 
  2  2  2   1 4 9 
 1 2 3 

 6  4  1 
 6 8 2 
   3 2.5 0.5
448 T 1 
adj  T  

 2  3 0.5  
 3  4  1 
T 2  
 1 1.5 0.5 

 3 2.5 0.5 0  0.5

449 Bt  T B   3  4  1  0    1 
1

 1 1.5 0.5  1  0.5

450 As none of elements of Bt  T 1 B are zero, the system is completely state controllable.

451 Observability

452

1 1 1 453
Ct  CT  1 0 0  1  2  3  1 1 1
 1 4 9 
454 As there is none zero element in matrix Ct, then the system is completely observable.
455

456

457 Example
458 Consider the system where the state space representation given by

 x1   0 1 0   x1  0 
459
 x    0 0 1  x2   0  u (t )
 2 
 x3   0  2  3   x3  1 

460 y (t )  3 4 1 x  t 

461 Estimate the state observability by Gilbert’s Test

462 Solution

CONTROL ENGINEERING III EEE 512 ENGR. DR. O. J. TOLA

  1 0
463  I  A  0  1  0
0 2  3

464  3  3 2  2  0

465     1   2   0

466 1  0,

467 2  1,

468 3  2

469 The eigenvalues are distinct

470 Therefore, the modal matrix is
471

1 1 1  1 1 1
  
472 T  1 2 3   0  1  2 
 
 2  2  2  0 1 4 
 1 2 3 

1 1 1
473 Ct  CT  3 4 1 0  1  2   3 0  1

0 1 4 

474 As there is one zero element in matrix Ct, then the system is not completely observable.
475 Repeated eigenvalues the modal matrix is given by

1 0 1 1 
 1 3 4 
T   2
1
476
 21 32 42 
 1 
 13 312 33 43 

477 Example
478 Consider a system, where the state space representation given by
479

CONTROL ENGINEERING III EEE 512 ENGR. DR. O. J. TOLA

  x1   0 1 0   x1  0 
480
 x    0 0 1   x2   0  u (t )
 2 
 x3    4  8  5  x3  1 

481 y (t )  3 4 1 x  t 

482

483 Evaluate the observability of this system using Gilbert’s test

484 Solution
485

 1 0
486  I  A  0  1  0
4 8  5

487  3  5 2  8  4  0

488    1   2    2   0
489 1  1,

490 2  2,

491 3  2

492 Because of the repeated eigenvalues, the modal matrix is given as

1 1 0  1 1 0 
   
493 T  1 2 1    1  2 1 
  2  2 2   1 4  4 
 1 2 2 
494 Transform the model to canonical form

1 1 0 
495 Ct  CT  3 4 1  1  2 1    0  1 0

 1 4  4 

496 There is one zero element in matrix Ct , then the system is not completely observable

497

498 Assignment

CONTROL ENGINEERING III EEE 512 ENGR. DR. O. J. TOLA

499 Using Gilbert’s test determine whether or not the following systems are completely state
500 controllable
501 (Assume that matrix A in each case is diagonal)

 1 0  1 
502 (i) x    x   u
 0  2  0
 1 0 5 
503 (ii) x    x   u
 0  5  2
504 Using Gilbert’s test determine whether or not the following systems are completely state
505 controllable and observability
506 (Assume that matrix A in each case is Jordan diagonal in each case)

 x1   1 1 0   x1   2 
507 (i)
 x    0 1 0   x2   0  u (t )
 2 
 x3   0 0  2   x3  3 
 x1   3 1 0   x1   x1 
 x     y1   3 0 5  
508 (ii) 0 3 1   x2  ,     x2
 2  y 2 0 1   
 x3   0 0 3  x3   2    x3 

CONTROL ENGINEERING III EEE 512 ENGR. DR. O. J. TOLA