Nonlinear Control + Consider the following problem: Fe f(ou),reR" find Eo state feedback yahoueR™ u=T(y)) output feedback 80 that the closed loop system = f(x,r(x)) or = f(x,T(x)) exhibits desired stability and performance characteristics. + Why do we use nonlinear control ~ Tracking, regulate state setpoint Ensure the desired stability properties ~ Ensure the appropriate transients ~ Reduce the sensitivity to plant parameters The Tracking Problem ‘Consider the system: ae flow [Need to accomplish tvo contol objectives: 1) Conttol Objective make x» x, (4, isa desired trajectory), assuming 1,4, L,. 2) Hidden Control Objective-keep everything bounded (ie. x, 8,u€ 1.) [Need to make some assumptions first 1) xis measureable. 2) ifre L,, then (2) ely 3) x has a solution 4) xOEL,The Tracking Problem (continued) Let the tracking error ¢, be defined as Now wecan substitute for — Exact Model Knowledge Now, solve the differential equation e(t) = e(O)exp(—e) Finally, insure all signals are bounded Syiy Lg CEL, 2X1, > fel, uel, Piel, Allsignals are bounded! Example Exact Model Knowledge conto linear Input vc constants Disturbance + Dynamics: + Tracking Control } rive e to zer0 Objective: + Open Loop Error System: Assume + Control Design: we ah are oo known Feedforward Feedback + Closed Loop Error System: + Solution: wa BponontaAdaptive Control Consider a linearity parameterizable finction Constant that can be factored out f= reve fiscal 09 =f xsi} {*) ‘where IV (x) is known, and @ is an unknown constant. By Assumption 2: both fix) and W(x) are bounded, lt 224,-W(9)0-u () =ourcontol where or Yetobe designed, eed wat Wo He @) Siete praetes Let be defined as 3-0-6 Now, combining (1) and (2), we et é=-ke-W(2)0 Adaptive Control (continued) ‘Choose the Lyapunov candidate va Mes Mor Q: Whyis this a good candidate? A: Itis lower bounded (not necessarily by zero) radially unbounded in. positive definite in 2 V “explodes” as e and Lemma: if — Lyapunov-like lemma veo 2) F-<-g(d), where g(20 3) (EL, if A(t) is bounded the g(0) is uniformly continuous then lim g()=0 ‘will use this lemma by getting get e and @ into g and satisfying the conditions on g Note: detailed in deQueiroz.Adaptive Control (continued) ‘With our candidate Lyapunov funtion v= Yes Yar closed-loop Taking the derivative gives error system Re an0 Poa-te78)-076 Vane +8"(H77e-8) design d whe Lyapunov fusion Letting We, we finally get Vente? did get 0 here = aalyssis more complicates ThereforeY € 1, > 6,¢,6,x€L, ue L, = allsignals are tounded! For his problem Now havea dynamic cant ‘g(0=ke? and A(t) 2kee € 1, (Control has dynamics) sompared ‘So our closed loop system is to state- feedback whick ia s static ke-WOand=W'e —_[Q: So,does6 -r02) Az Not necessarily! Sed srox, = cant identify the parameters @ 20) Example Unknown Model Parameters + Open Loop Error cnknown ae temere BURR Design: eee . Same controleras efore,but andre functions orume Howdoweadust and? Use the Lyapunov Stabil is to develep an adaptive control design (ool for compensation of parametric uncertainty parameter + Closed Loop Error System: [At this point, we have not fully developed the controller since and ‘are yet to be determined.Example Unknown Model Parameters Fundamental Theorem pie iif iii) isbounded( is UC Sf tet ot conten an) Bounded £> fiat becomes a} ettocts of conatton constant } is bounded -_ constant + Non-Negative Funetion: J sats contton + Time Derivative of V/V: } examine condition ay ‘substitute the dynamics for Example Unknown Model Parameters + Substitute Error System: ur Howdoweselect and _—_suchthat 2 + Update Law Design: + Substitute in Update } and Laws: Fundamental wm > —_is boundedm=>all signals are _— wm > is bounded bounded a es => }Site ‘Theorem : achieved control } Sistine derived from fF statiity Feedforward FeedbacknalysisHow Can We Use the Adaptive Controller? —— Design adaptive control to tack a desired trajectory while compensating for unknown, ‘constant parameters (parametric uncertainty) ush,(x4) i= s(x0)+u b=) ‘Adaptive control with backstepping in cascaded subsytems to track a des ted trajectory ‘hile compensating for unknown, constant parameters (parametric uncertainty) y= fy(.8,) +m a= f(x8)+y| 7 Backstepping - intermediate controller is adaptive 1», =h(x,6),0=h,(x) How Can We Use the Adaptive Controller? (continued) Adaptive control with backstepping in cascaded subsytems to track a desired trajectory’ while compensating for unknown, constant parameters (parametric uncertainty) ‘Backstepping - intermediate and input controller are adaptive 4 =h(x,6),0=h, (x) and = hy (x,9.,).4, = hey) \ Backstepping - input controller is adaptive iu (ssnds) =o) P= A(0) +m r= (a0) +y i (¥50,) + -—9 = F(x.) +9}How Can We Use the Adaptive Controller? ‘What about the case where input multiplied by an unknown parameter, can we design ‘adaptive control to track a desired trajectory while compensating for unknown, constant parameters (parametric uncertainty) L(x0)+8; Homework A.2-2 Robust Control Recall the system defined bythe following: = footw emi, S)—u ‘We can ty to make several assumptions about the sytem: D ryigels 2) ifre L,, then f(a)e 3) x->x, andall signals are bounded 4) f(2) is lineatily parameterizable (e.,f(2) =”) => Adaptive control ONLY Restriction on the structure but not the uncertainty = We use this assumption for Robust (Sliding Mode) control ONLY!Robust Control (continued) Feedback and feedforward motivated by: €=,—f(x)— Now, let control be_— auxiliary contol wakes k, 4% where Fis function that we can choose. Consider the thre following factions co frequency In reality can't implement tis You of peSliding mode sone) control because it is not really defined at 0. Practically, asin MATLAB, use Vex tees Robust, high gain oe aideg where €>0. Mie will consider each F, separately. Va= => Robust, high frequency ii sol tent ooks Vp md Vy aa response like the high frequeney controller fx ghis mathematically (> frequency) Robust (Sliding Mode) Control Let's try the first function y:,gtferental equation will not have solution ete fC) Fo, Now, take a Lyapunov candidat ae Use assumption 5 here: ([7(9]< a) “he por pose -> less nate bund on 7 Pmebm e(-te- f(2)-Vy Ps-tet Wipers ‘Substitute proposed contro. Vy = p= Note it doesn't exist at e=0 where we are trying to drive the error system <-2kY 3 V+2kV <0 Va 2kV =—s(0), where s()20 differential inequaity — differential equality —> solveRobust (Sliding Mode) Control (continued) Solving the differential equation, we get V(t) =V(O)exp(-2kt)—exp(-2kt) exp(2ke)s(e)dr “~ always negative Ve)sV(OexM-2k) <= Bes fe Oer(-2H) upper bound le(0|s le(Ofexp(-kr) So, the systemis globally exponentially stable, and all signals are bounded! Robust (High-Gain) Control Recall, we proposed u= ke +i, +V_, ‘Now let try it with Pq, andthe same Lyapunov function Same basic proofs Vi “ke= 10)-Vr new robot control term Pecke?+|elp)-e¥n Peote sfelo0y-{ tee ole? ate +l a(x). lleco( Vato’ ae] eine ot npn \, 9) [enh at > 9920Robust (High-Gain) Control (continued) Lyapunov analysis continued, As a reminder we started with ‘This is what the new robust contol term accomplished. sit useful? Pete +Hoco( Case Li ffp)>e ie) a Z z afiHeoe)<0 epcn/ gklew)>1 0 glee) }<0— elect = Ps-ke case it po) se Helo s1-r0 = Ps-tese Viste +6 Ps-24V 46 P-+2kV =e~s(t) Sobeditemetil enation (i befor) P4287 =-(40)-#) 19 (4) <1-»}4oe0( 1-2 lle} sates Robust (High-Gain) Control (continued) Solving the differential equation yiekis igV ert) enya ested eery-2h)fenilde discard this negative term con 20) VE) SV O)exp(-2k)+ 5 [exp(-24)] (0) is less than a constart. eo < Yee exp(-241) + [t-exp(-240)] € = [iexp(-2k9] 0 will gotoaballofsize = bids Prema ‘The system is Globally Uniformly Ulimately Bounded (GUUB),, and all signals are bounded. Signal chasing: ¢x, are bounded > x bounded —> is bounded Canis ¢sll ede he ie of he bal but the trade-offs thatthe control tem Vs