Jtam.dvi JOURNAL OF THEORETICAL AND APPLIED MECHANICS 42, 1, pp. 139-162, Warsaw 2004 ENERGY METHOD FOR POSITION STABILITY ANALYSIS OF CRITICAL POINTS OF ROBOT MANIPULATORS Przemysław Szumiński Tomasz Kapitaniak Andrzej Stefański Division of Dynamics, Technical University of Łódź e-mail: przeszum@p.lodz.pl An energy method for the position stability analysis of critical points (equilibrium positions) of dynamical systems with an arbitrary finite numberofdegreesof freedomispresented.Themethodproposedconsists in the analysis of the potential energy balance of a perturbation of an arbitrary number of generalised co-ordinates of the dynamical system in order to establish the energy stability of individual degrees of freedom and of the whole system. In the case ofmechanical systems, the position stabilitycriterion,understoodasminimisationof loadsofdriving systems with conservative forces, is discussed. Some examples of application of the method to the position stability analysis of critical points of one- and two-degree-of-freedom systems are analysed. Key words: critical point, load of driving system, position stability 1. Introduction Aphysical system in the equilibriumposition is subject to various types of external perturbations. Position perturbations can be caused both by external loads acting on the system and by inaccuracies of the physical system posi- tioning. The problem of near-critical analysis of a dynamical system subject to a perturbation of generalised co-ordinates has practical meaning for the determination of a tendency of the system behaviour. The dynamical system remaining in the vicinity of the equilibriumposition, such as the critical point, is stable in the positioning sense if there is a tendency towards conservation of the equilibrium position despite the effect of perturbations. Otherwise, the 140 P.Szumiński et al. critical point is unstable. Additionally, in the case of mechanical systems, it is advantageous to position links in such a way so that driving systems are unloaded. Critical points, which are equilibrium positions, are such special configurations ofmechanical system links, in which the work necessary toma- intain the system at rest is minimised, and thus the driving system loading with conservative forces is decreased. In general, among methods of stability analysis of dynamical systems, we can distinguishmethods based on analysis of equations of motion of a system and energymethods. Commonly usedmethods of the stability analysis of cri- tical points are methods based on eigenvalues (Kapitaniak, 2001; Szumiński, 1995; Wiggins, 1990) or Lapunov’s exponents (Kapitaniak, 1991; Kapitaniak and Wojewoda, 2001; Ott et al., 1990; Wiggins, 1990). These methods allow for a qualitative description of the dynamical system behaviour in the vici- nity of the critical point through the statement of its stability or instability. Well-known are also methods related to limitations of solutions to equations of motion, for instance investigations of the stability of solutions to equations of motion in terms of Lagrange’s or Lapunov’s analysis (Kapitaniak, 1991; Gutowski and Świetlicki, 1986; Demidowicz, 1972). In turn, a method based on perturbation equations (Szumiński, 1996, Minorsky, 1962) that consists in analysis of equations of a perturbedmotion allows for analysis of the dynami- cal systembehaviour in the vicinity of the critical point as a function of one or more motion perturbations introduced to the system. Among the energy me- thods,well known is amethod of analysis of derivatives of the potential energy of a dynamical system in the critical point. It is possible to obtain an answer concerning the stability of the dynamical system in the vicinity of the critical point in an easy way on the basis of analysis of the sign of the second deriva- tive of the system potential energywith respect to the generalised co-ordinate for parameters of the critical point (Kapitaniak, 2001; Cesari, 1959; Langha- ar, 1962; Kruszewski et al., 1993; Schuster, 1993). Employing the principle of virtual works, one can show that the necessary and sufficient conditions for the stability of critical points can also be generalised for systems with ma- ny degrees of freedom. A separate group of methods for the investigations of stability of some defined types of dynamical systems are stability criteria, for example Routh-Hurwitz’s criterion, which are used for the stability analysis of dynamical systems described by linear equations of motion with constant coefficients (Kheir, 1996). The method proposed here is an energy method. It allows one to answer the question concerning the position and energy stability of the critical point, and to carry out quantitative analyses of the dynamical system behaviour in Energy method for position stability analysis... 141 the vicinity of the critical point as a function of the number andmagnitude of the perturbation of generalised co-ordinates, without necessity of formulation of differential equations of motion. Themethod rests on recording the pertur- bation potential energy of a dynamical system and on the static analysis of its changes in the vicinity of the critical point. It allows one to determine the character of motion in the vicinity of the critical point. The paper consists of four sections. Theoretical grounds of themethod are discussed in the secondpart.The third sectionpresents examples of the stabili- ty analysis and determination of characteristics of critical points of exemplary dynamical systems. Dynamical systems with one and two degrees of freedom have been assumed as examples. In the fourth part conclusions are included. 2. Theoretical grounds Let us assume that a vector of generalised co-ordinates of a dynamical system assumes the following form q= [q1,q2, ...,qi, ...,qn] ⊤ (2.1) where qi ∈M and M is the set of perturbed generalised co-ordinates of the dynamical system. Let us now introduce a perturbation of the ith generalised co-ordinate in the form qi = qi+ϕi (2.2) where ϕi is the perturbation of the ith generalised co-ordinate of the system. Let us assume that the relation expressing the potential energy of the no- minal systemposition as a function of its vector of the generalised co-ordinates is known.Thepotential energy, taking into account the effect of aperturbation of selected generalised co-ordinates, is expressed as follows Ep =A+Z(ϕi) (2.3) where A are potential energy terms dependent on the location of the nominal system in the system state space and the data concerning the mass and geo- metry of the system structure.Here, the potential energy terms describing the nominal position of the dynamical systemare included; Z(ϕi) are terms of the potential energy equation describing perturbations of the dynamical system position. 142 P.Szumiński et al. In general, the potential energy of the dynamical system is expressed as Ep =E n p +E p p (2.4) where Enp is the potential energy of the nominal position of the dynamical system, and Epp – potential energy of the nominal position perturbation. As a result of the assumption of the analysis in the close vicinity of the dynamical system equilibrium points, the trigonometric functions describing the potential energy are replaced by a Taylor series. The obtained relations represent changes of the potential energy of the perturbation in vicinities of individual configurations in the system state space as a function of the per- turbation magnitude of individual generalised co-ordinates and critical point parameters in the space of motion parameters of the dynamical system. The perturbation potential energy is understood as the work that should be performed to bring the perturbed dynamical system to the equilibrium position. In order to carryout ananalysis of the systemdynamics inneighbourhoods of critical points, the stability criteria based on the following cases are given. The dynamical system is stable in the position sense in the critical point if each generalised co-ordinate is stable in this point, that is to say, when ∀i E p pi > 0 i=1,2, ...,n The dynamical system is unstable in the critical point if at least one gene- ralised co-ordinate of the dynamical system is unstable, i.e. when ∃i E p pi < 0 i=1,2, ...,n We can distinguish the following cases of the sign of the perturbation potential energy of the dynamical system: • for Epp < 0 a loss of the system potential energy resulting form the occurrence of perturbations; energy should be delivered to the system (work should be performed), the equilibrium point is unstable – some generalised co-ordinates can be stable; when upset from the equilibrium position, the system escapes from the critical point; • for Epp > 0 an increase in the system potential energy resulting from the occurrence of perturbations; energy can be taken from the system as a whole, the equilibriumpoint can be stable or unstable, depending on the signs of values of potential energies of the perturbation Eppi of individual generalised co-ordinates of the system; in the case of the stable system, when upset from the equilibrium position, it tends to return to it; Energy method for position stability analysis... 143 • for Epp =0 lack of perturbations, i.e. the perturbation potential energy equals to zero; in the case when perturbations take place, a stability boundary of the dynamical system in the energy sense occurs. We can distinguish the following cases of the sign of the perturbation potential energy of the ith generalised co-ordinate of the dynamical system: • for Eppi < 0 when the position is upset, energy should be supplied to the system (work should be performed) to bring the system to the equ- ilibrium position; the generalised co-ordinate is unstable; in mechanical systems this case is disadvantageous due to an increase in the load of the driving system of the generalised co-ordinate; • for Ep pi > 0 after the position is upset, the system returns to the equili- briumposition; energy can be taken back from the system; the generali- sed co-ordinate is stable; inmechanical systems this case is advantageous due to the fact that the driving system of the generalised co-ordinate self-unloads. Apart from the presented position stability criterion of the vicinity of the critical point, an energy criterion can be assumed. It is as follows: a dynamical system is stable in the energy sensewhen thebalance of the energydelivered to the system and taken back from it (i.e. its individual generalised co-ordinates) in order tobring the system to the critical point is positive, that is to say,when the potential energy of the system perturbations is positive. This criterion has beenassumed in the analysis of the energynecessary tomaintain the system in the critical point, and to eliminate stresses related to theposition perturbation in driving systems. As a result of numerical simulations, one can perform an analysis of chan- ges in the potential energy with respect to the perturbation location of sub- sequent generalised co-ordinates of a dynamical system as a function of the perturbationmagnitude, and for the case when: • perturbations of a selected generalised co-ordinate of the dynamical sys- tem under consideration occur; • perturbations of a higher number or of all generalised co-ordinates of the system under consideration occur. As a result, some information on types and characters of individual critical points of the dynamical system analysed can be, for instance, obtained. In thenext section, examples of the application of themethod to the analy- sis of types of characteristic points of sample dynamical systems is presented. 144 P.Szumiński et al. 3. Analysis of characteristic points of a double pendulum In the presented example only gravitational forces occur. The potential energy of the double pendulum in Fig.1 with respect to the X-axis is Ep = 1 2 g[m1L1+m2(2L1+L2)−L1(m1+2m2)cosq1−m2L2cosq2] (3.1) where g is the gravity acceleration. Fig. 1. Double pendulum Nowwe introduce a perturbation of each generalised co-ordinate, Eq. (2.2) q1 = q1+ϕ1 q2 = q2+ϕ2 (3.2) Introducing Eq. (3.2) into Eq. (3.1), the potential energy is Ep =A+Bcos(q1+ϕ1)+C cos(q2+ϕ2) (3.3) where the terms of the potential energy describing the basic motion are A= 1 2 g[m1L1+m2(2L1+L2)] and the termsof thepotential energydescribing theperturbationare as follows B=− 1 2 L1(m1+2m2)g C =− 1 2 m2L2g Energy method for position stability analysis... 145 The potential energy of the system connected with the perturbations, Eq. (2.4), is equal to Epp =Bcos(q1+ϕ1)+C cos(q2+ϕ2) (3.4) Now the trigonometric functions in Eq. (3.4) are replaced by a Taylor series, and the series terms higher than four are rejected. Then, the potential energy of the system perturbations is equal to Epp = −B [(ϕ21 2 − ϕ41 24 ) cosq1+ ( ϕ1− ϕ31 6 + ϕ51 120 ) sinq1 ] + (3.5) − C [(ϕ22 2 − ϕ42 24 ) cosq2+ ( ϕ2− ϕ32 6 + ϕ52 120 ) sinq2 ] If we assume the following values of the system parameters m1 =1kg m2 =1kg L1 =0.2m L2 =0.2m the coefficients B and C are equal to B=−2.943Nm C =−0.981Nm Taking into account the parameters of critical points of the system, the po- tential energy of perturbations, Eq. (3.5), assumes forms presented inTable 1. Table 1 Parameters of the system Potential energy of perturbations in the critical point of the system q1 =0, q2 =0 E p p =−B (ϕ21 2 − ϕ41 24 ) −C (ϕ22 2 − ϕ42 24 ) q1 =0, q2 =π E p p =−B (ϕ21 2 − ϕ41 24 ) +C (ϕ22 2 − ϕ42 24 ) q1 =π, q2 =0 E p p =B (ϕ21 2 − ϕ41 24 ) −C (ϕ22 2 − ϕ42 24 ) q1 =π, q2 =π E p p =B (ϕ21 2 − ϕ41 24 ) +C (ϕ22 2 − ϕ42 24 ) The expressions inTable 1 allow for an analysis of changes of the potential perturbation energy in the close vicinity of equilibriumpositions as a function of the perturbationmagnitude ϕ1 and ϕ2. 146 P.Szumiński et al. 3.1. Examples of numerical analysis In Fig.2, cases of stability and instability of the system in the vicinity of individual critical points for a perturbation of the generalised co-ordinate q2 are presented. Fig. 2. Perturbation potential energy distribution in the case of an unstable and stable point The sign of theperturbation energydecides about stability of the system in thevicinity of the equilibriumposition, according to theassumptionspresented in Section 2. A perturbation of one generalised co-ordinate fulfils the basic criterion of stability. Let us consider an influence of both the generalised co-ordinates on the system stability. In Fig.3, the vicinity of the equilibrium position: q1 = 0, q2 =π is shown. The potential energy of the system perturbation has a positive value, Epp > 0. The system satisfies the energy criterion of stability. However, the system is unstable as the generalised co-ordinate q2 is unstable, E p p2 < 0. On the other hand, the potential energy of the system perturbation in the vicinity of the critical point q1 = q2 =0 is presented in Fig.4. The system is stable near the critical point because both the generalised co-ordinates of the system are stable, Epp1, E p p2 > 0, that is to say, the potential energy of the system perturbation is positive, Epp > 0. The system fulfils also the energy criterion of stability. Energy method for position stability analysis... 147 Fig. 3. Energy stability of the equilibrium q1 =0, q2 =π Fig. 4. Stability of the equilibrium q1 =0, q2 =0 In Fig.5 the potential energy of the system perturbation in the vicinity of the critical point q1 = π, q2 = 0 is shown. In this case, the generalised co-ordinate q1 is unstable, E p p1 < 0, whereas the co-ordinate q2 is stable, E p p2 > 0. The system under consideration is unstable, which also follows from the fact that the value of the potential energy of the system perturbation is negative, Epp < 0. In this case, the system is not stable in terms of energy. 148 P.Szumiński et al. Fig. 5. Unstability of the equilibrium q1 =π, q2 =0 4. Pendulum with an elastic support Let us consider a systempresented inFig.6. Thepotential energy of conse- rvative forces in the form of forces of gravity and forces of elasticity, expressed in a system of the co-ordinates XY , is as follows Ep = 1 2 kq2− 1 2 mgL(1− cosq) (4.1) Fig. 6. Elastically suported pendulum Asweknow, equatinghepotential energyderivative,Eq. (4.1),with respect to the variable q to zero,weobtain the equilibriumpoints q=0anda solution in the form 2kq=mgLsinq (4.2) Energy method for position stability analysis... 149 In such cases, this approach is useful and easy. Let us assume that the stiffness coefficient k is variable. For k> 1 2 mgL there is only one equilibrium position q=0, and it is stable. For k< 1 2 mgL the system has three equilibrium positions: q = 0 that is unstable, and two equilibrium positions that are stable and determined by the relation 2kq=mgLsinq Now, let us introduce a perturbation of the generalised co-ordinate q= q+ϕ into the potential energy equation, Eq. (4.1), where ϕ is a perturbation of the position of the system under consideration in the equilibrium point. Then, the potential energy of the system is as follows Ep = 1 2 k(q+ϕ)2− 1 2 mgL[1− cos(q+ϕ)] (4.3) Expanding the trigonemetric function in Eq. (4.3) into a Taylor series and rejecting the terms of higher orders, we obtain Ep =E n p +E p p where Enp is the nominal potential energy of the system Enp = 1 2 [kq2−mgL(1− cosq)] whereas Epp is the potential energy of the perturbation Epp = kϕ ( q+ ϕ 2 )− 1 2 mgL [ cosq (ϕ2 2 − ϕ4 24 ) +sinq ( ϕ− ϕ3 6 + ϕ5 120 )] (4.4) In the case of the equilibrium point q=0, the perturbation potential energy, Eq. (4.4), has the form Epp = 1 2 [ kϕ2− 1 2 mgL ( ϕ2− ϕ4 12 )] (4.5) 150 P.Szumiński et al. Fig. 7. Transition from the unstable system to the stable one For given m,Lof the systemunder consideration, the stiffness coefficient k decides whether the critical point is stable or not. In Fig.7, the distribution of the perturbation potential energy as a function of the stiffness coefficient k and the perturbationof the systemgeneralised co-ordinate, basedonEq. (4.5), is shown. The following data of the system have been assumed: m=1kg, L=1m. A transition from the unstable system, Epp < 0, to the stable one, E p p > 0, can be seen with an increase in the stiffness k. The boundary stability occurs for the stiffness coefficient k=4.905Nm/rad. Fig. 8. Stability of the equilibrium q=0 Energy method for position stability analysis... 151 In Fig.8, changes in the perturbation potential energy in the vicinity of the critical point q = 0 for the stiffness k = 6Nm/rad are presented. The perturbation potential energy is positive, so the system is stable in the vici- nity of the equilibrium point. In turn, Fig.9 shows the perturbation potential energy in the vicinity of the equilibriumpoint q=0 for the stiffness coefficient k=4Nm/rad. Fig. 9. Unstability of the equilibrium q=0 Fig. 10. Identification of the stability boundary for the equilibrium q=0 The perturbation potential energy is negative, so the system is unstable in the critical point. Identification of the stability boundary for the equilibrium point q=0 is performed throughananalysis of the change in the perturbation potential energy of the system. In Fig.10 an energymap of the vicinity of the equilibrium point is shown. It allows for the identification of the stability boundary of the system, which can be stable or unstable, through finding a 152 P.Szumiński et al. value of the parameter atwhich a bifurcation occurs – a transcritical fork-type bifurcation in the case under consideration. In the figure, the system stability boundary as a function of the position perturbation is shown as well. In the case of dynamical systems with a higher number of degrees of freedom than one, the energymap refers to the energy criterion of stability. In other words, it is the balance of energy needed to bring the dynamical system to the critical point. For the value of the stiffness coefficient k corresponding to the system sta- bility boundary (boundary value of the stiffness coefficient k between stability and instability of the system), the perturbation potential energy is positive, and thus the system remains stable – Fig.11. Fig. 11. Perturbation of the system for the coefficient k corresponding to the stability boundary Fig. 12. Stability of the system for the equilibrium 2kq=mgLsinq In turn, changes in the perturbation potential energy as a function of the perturbation in the vicinity of the equilibriumpoint described by the solution (4.2), for the stiffness coefficient k=4Nm/rad, are presented in Fig.12. Energy method for position stability analysis... 153 As theperturbationpotential energy is positive, the system remains stable. While drawing a plot of the changes in the perturbation potential energy as a function of the stiffness coefficient k, one can determine the system stability boundary – Fig.13. Fig. 13. Finding the stability boundary of the system It can be seen that the equilibrium point described by the solution (4.2) vanishes for the stiffness coefficient k=4.905Nm/rad. 5. Analysis of characteristic points of an exemplary robot manipulator Let us consider a robot manipulator with three degrees of freedom as an example. A scheme of such a manipulator is shown in Fig.14. It is a four- link manipulator with the following kinematic chain structure: R-R-P (R – rotational kinematic pair, P – prismatic kinematic pair). In order to simplify the mathematical model, ideal kinematic pairs of the manipulator have been assumed. In the assumed manipulator model, conservative forces in the form of gravity forces occur. Now let us consider a relation describing the potential energy of the ma- nipulator in the second stage of motion (motion of the second and third gene- ralised co-ordinate, whereas the first co-ordinate remains stationary). 154 P.Szumiński et al. Fig. 14. Scheme of the robot manipulatorMAR Let us introduce perturbations of the co-ordinates.We obtain then a rela- tiondescribing themanipulatorpotential energy, in the second stageofmotion, in the form Ep2 =A+Z(ϕ2,ϕ3) (5.1) where A – terms of the potential energy equation describing basic motion; factors depending on the robot configuration and on themass and geometrical data of its links; in the case under consideration A= ( − 1 2 m0L0+ 1 2 m1L1cosq1+m2L1cosq1+m3L1cosq1)g Li – length of the ith manipulator link mi – mass of the ith manipulator link Z – terms of the potential energy equation describing themanipulator perturbation or transforming Eq. (5.1) Ep2 =A+Bcos(q2+ϕ2)+C sin(q2+ϕ2)+Dϕ3 sin(q2+ϕ2) (5.2) Energy method for position stability analysis... 155 where B= (1 2 m2+m3 ) L2gcosq1 C =m3(s3+q3)gcosq1 (5.3) D=m3gcosq1 and s3 is the initial position of the centre of gravity for link 3, with regard to the position of point C – Fig.14. Thus, when the perturbation occurs, we can express the manipulator po- tential energy as a sumof themanipulator nominal potential energy andof the potential energy, called the potential energy of themanipulator perturbation, i.e., generally in the form Ep2 =E n p2+E p p2 (5.4) After somemathematical transformations and taking into account the clo- se vicinity of manipulator equilibrium points, the trigonometric functions are replaced by aTaylor series, and the series terms higher than four are rejected. Then, on the basis of Eq. (5.1), the potential energy of the system perturba- tions is E p p2 = −(Bcosq2+C sinq2) (ϕ22 2 − ϕ42 24 ) + − (B sinq2−C cosq2−Dϕ3cosq2) ( ϕ2− ϕ32 6 + ϕ52 120 ) + (5.5) + Dϕ3 sinq2 ( 1− ϕ22 2 + ϕ42 24 ) If we assume the values of the parameters of individual equilibriumpoints, then on the basis of Eq. (5.3), we obtain the values of the coefficients B, C, D of Eq. (5.5) in the neighbourhood of manipulator individual equilibrium points. A collection of the obtained results is presented in Table 2. Employing the data included in Table 2, Eq. (5.5) assumes the form E p p2 =− B sinq2 120 ϕ52+ Bcosq2 24 ϕ42+ B sinq2 6 ϕ32− Bcosq2 2 ϕ22+ −B sinq2ϕ2+Dsinq2ϕ3+ Dcosq2 120 ϕ52ϕ3+ Dsinq2 24 ϕ42ϕ3+ (5.6) − Dcosq2 6 ϕ32ϕ3− Dsinq2 2 ϕ22ϕ3+Dcosq2ϕ2ϕ3 156 P.Szumiński et al. Table 2 Maniulator Parameters of the critical Coefficient configuration point in kinematic pairs B C D q1 =0 High q2 =0 (m2/2+m3)L2g 0 m3g q3 =−s3 q1 =0 High q2 =π (m2/2+m3)L2g 0 m3g q3 =−s3 q1 =π Low q2 =0 −(m2/2+m3)L2g 0 −m3g q3 =−s3 q1 =π Low q2 =π −(m2/2+m3)L2g 0 −m3g q3 =−s3 Equation (5.6) allows for an analysis of changes of the manipulator per- turbation potential energy as a function of the perturbation magnitude of individual degrees of freedom.We can distinguish here the casesmentioned in the theoretical introduction. In the case of a high configuration of the manipulator, and at the genera- lised co-ordinate q2 = 0 as well as in the case of a low configuration of the manipulator, and at the generalised co-ordinate q2 =π, Eq. (5.6) assumes the form E p p2 = 1 2 Bϕ22 ( 1 12 ϕ22−1 ) +Dϕ2ϕ3 ( 1 120 ϕ42ϕ3− 1 6 ϕ22+1 ) (5.7) In the case of a high configuration of the manipulator, and at the genera- lised co-ordinate q2 = π as well as in the case of a low configuration of the manipulator, and at the generalised co-ordinate q2 =0, Eq. (5.6) assumes, in turn, the form E p p2 = 1 2 Bϕ22 ( 1− 1 12 ϕ22 ) −Dϕ2ϕ3 ( 1 120 ϕ42ϕ3− 1 6 ϕ22+1 ) (5.8) 6. Examples of numerical analysis results Analysing Fig.15-Fig.18, one can observe changes in the potential energy of the manipulator location perturbation as a function of the perturbation magnitude, and the case when: Energy method for position stability analysis... 157 • a perturbation of one of the generalised co-ordinates of themanipulator occurs – in this case it is the generalised co-ordinate q2, see Fig.15 and Fig.16 • perturbations of a higher number of the generalised co-ordinates occur – in the considered case these are q2 and q3, see Fig.17 and Fig.18. Fig. 15. Perturbation potential energy distribution in the case of an unstable point Fig. 16. Perturbation potential energy distribution in the case of a stable point InFig.17 andFig.18 a distribution of the perturbation potential energy as a function of the two-dimensional perturbation ϕ2 and ϕ3 for selected cases of equilibrium points is presented. The perturbations of the location of themanipulator links in the following ranges ϕ2 =±0.01 rad ϕ3 =±0.0005 rad 158 P.Szumiński et al. Fig. 17. Perturbation potential energy space distribution in the case of an unstable point Fig. 18. Perturbation potential energy space distribution in the case of a stable point have been introduced into the system under analysis, [Asada 1986]. Taking into account the stability criteria mentioned in theoretical intro- duction, the types of individual critical points of the considered manipulator are presented in Table 3. Energy method for position stability analysis... 159 Table 3 Distribution of perturbation Kind of critical point potential energy Figure 15 Unstable point (anti-attractor) Figure 16 Stable point (attractor) Figure 17 Energy and position unstable point Figure 18 Energy and position stable point The notion of an attractor is understood as the effect of ”attraction” of a trajectory of changes of theperturbationpotential energyby the critical point, whereas the notion of an anti-attractor refers to the property of the critical point consisting in ”repulsing” the trajectory of the perturbation potential energy changes. The analysis of Eqs (5.7) and (5.8) and of Fig.17 and Fig.18 allows one to observe that in the case of lack of the perturbation of the critical point of the second generalised co-ordinate a perturbation of the critical point related to the third co-ordinate does not affect the magnitude of the perturbation potential energy of the manipulator. Thus, one can state that the behaviour of the system in the neighbourhood of the critical point is influenced by the perturbation of the second generalised co-ordinate of the manipulator. In Fig.15 and Fig.16, the character of the potential energy distribution of the manipulator perturbation for the case of a zero perturbation of the ma- nipulator third generalised co-ordinate in vicinities of various types of critical points, in the second stage of robot motion, is presented. The figures confirm the obtained results referring to the types of individual critical points deter- mined by the method of eigenvalues and by the motion perturbation method (Kapitaniak andWojewoda, 2001; Szumiński, 1995, 1997). Table 4 Manipulator link number i 1 2 3 Kind of kinematic pair R-rotational R-rotational P-prismatic Linkmassmi [kg] 12.7 12.7 15 Link lenghtLi [m] 0.17 0.18 0.42 Position of xrsi [m] −L1/2 −L2/2 0 the centre yrsi [m] 0 0 0 of gravity zrsi [m] 0 0 −L3/2 S3 =0.18 m 160 P.Szumiński et al. Stability simulations for the cases of the determined critical points of the low configuration of the manipulator (q1 = π) yield analogous results to the presented simulation for the high configuration (q1 =0). Physical and geometrical data of the analysedmanipulator are included in Table 4. 7. Conclusions Thepresentedmethod of the potential energy perturbation of a dynamical system allows for the analysis of position and energy stability in the critical point vicinity as a function of the perturbation of an arbitrary number of the system generalised co-ordinates. It enables simulation of the dynamical beha- viour for various cases of generalised co-ordinate perturbations, including the effect of perturbations of individual generalised co-ordinates on the dynamical system stability in the critical point. Formulation of equations of motion of the system is not required. Themethod is based on the analysis of changes in the potential energy of the system upset from the equilibrium position. The formsandkindsof critical points are closely related to themanipulator model, which has been assumed in the investigations. This method can be applied for the determination of control algorithms of motion in terms of subsequent generalised co-ordinates of the dynamical system in order tomaintain the conditions of its position and energy stability in the critical point. References 1. Asada H., 1986,Robot Analysis and Control, Wiley, NewYork 2. Cesari L., 1959, Asymptotic Behaviour and Stability Problems in Ordinary Differential Equations, Springer Verlag, NewYork 3. Demidowicz P., 1972, Mathematical Theory of Stability, PWN Warsaw (in Polish) 4. Gutowski R., Swietlicki W.E., 1986, Dynamics and Vibration of Mecha- nical Systems, PWNWarsaw (in Polish) 5. KapitaniakT., 1991,ChaoticOscillations inMechanical Systems,Manchester University Press,Manchester, England Energy method for position stability analysis... 161 6. Kapitaniak T.,Wojewoda J., 2001,Bifurcations andChaos, PWNWarsaw (in Polish) 7. Kapitaniak T., 2001, Introduction to Vibration Theory, PWN Warsaw (in Polish) 8. Kheir N.A., 1996, Systems Modeling and Computer Simulation, Marcel Dekker, Inc., USA 9. Kruszewski J., Wittbrodt E., Walczyk Z., 1993, Mechanical System Vibration in Computer Aspect, PWNWarsaw (in Polish) 10. Langhaar H.L., 1962, Energy Methods in Applied Mechanics, Willey, New York 11. Minorsky N., 1962, Nonlinear Oscillations, D.van Nonstrand Company, London 12. Ott E., Grebogi C., Yorke J.A., 1990, Controlling chaos,Physical Review Letters, 64, 1196-1199 13. Schuster H.G., 1993,Deterministic Chaos, PWNWarsaw (in Polish) 14. Szumiński P., 1995,The analysis of stability of characteristic points of robots, Proc. VIII Int. Conf.Microcomputers in Science andTechnology, Poland-Czech Republic, 80-86 15. Szumiński P., 1997, Analysis of selected nonlinear phenomena of dynamics of manipulators with flexibility, Doctoral Thesis, Łódź, Poland (in Polish) 16. Szumiński P., 1996, Perturbations of equations of motion versus stability, Inter. Conference of Applied Chaotic Systems, Inowłódz, Poland 17. Wiggins S., 1990, Introduction to Applied Nonlinear Dynamical Systems and Chaos, Springer Verlag Metoda energetyczna analizy stateczności punktów krytycznych manipulatorów Streszczenie Wartykuleprzedstawionometodę energetycznąanalizy statecznościpunktówkry- tycznych (położeń równowagi) układów dynamicznych o dowolnej skończonej liczbie stopni swobody. Proponowanametoda polega na analizie bilansu energii potencjalnej zaburzenia dowolnej liczby współrzędnych uogólnionych układu dynamicznego w ce- lu okreslenia stateczności energetycznej poszczególnych stopni swobody oraz całego 162 P.Szumiński et al. układu.Wprzypadku układówmechanicznych przedstawiono kryterium stateczności w sensie minimalizacji obciążeń układów napędowych siłami zachowawczymi. Przed- stawionoprzykłady zastosowaniametodydla celówanalizy stateczności punktówkry- tycznych układów o jednym i dwóch stoopniach swobody. Manuscript received September 8, 2003; accepted for print October 28, 2003