betaFeedback

WorksComputer Science

Constrained multibody kinematics and dynamics in absolute coordinates: a discussion of three approaches to representing rigid body rotation

Abstract

We compare three approaches to posing the index 3 set of differential algebraic equations (DAEs) associated with the constrained multibody dynamics problem formulated in absolute coordinates. The first approach works directly with the orientation matrix and therefore eschews the need for generalized coordinates used to produce the orientation matrix A \mathbf{A} . The approach is informed by the fact that rotation matrices belong to the SO(3) Lie matrix group. The second approach employs Euler parameters, while the third uses Euler angles. In all cases, the index 3 DAE problem is solved via a first order implicit numerical integrator. We note a roughly twofold speedup of rA \mathbf{rA} over rε \mathbf{rε} , and a 1.2 – 1.3 times speedup of rε \mathbf{rε} over rp \mathbf{rp} . The tests were carried out in conjunction with four 3D mechanisms. The improvements in simulation speed of the rA \mathbf{rA} approach are traced back to a simpler form of the equations of motion and more concise Jacobians that enter the numerical solution. The contributions made herein are twofold. First, we provide first order variations of all the quantities that enter the rA \mathbf{rA} formulation when used in the context of implicit integration; i.e., sensitivity of the kinematic constraints for all lower pair joints, as well as the sensitivity of the constraint reaction forces. Second, to the best of our knowledge, there is no other contribution that compares head to head the solution efficiency of rA \mathbf{rA} , rp \mathbf{rp} , and rε \mathbf{rε} in the context of the multibody dynamics problem posed in absolute coordinates.

Introduction

In rigid multibody dynamics simulation using absolute coordinates, several approaches can be employed to track the attitude of a body in 3D motion. This paper is concerned with comparing the performance (accuracy and efficiency) of three approaches. The first, called herein rA, solves the kinematics and dynamics problems by computing the orientation matrix Ai∈R3×3\mathbf{A}_i \in\mathbb{R}^{3 \times3} of each body ii in the mechanical system directly. The other two approaches use Euler parameters (rp) and Euler angles (rε) to express the orientation matrix Ai\mathbf{A}_i. To the best of our knowledge, the results of such a study has not been reported in the literature.

While rA, rp, and rε use different paths to produce the orientation matrix A\mathbf{A}, all three formulations share the same high-level process for generating a numerical solution. Specifically, for dynamics analysis, the numerical solution is produced using a direct index 3 approach that uses a first order implicit integrator to discretize the Newton-Euler constrained equations of motion [1, 2]. Herein, we compare the simulation times and the number of Newton-Raphson iterations taken until convergence of rA, rp, and rε both kinematics and dynamics. We also confirm the first order accuracy of the three dynamics solvers.

Kinematics Analysis

In multibody kinematics and dynamics analyses, the kinematic constraints modeling the lower-pair joints, e.g., revolute joint, spherical joint, hook joint, translational joint, etc., can be formulated in terms of a small set of geometric constraints (GCONs) [3]. The ones chosen herein are called DP1, DP2, D, and CD, and summarized in Table 1.

DP1DP2DCD
Intermediate constraint
Perpendicular 1 (⊥1\perp1)××
Perpendicular 2 (⊥2\perp2)××
Joints
Spherical (SJ)× × ×
Universal (UJ)×[× × ×]SJ_{SJ}
Cylindrical (CJ)[× ×]⊥1_{\perp1}[× ×]⊥2_{\perp2}
Revolute (RJ)[× ×]⊥1_{\perp1}[× × ×]SJ_{SJ}
Translational (TJ)×[× ×]⊥1_{\perp1}[× ×]⊥2_{\perp2}

Table 1. Summary of modeling various joints using four basic GCONs. Perpendicular 1 constrains a vector cj\mathbf{c}_j on body jj to be perpendicular to a plane of body ii defined by two noncolinear vectors ai\mathbf{a}_i and bi\mathbf{b}_i (from where the two × signs). Perpendicular 2 constrains a vector PiQj→\overrightarrow{P_iQ_j} from body ii to body jj to be perpendicular to a plane defined by two noncolinear vectors ai\mathbf{a}_i and bi\mathbf{b}_i.

The DP1 (“dot product 1”) GCON constrains the relative motion between two bodies by requiring that the dot product between a vector attached to body ii and a second vector attached to body jj assumes a specified value. For instance, if the vectors remain at all times perpendicular, this value should stay zero throughout the simulation. DP1 has the following attributes; i.e., it requires the following pieces of information to be properly set up: (i) body ii, the associated local-reference frame L-RFi_i, and the vector ai\mathbf{a}_i that enters the dot product condition; (ii) body jj, the associated local-reference frame L-RFj_j, and the vector aj\mathbf{a}_j that enters the dot product condition; and, (iii) the value that the dot product should assume, which is prescribed through a function f(t)f(t). Most often, f(t)=0f(t)=0, which indicates that the two vectors are orthogonal. If the function ff depends on time, this leads to ΦDP1\Phi^{DP1} being a driving (rheonomic) constraint. The DP1 geometric constraint is captured in the following kinematic constraint equation:

ΦDP1(i,aˉi,j,aˉj,f(t))=aˉiTAiTAjaˉj−f(t)=0.(1)\Phi^{DP1}(i,\bar{\mathbf{a}}_i,j,\bar{\mathbf{a}}_j,f(t))=\bar{\mathbf{a}}_i^{T}A_i^{T}A_j\bar{\mathbf{a}}_j-f(t)=0 . \tag*{(1)}

The DP2 (“dot product 2”) GCON constrains the relative motion between two bodies by requiring that the dot product between a vector aˉi\bar{\mathbf{a}}_i on body ii and a second vector PiQj→\overrightarrow{P_iQ_j} from body ii to body jj assumes a specified value. DP2 has the following attributes: (ii) body ii, the associated L-RFi_i, the algebraic vector aˉi\bar{\mathbf{a}}_i, and the location sˉiP\bar{\mathbf{s}}^P_i of the point PiP_i; (iiii) body jj, the associated L-RFj_j, and the location sˉjQ\bar{\mathbf{s}}^Q_j of the point QjQ_j; and, (iiiiii) the value that the dot product should assume, which is prescribed through a function f(t)f(t). Most often, f(t)=0f(t)=0, which indicates that the two vectors are orthogonal. If the function ff depends on time, this leads to ΦDP2\Phi^{DP2} being a driving (rheonomic) constraint. The DP2 geometric constraint is captured in the following kinematic constraint equation:

ΦDP2(i,aˉi,sˉiP,j,sˉjQ,f(t))=aˉiTAiTdij−f(t)=aˉiTAiT(rj+AjsˉjQ−ri−AisˉiP)−f(t)=0.(2)\begin{aligned} \Phi^{DP2}(i,\bar{\mathbf{a}}_i,\bar{\mathbf{s}}^P_i,j,\bar{\mathbf{s}}^Q_j,f(t))&=\bar{\mathbf{a}}_i^{T}A_i^{T}\mathbf{d}_{ij}-f(t)\\ &=\bar{\mathbf{a}}_i^{T}A_i^{T}\left(\mathbf{r}_j+A_j\bar{\mathbf{s}}^Q_j-\mathbf{r}_i-A_i\bar{\mathbf{s}}^P_i\right)-f(t)\\ &=0 . \tag*{(2)} \end{aligned}

The D (“distance”) GCON constrains the relative motion between two bodies by requiring that the distance between point PP on body ii and point QQ on body jj assumes

a specified value strictly greater than zero. D has the following attributes: (ii) body ii, the associated L-RFi_i, and the location sˉiP\bar{\mathbf{s}}^P_i of the point PP; (iiii) body jj, the associated L-RFj_j, and the location sˉjQ\bar{\mathbf{s}}^Q_j of the point QQ; and, (iiiiii) the value that the distance between the two points should assume, which is prescribed through the function f(t)f(t). Most often, f(t)=C2>0f(t)=C^2>0, which defines a scleronomic kinematic constraint (the power 2 emphasizes that the constant function assumes a non-negative value). If f(t)f(t) depends on time, this leads to ΦD\Phi^D being a driving (rheonomic) constraint. The D geometric constraint is captured in the following kinematic constraint equation:

ΦD(i,sˉiP,j,sˉjQ,f(t))=dijTdij−f(t)=(rj+AjsˉjQ−ri−AisˉiP)T(rj+AjsˉjQ−ri−AisˉiP)−f(t)=0.(3)\begin{aligned} \Phi^D(i,\bar{\mathbf{s}}^P_i,j,\bar{\mathbf{s}}^Q_j,f(t))&=\mathbf{d}_{ij}^{T}\mathbf{d}_{ij}-f(t)\\ &=\left(\mathbf{r}_j+A_j\bar{\mathbf{s}}^Q_j-\mathbf{r}_i-A_i\bar{\mathbf{s}}^P_i\right)^T\left(\mathbf{r}_j+A_j\bar{\mathbf{s}}^Q_j-\mathbf{r}_i-A_i\bar{\mathbf{s}}^P_i\right)-f(t)\\ &=0 . \tag*{(3)} \end{aligned}

The CD (“coordinate difference”) GCON constrains the relative motion between two bodies by requiring that the difference between the xx (or yy or zz) coordinate of point PP on body ii and the xx (or yy or zz) coordinate of point QQ on body jj assumes a specified value. CD has the following attributes: (ii) the coordinate c∈{i,j,k}\mathbf{c}\in\{i,j,k\} of interest; (iiii) body ii, the associated L-RFi_i, and the location sˉiP\bar{\mathbf{s}}^P_i of the point PP; (iiiiii) body jj, the associated L-RFj_j, and the location sˉjQ\bar{\mathbf{s}}^Q_j of the point QQ; and, (iviv) the value that the coordinate difference should assume, which is prescribed through the function f(t)f(t). Note that if f(t)=const.f(t)=\mathrm{const.}, ΦCD\Phi^{CD} defines a scleronomic kinematic constraint. Otherwise, it defines a driving (rheonomic) constraint. Also, often times the second body jj is the ground. In this case, by convention, j=0j=0 (the global reference frame G-RF is attached to body 0). The CD geometric constraint is captured in the following kinematic constraint equation:

ΦCD(c,i,sˉiP,j,sˉjQ,f(t))=cTdij−f(t)=cT(rj+AjsˉjQ−ri−AisˉiP)−f(t)=0.(4)\begin{aligned} \Phi^{CD}(\mathbf{c},i,\bar{\mathbf{s}}^P_i,j,\bar{\mathbf{s}}^Q_j,f(t))&=\mathbf{c}^{T}\mathbf{d}_{ij}-f(t)\\ &=\mathbf{c}^{T}\left(\mathbf{r}_j+A_j\bar{\mathbf{s}}^Q_j-\mathbf{r}_i-A_i\bar{\mathbf{s}}^P_i\right)-f(t)\\ &=0 . \tag*{(4)} \end{aligned}

References

  1. [1]N. Orlandea, M. A. Chace, and D. A. Calahan. A sparsity-oriented approach to the dynamic analysis and design of mechanical systems – part I and part II. Transactions of the ASME Journal of Engineering for Industry, pages 773–784, 1977.DOI
  2. [2]K. E. Brenan, S. L. Campbell, and L. R. Petzold. Numerical solution of initial-value problems in differential-algebraic equations. SIAM Classics in Appl. Math. SIAM, Philadelphia, Second Edition, 1996.DOI
  3. [3]E. J. Haug. Computer-Aided Kinematics and Dynamics of Mechanical Systems Volume-I. Prentice-Hall, Englewood Cliffs, New Jersey, 1989.
  4. [4]J. E. Marsden and T. S. Ratiu. Introduction to Mechanics and Symmetry, volume 17 of Texts in Applied Mathematics. Springer-Verlag, New York, 1994.
  5. [5]Veeravalli Varadarajan. Lie groups, Lie algebras, and their representations, volume 102. Springer Science & Business, 2013.
  6. [6]A Müller and P. Maisser. A Lie-group formulation of kinematics and dynamics of constrained MBS and its application to analytical mechanics. Multibody system dynamics, 9(4):311–352, 2003.DOI
  7. [7]Jonghoon Park and Wan-Kyun Chung. Geometric integration on Euclidean group with application to articulated multibody systems. IEEE Transactions on Robotics, 21(5):850–863, 2005.DOI
  8. [8]Olivier Brûls, Alberto Cardona, and Martin Arnold. Lie group generalized-α time integration of constrained flexible multibody systems. Mechanism and Machine Theory, 48:121–137, 2012.DOI
  9. [9]Zdravko Terze, Andreas Müller, and Dario Zlat ar. Lie-group integration method for constrained multibody systems in state space. Multibody System Dynamics, 34(3):275–305, 2015.DOI
  10. [10]Victoria Wieloch and Martin Arnold. BDF integrators for constrained mechanical systems on Lie groups. Journal of Computational and Applied Mathematics, 387:112517, 2021. Numerical Solution of Differential and Differential-Algebraic Equations. Selected Papers from NUMDIFF-15.
  11. [11]Leonhard Euler. Formulae generales pro translatione quacunque corporum rigidorum. Novi Commentarii academiae scientiarum Petropolitanae, pages 189–207, 1776.
  12. [12]Olinde Rodrigues. De l’attraction des sphéroïdes, Correspondance sur l’École Impériale Polytechnique. PhD thesis, Thesis for the Faculty of Science of the University of Paris, 1816.
  13. [13]Jay Taves, Alexandra Kissel, and Dan Negrut. On an exponential map approach for rigid body kinematics and dynamics applications. Technical Report TR-2020-08: https://sbel.wisc.edu/technicalreports/, Simulation-Based Engineering Laboratory, University of Wisconsin-Madison, 2020.
  14. [14]Arieh Iserles, Hans Munthe-Kaas, Syvert Nørsett, and Antonella Zanna. Lie-group methods. Acta Numerica, 9:215–365, 2000.DOI
  15. [15]Dan Negrut, Rajiv Rampalli, Gisli Ottarsson, and Anthony Sajdak. On the use of the HHT method in the context of index 3 differential algebraic equations of multibody dynamics. ASME Journal of Computational and Nonlinear Dynamics, 2, 2007.
  16. [16]D. Negrut, L. Jay, and N. Khude. A discussion of low-order numerical integration formulas for rigid and flexible multibody dynamics. Journal of Computational and Nonlinear Dynamics, 4:021008–1, 2009.
  17. [17]Jay Taves, Alexandra Kissel, and Dan Negrut. Software and models for the r – A formulation. https://github.com/uwsbel/public-metadata/tree/master/2021/ASME/rA-formulation, 2021. Simulation-Based Engineering Laboratory, University of Wisconsin-Madison.
  18. [18]P. Nikravesh. Computer-aided analysis of mechanical systems. Prentice-Hall, Englewood Cliffs, New Jersey, 1988.
  19. [19]ECMA. The JSON data interchange format. Technical Report ECMA-404, ECMA International, 2013.
  20. [20]R. A. Wehage and E. J. Haug. Generalized coordinate partitioning for dimension reduction in analysis of constrained dynamic systems. J. Mech. Design, 104:247–255, 1982.DOI
  21. [21]Peter Betsch and Sigrid Leyendecker. The discrete null space method for the energy consistent integration of constrained mechanical systems. Part ii: Multibody dynamics. International journal for numerical methods in engineering, 67(4):499–552, 2006.
  22. [22]D. Negrut, R. Serban, and A. Tasora. Posing multibody dynamics with friction and contact as a differential complementarity problem. ASME Journal of Computational and Nonlinear Dynamics, 13(1):014503, 2017.

Paper details

vlnc:W_aw8ehnzn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Not recorded for this paper

View statement
Looking for related papers…
Contents