## Interactive Manipulation of Articulated Objects with Geometry Awareness (1999)

Venue: | In Proceedings of IEEE International Conference on Robotics and Automation |

Citations: | 2 - 1 self |

### BibTeX

@INPROCEEDINGS{Choi99interactivemanipulation,

author = {Min-hyung Choi and James F. Cremer},

title = {Interactive Manipulation of Articulated Objects with Geometry Awareness},

booktitle = {In Proceedings of IEEE International Conference on Robotics and Automation},

year = {1999},

pages = {592--598},

publisher = {}

}

### OpenURL

### Abstract

Techniques for interactive 3D manipulation of articulated objects in cluttered environments should be geometrically aware, going beyond basic inverse or forward kinematics to allow contact while preventing interpenetration. This paper describes a general purpose interactive object manipulation technique using nonlinear optimization. The method converts geometry awareness into sets of inequality constraints and handles nonlinear equality and inequality constraints efficiently without restricting object topology. Our iterative algorithm has a quadratic convergence rate and each iteration can be solved in O(n nz (L)), where n nz (L) is the number of non-zeros in L, a Cholesky factor of a sparse matrix. To promote additional speedup, symbolic factorization is separated from numerical computation. Our approach provides a framework for using optimization techniques in interactive tools for building and manipulating models in constrained, cluttered environments. 1 Introduction Many interact...

### Citations

278 |
Simulating Humans: Computer Graphics, Animation, and Control
- Badler, Phillips, et al.
- 1993
(Show Context)
Citation Context ...lue in computer graphics and virtual environments applications. In computer graphics, there have been a number of attempts to address the problem of manipulation of redundant articulated objects. Jack=-=[5]-=- uses a reduced coordinate system and nonlinear optimization techniques to solve inverse kinematics for a human model. Zhao[17] added collision avoidance to Jackbut this simple augmentation using boun... |

253 | An approximate minimum degree ordering algorithm
- Amestoy, Davis, et al.
- 1996
(Show Context)
Citation Context ...ined such that the ll-in in the Cholesky decomposition L is minimized. Since the problem of choosing an ordering that minimizes ll-in is NP-hard[2], weuse the approximate minimum degree heuristic(AMD)=-=[1]-=-. Second, the non-zero structure of the Cholesky factor is determined and sparse storage is allocated accordingly. This takes care of indexing and associated bookkeeping so that the actual numerical f... |

166 | V-Clip: Fast and Robust Polyhedral Collision Detection
- Mirtich
- 1998
(Show Context)
Citation Context ...ple \interpenetrating/not interpenetrating" results, while others compute minimum distances without robustly identifying nearest features. We use a new geometric computing package developed by Mi=-=rtich[12]-=- that provides very fast collision detection, distance computation, and nearest feature identi cation. In our manipulation algorithm, interpenetration checks are performed at every iteration. When int... |

161 |
Practical methods of Optimization (2nd ed
- Fletcher
- 2000
(Show Context)
Citation Context ... from a starting point at which allconstraints except the goals are satis ed. Assuming that inequality constraints have been replaced by equalities as described in Section 5.5, the Lagrangian function=-=[10] i-=-s de ned as L(q� )=E(q)+ X i i(q): (3) The rst order necessary condition for the existence of a local minimizer q of the objective function of Equation 1 requires the existence of a Lagrange multipl... |

89 | W.: Interactive dynamics
- WITKIN, GLEICHER, et al.
- 1990
(Show Context)
Citation Context ...bjects. The work investigated the generation of stable non-interpenetrating con gurations of curved-surface objects. Interactive dynamics simulation may also be used as a basis for object manipulation=-=[6, 15]-=-. Interactive dynamics simulation can provide fairly physically accurate results, accounting for friction and inertia in addition to geometry awareness. As good haptic feedback interfaces are develope... |

69 | Implementation of interior-point methods for large scale linear programs
- ANDERSEN, GONDZIO, et al.
- 1996
(Show Context)
Citation Context ...uping columns using the supernode technique, all columns in the supernode update exactly the same position in the subsequent columns. Furthermore, loop unrolling is possible by employing the supernode=-=[3]-=-. With most common memory architectures having at least 2 levels of cache, loop unrolling leads to speedup by taking advantage of locality of memory access. Since the loop from step 4 to step 6 is per... |

27 |
An interactive tool for placing curved surfaces without interpenetration
- Snyder
- 1995
(Show Context)
Citation Context ...n objects in close proximity. Gleicher's[11] di erential method casts the graphical manipulation problem to a nonlinear optimization problem, linearizing it to make the problem easier to solve. Snyder=-=[13]-=- addresses geometry awareness but not for articulated objects. The work investigated the generation of stable non-interpenetrating con gurations of curved-surface objects. Interactive dynamics simulat... |

24 | On a homogeneous algorithm for the monotone complementarity problems. Mathematical Programming 84(1999)375–400
- Andersen, Ye
(Show Context)
Citation Context ...ke advantage of small system size of the normal equation and use e cient LL T Cholesky factorization. Studies show that the Cholesky factorization based on normal equation method is usually preferred =-=[16]-=- if the upper left corner block of the linear system 8 is a diagonal matrix. 5.2 The dense columns in JJ T To avoid dense decomposition we employ the modi ed Schur complement method[4]. The main idea ... |

10 |
The apos linear programming solver: an implementation of the homogeneous algorithm. CORE Discussion paper 9337, CORE, Universite Catholique de Louvain
- Andersen, Andersen
- 1997
(Show Context)
Citation Context ... computation. First, row and column orderings are determined such that the ll-in in the Cholesky decomposition L is minimized. Since the problem of choosing an ordering that minimizes ll-in is NP-hard=-=[2]-=-, weuse the approximate minimum degree heuristic(AMD)[1]. Second, the non-zero structure of the Cholesky factor is determined and sparse storage is allocated accordingly. This takes care of indexing a... |

7 |
A graphical, language-based editor for generic solid models represented by constraints
- Barford
- 1987
(Show Context)
Citation Context ...nts. Furthermore, incorporation of inequality constraints is not very di cult in our optimization approach. Equation 1 represents an objective function to achieve the \Principle of Least Astonishment&=-=quot;[7]-=- which suggests the system should move as little as possible from the previous con guration. E(q) = 1 2 kq ; q0k (1) 2 Combining this objective function E(q) with kinematic equality constraints, i(q),... |

7 |
Near real-time body awareness
- Zhao, Badler
- 1994
(Show Context)
Citation Context ...address the problem of manipulation of redundant articulated objects. Jack[5] uses a reduced coordinate system and nonlinear optimization techniques to solve inverse kinematics for a human model. Zhao=-=[17]-=- added collision avoidance to Jackbut this simple augmentation using bounding spheres and cylinders does not support precise geometry awareness to allow contact or prevent interpenetration between obj... |

6 |
J.F.: Geometric Awareness for Interactive Object Manipulation
- Choi, Cremer
- 1999
(Show Context)
Citation Context ...nstraints. To ensure e ciency, it is important tokeep the set of active inequality constraints as small as possible. As mentioned in Section 4.2, details of our geometric analysis are presented in [9]=-=[8]-=-. The monitoring and management ofinequalities are important for the overall performance, because constraint addition or deletion can alter the block structure and necessitate further symbolic manipul... |

4 |
Linear�time dynamics using lagrange multipliers
- Bara��
(Show Context)
Citation Context ...bjects. The work investigated the generation of stable non-interpenetrating con gurations of curved-surface objects. Interactive dynamics simulation may also be used as a basis for object manipulation=-=[6, 15]-=-. Interactive dynamics simulation can provide fairly physically accurate results, accounting for friction and inertia in addition to geometry awareness. As good haptic feedback interfaces are develope... |

4 |
Symmetric quasi-de nite matrices
- Vanderbei
- 1991
(Show Context)
Citation Context ...rse, due to the existence of single dense column. The other solving technique is to solve 8 directly using LDL T factorization, because it is always sparse regardless of the structure of J. Vandervei =-=[14]-=-showed that any symmetric permutation of a quaside nite matrix has a factorization LDL T and proposed an e - cient sparse algorithm. The main advantage of this approach is that it is not hampered by a... |

3 |
A modi ed Schur-complement method for handling dense columns in interior point methods for linear programming
- Andersen
- 1996
(Show Context)
Citation Context ...ally preferred [16] if the upper left corner block of the linear system 8 is a diagonal matrix. 5.2 The dense columns in JJ T To avoid dense decomposition we employ the modi ed Schur complement method=-=[4]-=-. The main idea is to rewrite the normal equation system 10 into 2 4 (AsA T s + FF T ) AD F A T D ;I 0 F T 0 I 32 54 3 5 = 2 4 b 0 0 3 5 (11) where As and AD are sparse and dense columns of J respecti... |

2 |
A Di erential Approach to Graphical Manipulation
- Gleicher
- 1994
(Show Context)
Citation Context ...ut this simple augmentation using bounding spheres and cylinders does not support precise geometry awareness to allow contact or prevent interpenetration between objects in close proximity. Gleicher's=-=[11]-=- di erential method casts the graphical manipulation problem to a nonlinear optimization problem, linearizing it to make the problem easier to solve. Snyder[13] addresses geometry awareness but not fo... |