Title: An Implicit TimeStepping Method for QuasiRigid Multibody Systems with Intermittent Contact
1An Implicit Time-Stepping Method for Quasi-Rigid
Multibody Systems with Intermittent Contact
- Nilanjan Chakraborty, Stephen Berard, Srinivas
Akella, and Jeff Trinkle - Department of Computer Science
- Rensselaer Polytechnic Institute
2Introduction
- Dexterous manipulation/Grasping
- Mechanical design
- Computer Games
Example Grasping experiment where the Circular
lock piece must be grasped by the parallel jaw
grippers as they close. (Brost and Christensen,
1996)
3Related Work (Dynamics Modeling)
- Differential Algebraic Equation (DAE)
- (Haug et al. 1986)
- Differential Equations (Motion model) Algebraic
Constraints - Requires knowledge of contact interactions
(sliding, rolling or separating) - Contact Interactions not known apriori!
- Differential Complementarity Problem (DCP)
- (Stewart and Trinkle 1996 Anitescu, Cremer and
Potra 1996 Pfeiffer and Glocker 1996 Trinkle et
al. 1997 Trinkle, Tzitizouris and Pang 2001) - Differential Equations (Motion model)
Complementarity constraints (Contact Model,
Friction Model) Algebraic Constraints
4Contact Modeling Assumptions
- Point Contact
- Quasi-Rigid or locally compliant objects each
object has a rigid core surrounded by a thin
compliant layer. (Song and Kumar, 2003, Song,
Kraus, Kumar and Dupont, 2001) - Deflections are small
- There is a maximum deflection beyond which the
body acts as a rigid body - Lumped parameter model of compliance (Force based
methods) - Kelvin-Voigt Model
- Hertz Model
- Hunt and Crossleys Model
5Motivation Integrate Collision Detection
Collision Detection
Solve Dynamics
Update Position
6Motivation Disc rolling on plane
- Eliminate sources of instability and inaccuracy
- Polyhedral approximations
- decoupling of collision detection from dynamics
- approximations of quadratic Coulomb friction model
100 vertices
10 vertices
7Motivation Disc rolling on plane
Discretization of geometry and linearization of
the distance function lead to a loss of energy in
current simulators
Results of simulating the rolling disc using the
Stewart-Trinkle algorithm for varying number of
edges and varying step-size. The top horizontal
line is the computed value obtained by our
geometrically implicit time-stepper using an
implicit surface representation of the
disc. (Error Tolerance 1e-06)
8Motivation Integrate Collision Detection
Collision Detection
Goal Integrate collision detection with
equations of motion
Solve Dynamics
Assumption Convex objects described as an
intersection of implicit surfaces.
Update Position
9Complementarity Problem
10Continuous Time Dynamics Model (Rigid Body)
Newton-Euler equations
Mass Matrix
Contact forces and moment
Applied force
Coriolis force
Kinematic map
Contact constraints
Friction Model Set of Complementarity
Constraints
Joint Constraints Set of Algebraic Constraints
11Discrete Time Dynamics Model (Rigid Body)
Discrete Time Model
(Stewart and Trinkle 1996, uses linearized
friction cone, subproblem at each time step is
LCP)
(Tzitzouris 2001 for closed form distance
functions, subproblem at each time Step is NCP)
12Compliant Contact Model (Quasi-Rigid Body)
(For tangential directions)
13Contact Constraints (Rigid Body)
Contact Constraint in Configuration Space
14Contact Constraints (Rigid Body)
From KKT Conditions
15Contact Constraints (Quasi-Rigid Body)
16Discrete Time Dynamics Model
The mathematical model is a Mixed Nonlinear
Complementarity Problem.
17Results
Rigid Ellipsoid Falling on a deformable half-plane
The MCP solver PATH was used to solve this
problem.
18Results
19Results
20Conclusion and Future Work
- We presented a geometrically implicit
time-stepper for quasi-rigid multi-body systems
that combines the collision detection and dynamic
time step to deal with a source of inaccuracy in
dynamic simulation. - Address the question of existence and uniqueness
of solutions - Implementation for intersection of surfaces.
- Extend to nonconvex implicit surface objects
described as an union of convex objects as well
as parametric surfaces. - Precisely quantify the tradeoffs between the
computation speed and physical accuracy
21THANK YOU!
22Results
23Results
24Results
25Future Work
- Address the question of existence and uniqueness
of solutions - Implementation for intersection of surfaces.
- Extend to nonconvex implicit surface objects
described as an union of convex objects as well
as parametric surfaces. - Precisely quantify the tradeoffs between the
computation speed and physical accuracy