CHAPTER 6 MOTION CONTROL he time history of the general to be developed by the joint actuators so as to gusrantee execution of the commanded ‘count, Finally, 28 premise tothe interaction control problem, the basic features of operational space control schemes ar illustrated, 6.1 THE CONTROL PROBLEM Several techniques can be employed for controlling a manipulator. The technique followed, as well as the way it is implemented, may have a significant influence oon the manipulator performance and then on the possible range of applications. For instance, the need for trajectory tracking control in the operational space may lead to hardware/software implementations w! +r from those allowing point-to-point control where only reaching of the final position is of concern. (On the other hand, the manipulator mechanical designhas an influence on the kind of control scheme utilized, For instance, the control problem of a Cartesian manipulator is substantially different from that of an anthropomorphic manipulator, ‘The driving system of the joints has also an effect on the type of control strategy used, Ifa manipulator is actuated by electric motors with reduction gears of high ratios, 199«9 cyte ann 4 G9) 30 MOKA UT your age sooo 28049 je (uw) 37 22 sub = bey diysuoneyar Suymoy(0} 241 pur pita oq 0} paunsse—suorsstuiswen Jo 10190 otf 10uap “D 197] yey worou! a4 Woy uv $9040 pazqtesoUo8 oy + bt +5(b°) 9 + BIBT 6g paquuosop axe (s eu romMAoI yt aI}>-PUO OU {yo a8pajouy atp Uo worsysasduut Aue 40 (Aaonse[a "ysepy9eG seo ‘uononnstios) asmonays oy) Jo Atureuaoun Aue yey seO[o waKY SE “aIMONS Ise} doo}-tado ue u} payjosues are yoryan ou Soop atays jonoo aoeds wo! B TEE paulisap sj awaups rufauinbai uonout LWO}SueN 0} PaAios St SoneUIOUry esveAu! FoveINdy EEL ‘yy ‘stusa) jesouo# ul “S9ayJa oouequMystp Jo wORNpa! ssoujsngos “at ‘yoeqpaay Aq papiaoad sauneay poo aut oydxe or sdooy pasojo sey woo au ‘Sausoyos tog Ul “(@'9 “Bls) ways yoxsU00 ga0ds youonsado ‘auo4os Jaumuoo 2o0ds juol sig, ‘aoeds yutof oq ur powzopsod ae joniuoa stasayas ‘eonds yeuonwiodo 2m, us sj29yjo-pus) HoREDy ur Jo ad ayfoads axp 0} ws20U09 At ‘goueuuiopsod yBry UrEIqo 0} Jo ]YSHOKR ‘5 01 ancy saYoreas Jonuoo wasaysp‘2ouanbasuos v sy jURAdfas Sow: D "uaamaq sSurtdnoo pu sarieautuou jo 1ysraM ay Inq YsEP}Ieg PUP oy anp syaeqresp of savewT}o S9AUIP TAMIP YH payeNIDe Yoqor F “PUY TaLNO at UO ‘quoy 05 pu 52030} s¥[o409 “wuaU wopuodap-uonesn3yu09 oy anp st ¥f Wer B1OUL soueuiopiad warsAs yn] Kew yey) YsEp{aeg pue AINSE]2 “wonatIy yUIOF Jo souaLsNIG0 (soou0y puv jndrueus yeorue ‘mp st ‘1axomoy ‘Ked of 213d ax, ‘soayJ9 KuseouMuoU Jo woRRNpAL a Jo AaIA UE SIMLOL amp ajdnozap 0) shi pue sorueusp warsAs azuaur| 0} spusy srea Jo souosoud ot yosnoo aon of o auoyps (1802) reau9out femme eam [on [mn foto fa = SS804V-1UINVHLLONOH 0 TORLNOD GNVONTIBGON OZ202 MODELING AND CONTROL.OF ROBOT MANIPULATORS oNuMEaR ecoUPLED FIGURE 63 ‘Bock scheme of manipulator with dives, that the diagonal elements of B(q) are formed by constant terms and By obser fependent terms (functions of sine and cosine for revolute joints), one configu can sot B(g) = B + AB(a) 64) srepresentthe resulting average yields where B isthe diagonal matrix whose cons inertia at each joint. Substituting (6.2)-(6.4) into tm = KP BK im + Fmd + 4 65) where Fn = KO RK> iS represents the matrix of viscous friction coefficients about the motor axes, and d= K7ABQ)K; "in + KOU @KrMam + Ke*9(q) 6-7) represents the contribution depending on the configuration, ‘xr illustrated by the block scheme of Fig. 6.3, the system of manipalator with drives is actualy constituted by two subsystems; one has 7m as input and drs as output daa hat dog dees and Gm as inputs, and d as output, The former is linear and vTorounled, since each component of Tm influences only the corresponding component MoniON cONIKUL 209 of qr. The later is nonlinear and coupled, si coupling terms of manipulator joint dynamics. ‘On the basis of the above scheme, several control algorithms can be derived with reference to the detail of knowledge of the dynamic model. The simplest approach that can be followed, in case of high gear reduction ratios and/or limited performance in terms of required velocities and accelerations, is to consider the component of the nonlinear interacting term d as a disturbance for the single joint servo. ‘The design of the control algorithm leads to a decentralized control structure, since each joint is considered independently of the others. The joint controller must ‘guarantee good performance in terms of high disturbance rejection and enhanced Itajectory tracking capabilities. The resulting control structure is substantially based con the error between the desired and actual output, while the input control torque at actuator i depends only on the error of output i. (On the other hand, when large operational speeds are required to a direct-drive ‘manipulator (K, = 1), the nonlinear coupling terms strongly influence system per- formance. Therefore, considering the effects of the components of d as a disturbance ray generate large tracking errors. In this case, it is advisable to design control al- ‘orithms that take advantage of a detailed knowledge of manipulator dynamics so as to compensate for the nonlinear coupling terms of the model. In other words, itis necessary to eliminate the causes rather than to reduce the effects induced by them; 10 generate compensa This leads to centralized control algorithm: knowledge of the manipulator dynamic model. Nevertheless, it should be pointed out that these of error contributions between the desired and the actual trajectory, they are implemented in a feedback or in a feedforward fashion. of the fact that the considered dynamic model, even though a anyhow an idealization of reality which does not it friction, gear backlash, dimension tolerance, and the simplifying assumption ‘model, e.g. link rigidity, and so on. ‘As pointed out above, the role ofthe drive system is relevant for the type of contro! chosen. In the case of decentralized control, the actuator will be described in terms ofits own model as a velocity-controlled generator. Instead, in the case of centralized control, the actuator will have to generate torque contributions computed on the basis of a complete or reduced manipulator dynamic model; it will be then considered as ‘a torque-controlled generator which is representative of the actuator/power amplifier system satisfying the above requirement. it accounts for all those nonlinear and 6.3 INDEPENDENT JOINT CONTROL as formed by n independent systems (the m joints) and controls each joint axis as a single-input/single-output system. Coupling effects between joints due configurations during motion are treated as disturbance inputs ‘in the case of interest, the system to control is joint drive i corresponding to204 MODELING AND CONTROL OF ROBOT MANIPULATORS “hy FIGURE 64 Block scheme of joint dive system. the single-inpuvsingle-output system of the decoupled and linear part of the scheme ‘6.3. The interaction with the other joints is described by component i of the din (6.7). Without loss of generality, the actuator is taken as a rotary electric de motor. Hence, the block scheme of joint i can be represented in the domain of the complex variable s as in Fig. 6.4. In this scheme 0 is the angular variable of the motor, J is the average inertia reported to the motor axis (J; = bi:/k2,), Ra is the armature resistance (auto-inductance has been neglected), and k, and k, are respectively the torque and ‘motor constants. Further, G, denotes the voltage gain of the power amplifier, and then the reference input is not the armature voltage V, but the input voltage V_ ofthe amplifier; note that the amplifier bandwidth has been assumed to be much larger than that of the controlled system. In the scheme of Fig. 6.4, it has been assumed also that Keak an the mechanical viscous friction coefficient has been neglected with respect to the clectrical friction coefficient. “The input/output transfer function of the motor can be written as cs s(1+8Tm)" Fn & M()= (63) where 1 Ral Kea Tm RR are respectively the velocity-to-voltage gain and time constant of the motor. 6.3.1 Feedback Control ‘To guide selection of the controller structure, start by noticing that an effective rejection of the disturbance d on the output @ is ensured by: "Subscript & has been dropped for notation compactness. Also, Laplace transforms of time-dependent fencions ae indiated by esptal leurs without specifying dependence ons. 2 A complete treatment of actuators is deere to Chapter 8. Marion conTROL 205 emp po e-fe) LE FIGURE 65 Block seheme of general independent joint conta « a large value of the amplifier gain before the point of intervention of the distur- bance, ‘© the presence of an integral action in the gravitational component on the output at steady state (const ‘These requisites clearly suggest the use of a proporti in the forward path whose transfer function is integral (PI) control action 0(6) = KAA, (69) this yields zero erro at steady state for a constant disturbance, and the presence ofthe zero at s = —1/T. offersa 1 action. To improve dynamic performance, it is worth choosing the contr ascade of elementary actions with local feedback loops closed around the disturbance Besides closure ofa position feedback loop, the most general solution is obtained obtain zero error at steady state fr aconstant di are the respective transducer constants, and the amplifier g the gain of the inmost controller, In the scheme of Fig. 6.5, notice that the disturbance torque D has been transformed into a disturbance voltage by the factor Re /ke In the following, a number of possible solutions that can be derived from the general scheme of Fig, 6.5 are presented; at this stage, the issue arising from possible lack of measurement of physical variables is not considered yet. Three case studies are considered which differ inthe numberof active feedback loops.206 MODELING AND CONTROL OF ROBOT MANIPULATORS: FIGURE 6.6 Block scheme of position feedback conto Position Feedback. In this cas, the conteol acti LisTe ky = kra =0 “The scheme of Fig. 6.6 shows that the transfer function ofthe forward path is kn Kp(1+sTp) S+5Tm) is characterized by Cpls) = Kp Gi s)=1 Cals) =2 P(s)= while that of the return path is H(s) = kre. performed as a function of the gain of the position Loop mis are illustrated for the poles of the closed-loop ‘A root locus analysis can kn KpkrpTp/Tm- Three system with reference to the closed-loop feedback system imposes some constr of the Pl controller. If Tp < T must be Tp > T, the two roots ofthe locus tending towards the asymptote render Tp > Tm ( any case, “The closed-loop inpuVoutpu transfer function is ed ow) _ ine. ei 3 FO +Ta) 10 1 EKpkre(+ To) ‘which can be expressed in the form 1 Ell tet) W() = 2 ——. (eS) are MorION CONTROL 207 FIGURE 67 Root loc forthe poston feedback control sche Where wa and ¢ ate respectively the natural frequency and damping ratio ofthe pair cof complex poles and —1/r locates the real pole. These values are assigned to define the joint drive dynamics as a function ofthe constant Tp; if Tp > Tp, then 1/ Cin > Tp > T ig. 6.70); if Tp > Tr, Fi. 6:7), for large values of the loop gain, then cancel the effect of the real pole. ‘The closed-loop disturbance/output transfer function is —— ee p(t stp) te a) E,Kpkre(1 + sTr) ccan be recognized imposed by the feedback gain on the amplitude ofthe output duc| 208 MODELING AND CONTROL OF ROBOT MANIPULATORS LG Lea rcune«s Blok er of psn and ety eck ot the quantity Xn=Kpkrp (6.12) can be interpreted as the disturbance re the gain Kp. However damping ratios would result leading to unacceptable o ate T'z of the output recovery time needed by the cont jons of the output, An system (o recover the by analyzing the - Such estimate is expressed by T,= sans (Tr, 7 }. (6.13) Gon. Position and Velocity Feedback. In this case, the control action is characterized by: Cp(s)=Kp Cys) = Kv. kra To carry out a root locus analysis as a function of the vel is worth reducing the veloc the position loop by following usual rules for moving blocks. From the scheme in Fig, 6.8 the transfer function of the forward pat ‘while that of the return path is H(s) = kre (14 som ) MomioN conTRoL 209 elie FIGURES Root loa forthe poston and velocity feedback contol scheme The zero of the controller at s = —1/Ty can be chosen so as to cancel the effects of the real pole of the motor at s = ~1/Tr,. Then, by setting T= the poles of the closed-loop system move on the root locus as a function of the loop gain ky. Kvkrv, large absolute values of the real part. Then, the actual location can be established by a suitable choice of Ky. ‘The closed-loop input/output transfer function is 1 kre kv, 6.14) Kpkrp * knKpkrpKy which can be compared with the typical transfer function of a second-order system = 6.15) It can be recognized that, with a suitable choice of the gains, any value of natural frequency 1 and damping ratio ¢. Hence, as design requirements, the following relations can be found: possible to obtain and ¢ are given Kykry = 2% (6.16) KpkrpKy = 2% 6.17)Z1U__ MODELING AND CONTROL OF ROBOT MANIPULATORS 4 > a “> Pere Ky 102 oXs) < cure io hed chm of ton, wey al kon fe cn For given transducer constants kp and kry once Ky has been chosen to satisfy (6.16), the value of K’p is obtained from (6.17). ‘The closed-loop disturbance/output transfer function is oR ___EKpkreKy tata) en skry 2 . Kekre * Keke Rv ‘which shows that the disturbance rejection factor is 1+ Xn = KpkrpKy (6.19) and is fixed, once Kp and Ky have been chosen via disturbance dynamics, the presence of a zero at the orig real pole at s = —1, be noticed. Hence, in time constant ) and (6.17). Concerning introduced by the PI, of a ‘m» and of a pair of complex poles having real part Cav, should this case, an estimate ofthe output recovery time is given by the 1 anmeslta ch), sos oe igven ht he pve 2 Fr dcp ee eta Position, Velocity, and Acceleration Feedback. In this case, the control action is characterized by: Crl)= Ke Cu(s)=Ky Cals) = Ky tHe onion CONTROL 211 imkvitrK TF Fn ake After some manipulation, the Mock scheme of Fig. 6:5 canbe reduced to that of ie 610 where (+) inate the following want fonction yy (1-4 bal akra) [ ‘G7 eKakra) Gs) = ‘The transfer function of the forward path is _ KekvK, P(s) s while that ofthe return path sorter (v4 2X). jon is worthy which can be achieved either Aon is case a suitable pole cao by sting _. ortymang km KakpaTa® Tm km Kakra > 1. Thetwosoluionare equivalents In both cases, the poles of the closed analogy with the previous scheme can be recognize system is again of second-order type.