CHAPTER 12 Bizt. Referring to Byuation (3-29), ve have stl o =f "| o GO= CCsI~AY'B HL 7 ol] -7 sta 0 | fe - eo sa} |! (s+2)(s+3) o st2 o [rr Al sea (saneses) 7 flo ‘ = (341) (842)(543) Qo Oo (st1Xs42)]|/ = 5+ = Sst2- (st) )(st2)(st3) 0 Pe bst a ste Comparing this transfer function with bo 3 bby St be S thy S) bay s*t aeSt ay (a) we obtain M26, Apel, ag 76 bozo, =O, bre}, b=3 (a) Controllable canonical form: Referring to Byuations (11-3) and (11-4), ve have, * x o 4 Olly]. Jo Bla] o 9 Maal elo lee (2) EA 6 UM —Elixy Z x gala 7 oy a). x - 241 -Note that because of the cancellation of the terms (8 + 1,2),in the transter function, the system defined by Equations (2) and (3) is state controllable, but not observable. (») le canonical form: Referring to Equations (11-5) and.(11-6), we have oo -él[a] [3 HEL R1 0 -sf[rlel rs] & (4) o % + -é{lal fo *% g=le o sx, «) Xa Because of the cancellation of the terms (s + 3) in the transfer given By Bien 2) fi speen Gatien (4) an (3) cheney ‘bat. ite controllable. Tt is important to note that when cancellation of the numerator and deno- minator of the transfer function occurs [eee Equation (1)], the system becomes controllable but ‘not observable or observable but not controllable depending on how one writes the state and output equations. Baza. 7 Stl o -/ o GO)=CCSE-AY BRL AY ms sez o | |e / oe ogo St2 1 (s+2)(s+3) 0 sez fo aoe . (stlste (3+) U Wy ses Gti )se3) 7 4 ° oO Cst1Xsee) Zs"Hes+ 8 as eerste (StDesv2\sta) Shee SHE Comparing this transfer function with bo S38 + by s*+baS +b, S74 4 S* + Aes+ayve obtain Ub, Gratl, agus bez 0, bay $223, bg =5 Referring to Equations (11-5) and (11-6), have the state-space equations in the following observable canonical form: - - x} loo ~é]lyl fe 4/5] 0 -vialele %) lo 7 -6lig]) fo % we[o o syix % BLI2-3. Referring to Equation (12-18), the state-feedback gain matrix K can ‘be given by Kale oie 4B BET ew Aa Are mgr rea taal ‘The values of 1, 02, and %3 are determined from the desired characteristic equation: [SE (A- BW) = (54 24 ja)(se2—jaycsre) = S’ + /¥s*+ fo5 + 200 a SHS + SH Oy ‘Thus, “HIt, e360, oly = 200 e4 oF ot el* foro 100 PW=|0 0 sestlo o 1|+6}0 0 142) gy 70 nl -$ 6) -/ -5 ~6 af eS Hb oo?Wwe e al-e up 97 “7 “#3 7 Since of oT Ig ag aele| ror mv io-t 66 we have the desired staté-feedback gain matrix K as follova: ord Kaloo ris ot mu 7 1 -n' 60 arse? =fo os) arsre OLERE eo X]-3 ss9 “7? #3 3B 3996 =[e 0 1)) ppaze9 29.983/ = [zh 983/ "T s99 O.8SD# 0.08/20 0.8 (20. SEES se 39 7 -¥3 9 -e “7 OnE 80120 20120 e 7 7 190-269 4A UR/F SSE / 2h783/ SSlh/ 29R0F 2428/9) Belz. - the MATIAB programs to obtain the state-feedback gain matrix K by use of ‘command “acker* or comand "place" are shown on the next page.96° Generating atx K by use of command asker" #0 Aro 1 00 0 Iyl -5 vie: nd 15 K- 287831 55181 2.4819 eee. Bl25. Substituting es ke=-[h ¢] 1 | ES SEP [Ee of ‘(7 7H ‘The characteristic equation becomesait hel s-2 Because of the presence of one eigenvalue (s = 2) in the right-haif s plane, the system is unstable vhatever values k, and kz may assume. |sz-Al -| +4 = Gtki )(s-2) =0 B12-6. Since YO. “a _ fo DE> (st 0s42) (542) ~ S24 es* 47/5 +E we have Peeg+ 9+ by = lon Using the state variables as defined in the problem statement, the state equ- ation becomes % o 7 ollat’ je m=] 0 9 Fixl+fola EA ~6 -w —8llx;| jo o7 @ @ A=|e oo -f}, Bslo -6 6 o Referring to Bquation (12-18), the state-feedback gain matrix K can ‘be given by 1 Kale o/ILB ae Atel’ #@ $A) = Ars x, Att At GT ‘Me values of Wy, 2, and & 3 are determined from the desired characteristic equation: |sz- (A-BK)| = (54245 215 )(s42 ~f2vF stv) = SOM s*+ SES+ 160 SEK, SAE St Oy Malt, Wr= 8h, wy =fbo = 246 ~o + of fos ef? of e@ gAap eo 0 s]*#lo o 7| tl 0 oe y “6-0 6 ~b + 6 “6 -M H% ‘Since o 0 so Ig A wel = oe 0 -60 we have . 0 0 wll ee 8 (o 2 so we -be) |-48 46-3 40 ~b6 2590 we nis BF 44 06 ollie ee ob of Ol—pp se -3 ate o we -1s 8% RE. BRE KY a6 33.6 as 18% as a8 =s[oe 7} — =[se aS a8] B:12-7. A MATIAB solution of Problem 3-12-6 1s given!on the next page.% ***6 Geucrating matrix K by use of command "acker” ***¢* Arf 1 a0 0 6 a1 B={0,0;10}, xe E24j*2°(oqnt(3)) -2-j*2%(eqet(3))_-10]; Keacka(ABJ) K- 15.4000. 4.5000 0.8000 512-8. From Figure 12-49 ve obtain Un bi Cr-H1) hada hts =m - Kx ther K=Dh fe hs) Noting that the rank of o 8 / Mele ae @ele|0 1 -6 4-6 3h is three, arbitrary pole placement is possible, ‘The characteristic equation for thie system is sof 0 |sZ-al= os -/ |= steoste ss “oe o Ss. Sté! = S44 S*4+disStdy m0 Hence U6, 425, azo Since the state equntion for the systen 1s already in the controllable cano- nical form, we have 2 = & desired characteristic equation is (se atjer(s t2jesein =S5 +t 53+ bos +200 = ST HKIS + Ae St hyfrom which ve obtain Walt, W,=6a, % =200 Ka(e-as wan ta] TA =[200-6 bo-S- /%~6]T in s[2e0 ss F] ‘The state equation for the designed systen is Z= Ant ge =Axt B(-Kxtk,r) =@G-B4Ox +BY ‘Since . o 7 o| fo o fe ArBR Elo os [-leflzee se ejale 6 7 oo o-s -6} |! -ae0 -60 “MH we have %] fo 7 offs] | o Ka} oo Ft |lxelt].a[r q) a, | [200 -40 -“# | 4%] | 200) ‘The output equation is : x) yell o ol|lx, (2) Ao unit-step response of the designed system can be obtained fram Raaticne o and (2) Dy subetitueing t= 1(2) and finding ye), A MATLAS poogeem to obtain the unit-step response curve [y(t) versus ¢ curve] is given on the pext page.% ***** Unitatep response ***** A=10 1 0:0 0 1200 -60 -14); grid Titlel’Unit-Step Response of Designed System’) ‘The resulting unit-step response curve is shown below. es ta Regen oOo! Pron 4 ed +——| A cr) + 1 ¥ ca : Tost B29. . . the _state-: dons £¢ y Referring to Section 12-4, a mathemal for the ‘pendulum system shown in Figure 12- 50 is given by (Nem + lo = mlb emh ze =mgla vhich canbe modified to : MLE = Mtma)go -u MZ =u-mgO Since the state variables are defined as Heb, %eebs RX, Keak we get the following state space equations:% o 4 0 Olly, ° Em ML O 0 Ollx, “it + “ a) % a Oo 0 Six o : m 4] Lat 8 0. Ole |e % 4 oO ny lf o Ol” @ de oo ¢ olla, Xy Substituting the given mmerical val ‘the following equations f¢ for the syovent % 0 oe % 1222s 0 0 Oo 0.5" % e oor %| | -zeer 0 0 oO x Hl _|/ o@ 2 ofy”] oe oO} Olly Xl Determination of the state-feedback gain matrix K: Since matrices A and B for \e system are 7” o oo ° 12.2628 0 0.6 B as ? ” o ood - o w2es25 9 0 0 as and the desired closed-loop poles are at - 251 -Sa-#tjt, Sa-@-—j#, S=-20, 2-20 the folieving ae MATLAB program can be written for thé determination of the state feedback gain matrix K. '% *9°* State feedback gain matrix K ***** A={O 1 0 0;12.2625 0 0 0,0 0 0 i;-24525 0 0 9); Bx b.050051 rend 49% 20 a0; - 1.004003 * 4.1381--1.0094 -2.6096 09134 Obtain’ to initial condition: To obtain the system response the lar elat conaitions wo first sunseleete a= Ke into the system equation sAxtpe and get the following equation: 2-8 ‘hich, vhen the numerical values are substituted, can be given by % o a00/ Q ° % H| [20588 -o.s0e7 -n3098 -on SED a r . 3) % ° o ° aeol lx Xp] | 20S 0.5047 = AOR —_ AHSET IIe ‘Let us rewrite Equation (3) as e-As where o 0.00! o ° A -20568 -o.S097 -h309R -0,7567 a= o o ° o.00/ 2.0666 C.SO¥F ASOGR B¢SS7. = 252 -Define the initial condition vector as 8, .or poiving ‘the following ‘The fol: initial g-|° slo / the response of the systen to the initial condition can be obtained by equations: . 4 on gsAzr+Be xn Azr8« MATLAB program will generate the response of the system to the tion. In the MATLAB program we used the following notations: A=, B=ee ‘% —— Response to initial condition — This progrem obteins the response of the Yanai} 20 the given Initial eondlidon (0) % #449 Entor matrices A, B, and K to produce matrix % AA = Anat #088 : A= [0 1 0 0;12.2625 0 0 0,0 0 0 1;2.4625 0 0 O; B = (0;-0.5:0;0.51; K 138. }009.4 -2609.6 -913.4]; AASA ‘% **°** Enter the initial condition matrix BB = Bhat ***** bea Sena xan0 oI 3; x2 1 Bre o x4=(0 00 %'#*"** Plot response curves x1 % and-x4 versus t on one diagram * Foti hero) voisus t vlabel'xt = Theta’ subplot(2, 2,2); tida'x2 (Theta dot) versus t') xlabel("t Sec") Fabel x2 = ‘Teta dor’ — 253 -112,2,31; Paes - 1 Cart) versus ') i jeplacement o xlaballt Sec") ylabell'x3 = Displacement of Cart’) ‘subplot2.2.4; plotit.x4):1 4 (alot of Cer) veraus ‘The resulting response curves are shown below. 24 Tota) vorma 72 (Trt dot veri 02, 5 B+12-10, We shall present three methods for the design of the full-order state cbeerver s We shal first transform the system equations into the cbeervable form, Define a transformation matrix Q by aa(uy” = 254 =where a) is a coefficient in the characteristic equation of the original state equation: . oe -/ st2 a Steast/ |sz-4l = = Stay St ae a3, 4=/ o / a= i o=-Z Define a new state vector § by a-ak ‘Then, the state and output equations become EQ"Age y= ces Sve EDL cant a; J=0 4 ‘The new state and output equations are in the observable canonical form. Re- ferring to Equation (12-61), the state observer gain matrix Ke can be given by At a= of oo"] where of) and % 2 are determined from the desired characteristic equation:(s-Mi)(S-fe) = (S42) (s45) = s*e 08 425 SS*+0/S+%, =O as = /0, % =2e Hence, o s[es-A _[7 &=|! We -(7] Method 2: Define eal = a | eel ‘Then, the characteristic equation of the observer. is Stithe, 9 ~/ SI-At keC] = [sE-A+ kee] “the, St2 = S*4 (Ht he, + 2)5 +/+ Zhe, + hey Sst poste Ithe t2=/0, I+2he the, 25 he = 7, hes =/0 el Next, we shall obtain the observer ‘trix, Aeer- Betted Femuda given to Berion (ise grin wa Be Br use of nore where ¢(s) is the desired characteristic polynomial, or $0) = Cs- MN s—pa) = Gres sts) = St 103 +25 ~ 256 =$() = Att Og F258 ~~ fF 1 f 49) [17 7 el deol dF al fo ate aife 7 7c olte| Lt? Ke=ly pollee of Lathe elke led “Le ‘The equation for the full-order state observer is given by Equation (12-60): 2 = (A-heg) E+ Ba + keg vhere 8 = 0 in this problem. Hence: Eff? Jae olf) ~ f-S Ch ‘hie is the equation for the full-order state obeerver. Bl2-11. A full-order state observer for the given system is designed by use ‘Of MATLAB. The MATLAB program used for the design of the state observer is given i 1% ***** Design of full-order state observer ***** A=[0 100 0 1-5 6 O} Caf 0 ‘Sy ‘Ke acker(A; ‘Warring: Pots locations are moore than 10% in error. Ko=Referring to Equation (12-60), the full-order state observer is given by g = (A~ RCE + Burkey z) [-s5 7 ollzt fo 3 Rla|-see NR +] Ol 4 +] see 19 S| [-r290 -6 ot} [/ sags B-12-12. We shall present two methods for obtaining the full-order state ob- server gain matrix Ke. A MATLAB solution is also given. Referring to 12-61), the state cbeerver matrix method 1 oe ‘ing to Equation ( gain Ke Wy—ds he Q |e. ae “letim ay Matrix Q is given by ga Ceury” vhere Sen Brg ~B30 w iH n=(e* ate* aver] = a2 % 1 We) 4, -¢ 00 4e@0 ‘The values of a; and a2 are determined from the characteristic equation of the original system. s cee a o . -/ |sT-al= WAZER = bIISE SHI SS- SPA BSS = 013959 SH ZKK SSG SH+42S+ay UFBSRS, Wy Rm O39ISE, Aya /2KK - 258 -‘thus O35 9FS B/E / WH) aver ‘ o é oe o ‘Tnerefore | [rears ares: T'ye @ : a= (wary slaves 7 ol=lo 7 SAS ™ ‘ ° 0 1 “BAS 102066) ‘he values of 1, 2, and & 3 are determined from the desired characteristic equation. (s> pu) (s~f C5 3) FC sts NEMS +OHINT)(s +70) = (s'+ fos + 100 )(s4/0) = S7+20s*+ 2005 4+/006 S72 bl S*t+ Wr Stoly =O ‘Tos % F220, = 200 , Ws =lO0 Hence . ws ay oo Z 1000 HAZE Kee Blwemala|o 7 ws rAs || 200 40.3956 Ms ar CBM 1 2866\| 20-37 KS iss S| 47387 SMF IE, Method 2: Define the state observer gain matrix as her ke =| hee = hex ‘The desired characteristic equation becomesSool lo / o IsI-At+ ke) = osol-lo 6 lt ty 023 - oo Ss] [szeg o3tsg —3/45] |hes| Sthke, -I e S| hee s ~ 120g they 63986 St 2. S| SF 4 (he, $I S) S* + (3185 he, then -OFISE)S F CM2KK they THHS he, - 0.3996 he, ) = S24 208° + 2005+ 108 =O hey HRMS = 20 StS he, + Ke, ~~ % FIFE = 200 Th 24K + hey + BARS hey —O39F6 he, = 1000 from which we obtain . fe = PSS, hex @lPT3E7, hes = SREP] ass ke =| 49.380 SKF .E/ Referring to Equation (12-60), the full-order state observer is 2s (A-e)E+Be thes o g 6. 8S / os ° Bt sSS | *|-2397 0 4 WR]t] @ [4+] passe et |%] [-seva7 oases -aeer||S] [neve eK 387 it A MATLAB progrem to obtain the state observer gain matrix Re ‘shown on the next page.‘1% 9**** Design of fill-order state observer ***** As( 1 0,070 1;1246 0.3956 3.145), Call 0 OF B23. gels al ‘The desired closed-loop poles for the pole-placement part are S=—2.97 Lage and the desired observer pole of the minimm-order observer is at Se -$ ‘The first step to design a observer controller is to determine the state feedback gain matrix K and the observer gain matrix Ke. By using the MATLAB program given below ve can determine K and Ke. ‘4 "90" Determination of K and Ke **4* A=(O 1:0 Of B= [015 + | J=[-0.20714j°0.7071 -0.70714°0.7071), K=scke(AB) ~ 261 -‘The feedback gin matrix fad the observer gein matrix Fe are cbtained a follows: Kel! n442], Kea Next, we obtain the transfer function of the observer controller. Noting that the minimm-order observer equation is given by Equation (12-89), we have T= (Abe Maw E+ [CAs Ke hae) Ke + Asa Ke Aue] 9 + (84 —Ke Ba) For the present system, Aa 0, fab=ls Are=O, Abb =0, Ba =O, By=/ Ke=l, kya hery2 , kere By substituting these numerical values into the minimum-order observer equation, we get % = (0-sx/ Ft. [lo-sx$ + 0- $¥0]y + C/-8x0) ue Ba-SF Byte transfc ‘the sion, the Taxing the Laplace form of last equation, assuming zero initial St) =-5 Gs) —26- 10) + Wes) [-25 ye) + 76)] a. fo=s— Referring to Equation (12-104) we have wae - KE =—k, F- (ka tkekedy SHh4/E2 F 9.071 g ‘Taking the Laplace transform of this last equation, we obtain Us) = ~/. 4182 FE) —P.0U Yo) @ Eliminating $ (e) from Bquations (1) and (2), ve have . . e=-n¢r42 Le [-25 VE) + Vey) 8.077 Ye) Simplitying, (4s) Ue) = - 1 $142 [-25 Ye) + VO)] ~ ts) 8-°7/ YE) from vhich we get = 262 -Te) _ _ 80 7/ st = £.07/ SHO EIS Ye) st by se2 *" St 6/¥2 vhich gives the transfer function of the observer controller. The same observer controller equation can be obtained by use of MATLAB. - See the MATLAB. program given below. "°° Design of Observer Couroer 6°" Ace, 30. 10 OB (oaks =; Aba = 0; Abb = 0; Ba= 6; Bb= 1; iL otaainens ‘B-12-14. We shall use the MATLAB approach to solve this problem. The first the page determines the state feedback gain mtrix matrix Ke. The observer to be designed is a full-order i in‘4 ***** Determination of K and Ke *¢*** A=[0 1 09 0 2,6 -l1 4 B= (0;0;1; ° caf 0 Foch 4 5% XK ack(AB.D) ‘The state feedback gain matrix K and the observer gain matrix K_ thus obtained are as follows: k=l# 7 7] 42 Ke =} 2 -72 ‘The second MATLAB program given below determines the transfer function of the observer controller. ‘% Obexinng trnafe Sanction of observer controller = fillorder observer A=[ 1 09 0 1,6 -11 5} B= (00,1), ‘01,0000 119.0000 618.0000 den 1.0000 19.0000 121.0000 257.0000‘The transfer function of the obeerver controller is Ue) _ S*HHPS + E8 YG) ~ 8? + /9s*4 2/5 +237 ‘The transfer function of the given system in state space form is / G6) = ——_—_-__- > SHES EH STE A block diagram of the designed system is shown below. red 7 Notice that the designed system is of gixth order. 312-15. ‘The transfer function of the plant is YO _ _s*t 2st 57 US) Sts t2¢s + The corresponding differential equation is F+OZpr+ 2 y= a+ 2+ sou Comparing thie plant differential equation with the standard third-order differen tial equation G+ugr Ga G tas g = bet by bt be tage we find %=10, Wg =2¢,8,=0, & =0, b =1, b2=2, bs =S0 Define the state variables as follows: mga poe Mad fa Xs 2 - fae pe=be=0 Br= 4~ 4) fo=/ - 265 =Pe *ba ay J, — 4 fy = 2- 0x10 ~-3 a= by —@ifa On fy — As fy = S04 /0X8 — 26K I~ 0 = 106 Refering to Equation (3-36), we have Latte = %y-Pue X= Ah, A Xa 0) Ha t fs UH DERE; Hose ‘The output equation is G=% Hence the state-space representation of the plant ts % ore ‘ Bl=lo 0 s|tl-3 fu %| [ao - -w| [oe % gel 0 e)lx: 3 th a observer: .We now obtain the state feedback gain matrix & and cheerver gain natrix fe vhen the observer is a fuli-order one. The MATLAB Program shown below produces K and Ke. ‘4 9990" Determination of K and Ke for the foll-onder observer **** A=[0 1 60 0 1;0 -24 -105, 106]; (0.50000000000000 -0,09040074557316 .0.03984156570363 L=0 -10 -10; “OLY‘The transfer function of the observer controller can be obtained easily with MATLAB. ‘The MATLAB program given below produces the transfer function of the observer cont- roller vben the observer is of full order. ‘% Obtaining transfix ftnction of observer controller — full-order observer An[O 1 09 0 1,0 24-105 0 12.6915 163,626 500.0000 den = 41,0000 27.0000 218.3085 554.8695 ‘The transfer. function of the observer controller obtained ie WG). __ 1269S 5* + 163-6265 + $00. 0000 -Ye) S!+ 27,0000 5*+ 2/6, 308S°S + SSR ICIS ‘System with 2 minimm-order cbeerver: We next consider the case vhere the state observer is a minimm-order observer. (The state feedback gain matrix K is the same as the case of the full-order observer.) ‘Te folioving MATLAB program produces ‘the observer gain matrix Ke when the observer is the minimm-order observer. '%°**** Determination of Ke for the mimimant-order observer ***** Abb=[0 1-24-10]; Asb= - 267 -The MATLAB program shovn below produces the transfer function of the observer controlier based on the minimm-order observer. .% Obtaining transfer fimction of observer controller — minimam-order observer [0.1 0:0 0 1,0 24 -10}; ‘B= [1;8;106), Scie asain es “Alsat = Abb - Ke*Anby 156%: Xe = 1281 Bb = [93106], ‘The transfer function of the cbeerver controller obtained is Wis) _ 0.8822 s*+ /2.69155 + 50.0000 “ys S* + $8785 + S2.9260 Dalt-step response: The closed-loop trenefer function vben the observer 18 of ‘full order is WO. Ss PASI OYSE S? + 14610 100235" + 91BS./3.5 + 25000 RE) SHEDS T+ SLES + PSPS SF+ 12250 S*4 225003 + 25000 The closed-loop transfer function when the observer is the minimm-order observer is given by S?+ W3S*+ 7245 t2500 Si 27S% + 2ESS?+ (02S3* +4 20008 + 2500 ‘the unit-atep response curves for both cases are-shown in the next page. Notice that the unit-step response curves for both systems are almost identical. Comparison of bandwidths of both systemg: Bode diagrams of both systems are ‘shown in the next page. The bandwidths are almost the same for both systems. ‘The bendwidth for the system with the full-order observer is 2.4771 rad/sec. bandwidth for the system with the miniswm-order observer is 2.4201 rad/sec. = 268 ~im J = Phan (ea: Magna (2) ‘The MATLAB program to obtain the bandwidths of both systems is shown on the next page.{6 994° Comparison of bandwidtha "8 sum] = [12,6915 189.0456 1461.9002 9183.13 25000]; denl = [I 37 525 3575 12250 22500 25000}, B216. The transfer function of the plant is Wet v6) (sts) ‘The corresponding differential equation is dri- Define the state variables by . =F AF ‘Then the state space representation of the plant becomes as follows: E)-E Et eur oft Nov ve obtain the transfer function of the observer controller vith MATLAB. The MATLAB program given on the next page produces the desired observer controller. = 270 =‘%00%6° Obrsinng the transfer Sanction of observer controller **#* As[0 10 -1; B= (051); 119 117 ‘The observer controller obtained with MATLAB is BABPStS/2 GW) = ps t779 Block disgrame for Syetene (2) aot (2) can nor be rotram as shen tn te nt page. system (b), we determined N 80 steady-state output to the unit-step input is unity. ‘The closed-loop transfer function ¥(s)/R(s) for System (a) is Ye). 2697S +S/Z : Sh 205734186 S* + BPE SH SSZ ‘The closed-loop transfer function ¥(s)/R(s) for System (b) is Ye), 4:9 76/ S* + 83.198 95 + 912 RE) SFHROSI + /BES* 4 3EFST SIZUIST SZ St+ 1951/7 (a)1% °° Unit-step response carves ***** sami= [267 512}; deol = [1 20 136 384 512}, 3-12-17. To determine the parameter a in matrix A, ve first determine matrix P from “ ATE+EA=-E or o 0 -/ "Pu Pe Ps Pu Pos Py 0 4 0 100 10-218 fs pal * |Pe fe pr ilo o i/=-lore a1 -4 Why Bs Pn Pa Ps Pall-7 -2 ~ ool ‘The result is : aYeoa 1 za*+3 a 2(2a-1) 2(2a-) 2 Pp =| tates BtAtar 7 abtat/ - 2lwa-/) 2(ta~/) 2lea-“) a abtat/ ats 2 “Ram 20a) ‘Then we can obtain the optimal value of the parameter a that minimizes the performance index J for any given initial condition x(0). Since x(0) is given zeal e o ‘the performance index J can be simplified to T= ZTE xo = pct - 27 -‘Therefore, we obtain asa -/ = Gane To ‘minimize J, we determine a from dJ/da = 0, or Par— ta ~6 (94-238 c* =o fram which we get: Aa/é23 , A= —OP2z Since a is specitied to be positive, we discard the negative vaine of a. ‘Ths, 823. Noting that a = 1.823 satisfies the condition for the Minime (220/éa2 > 0), ‘the optimal value of a is 1.823, Baze. os) 25K as) SHASTHOSHUEK From this closed-loop transfer function, we obtain 2San LS, wat OS +2ASK Since ¢ is given as 0.5, we obtain by HLS =JlOS+25K from vhich ve get K = 0.7, W, = 1.5. ‘Then we obtain LH. L9E- RG) SH ASSH RAS E(s)/R(s) can then be obtained as EG) _ R-C) _ _Steesstos REY RG) ~ Ste Ass +225 E+ ssé+2azce sF+spsi+rasr Since r # 0 in this problem, we have eC +45E +2,28€ =O Define e; =e, eg = 6. Then ve have ~ 274 =6 _| ? te &] |-zes -zslle,| —S"etmat =f" e%@@ cleat co =|, e-[/ ‘ “ Caft)) o 0 Tet us solve ATP + PA=-Q for P. Using thie P we obtain T= ["erwat = 2% Pew Note that for a general case of ve have Coe L P BSan wn zon wm zug aed ‘Tus, \retoat = [e@) ee} he “l- U az] s 1 = Pu = Gan * Gn By substituting $ = 0.5 and @y = 1.5 into this last equation, we obtain Sede: = + Be1Z-19. “The optimal control signal u will have the formu = -R&. Tharg- fore, the performance index J becomes . a=(" @lgteDde = [x re KT xd e ° ae ee - 275 -Since R = 1 in this problem, Byation(12-125) becomes GQ-B"P + PBB) =~ (It RTD and Equation (12-217) becomes . KTR BRA EL ‘where P is determined from the reduced matrix Riccati equation: AE+ LA PeeTe +E =o Solving for P, requiring that it be positive definite, ve obtain ep - oo7 2¢ KBP slo 7 ? ee ral ‘Thus, the optimal control signal u is given by Sok Ea Anh 2-12-20. We first solve the reduced matrix Riccati equation: AL + PALER BELT O =O Noting that matrix A is real and matrix Q is real symmetric, matrix P is a real symmetric matrix. “Hence the reduced mitrix Riccati equation can be written as 2 ell ay pul, far al? 4 1 O3lPn pa fx pPalLo 9 _ [re Ae] [e fh tal [ o]_loo fr mlleae abe hE A]-Le 3] ‘This 2ast equation can be simplified to oe) fe poll we Aa |y]/ elle 2 ba Piz] | O Pra Peper be om oo from whicti we obtain the following three equations:[ahs =e Pam Pra Pox =O MA2pa~ Pa #0 Solving ‘these three simultaneous equations for Pui, P12, and p72 requiring P to be positive definite, we get e=[n a)-["" aa] ‘The optimal feedback gain matrix K is obtained as pb Pe Pu Pr Keeree cise al Pl =A: el=l re] ‘Thus, the optimal control signal is an ~ Kx = % J M+2 ze 212-21. A MATLAB program to solve the given quadratic optimal control ‘pro- ‘blew is shown below. : % ***** Quadcatic optimal control *7>+* A = 10 10 0;20.601 0 0 0;0 0 0 1;-0.4908 0 0 Ol; B= 1 1: 00.8 = Hoo’ @ 80:0 1 0 0:0 0 1 90 00 11 2 a KS qiA8.O,R Ke (64.0554 -11.8078 ‘The state-feedback gain matrix is obtained as follows: Ka=[-smesse ~-bo7¢ —t.000e -2.2965) Next, ve shall obtain the response to the given initial condition. We sub- stitute aan ke into the original state-space equation and obtain the following equation: X=Axts Bu =Ax- BkKx =(A-BKDZ‘The MATLAB program given below produces the response to the given initial condition. Note tint 2 — BAK is written as MA and the initial condition (0-1) 07,0; 0} is Fepresented by BB. The remulting response curves are shown bal next ‘% ***** Responss to initia! condition ***** x4—=10 0 0 1 plottt.x1) {Wtie("Response of x1, Thets') xlabel("t Sec’) ylabell'x1 = Theta’) pottexay ttle Ree saponse of x2, Theta dot) xlabell't ylabolt"x2 = Theta dot’) of x3, Displacement of Cart’) Silo Ret 180 Of x4, Velocity of Cart’) yebeltxa = ‘Velocity of Cart’) Rowan oe, Ta netvienna Al} Upper Saddle River, NJ 07458 https/www.prenhall.com =——! PF SE Ss 4 a=.