adaptive resolution of 1d mechanical b-spline
DESCRIPTION
Adaptive resolution of 1D mechanical B-spline. Julien Lenoir , Laurent Grisoni, Philippe Meseure, Christophe Chaillou. Problem. Real-time physical simulation of a knot. Fixed resolution simulation. Goal: adaptive resolution simulation. Related work. 1D model and knot tying - PowerPoint PPT PresentationTRANSCRIPT
Adaptive resolution of 1D mechanical B-spline
Julien Lenoir, Laurent Grisoni, Philippe Meseure, Christophe Chaillou
Problem
Real-time physical simulation of a knot
Fixed resolution simulation Goal: adaptive resolution simulation
Related work
1D model and knot tying [Wang et al 05] Mass-spring model, not adaptive [Brown et al 04] Non physics based model, ‘follow
the leader’ rules, not adaptive Generality on multiresolution physical model
Discrete model: [Luciana et al 95, Hutchinson et al 96, Ganovelli et al 99]
Continuous model: [Wu et al 04, Debunne et al 01, Grinspun et al 02,Capell et al 02]
Outline
Physical simulation of 1D B-spline Geometric subdivision of a B-spline Mechanical multiresolution Results Side effect Conclusion
Physical model: Lagrange formalismVariational formulation
+ Mechanical system defined via DOF
= Energy minimization relatively to DOFs
Physical simulation of 1D B-spline
Geometric model: B-spline
n
kkk sbtts
0
)()(),( qPqk=(qkx,qky,qkz) position of the kth control pointsbk are the spline base functions
t is the time, s the parametric abscissa
Physical simulation of 1D B-spline Physical Model:
Definition of the DOFs:iq
Physical simulation of 1D B-spline Physical Model:
Definition of the DOFs: Lagrange equations applied to B-spline:
i
i
iq
EQ
q
K
dt
d
)(
iq
Kinetic energyEnergies not derived from a potential(Collisions, Frictions…)
Potential energies(Deformations, Gravity…)
Physical simulation of 1D B-spline Physical Model:
Definition of the DOFs: Lagrange equations applied to B-spline:
i
i
iq
EQ
q
K
dt
d
)(
iq
i
zyx q ),,( AAAA
M
M
M
00
00
00
Μ dssbsbm j
s
iij )()(Μ
BA Μ
),,( zyx BBBB
Generalized mass matrix:
Gather the and termsi
Q iq
E
Physical simulation of 1D B-spline Continuous deformation energies
Stretching [Nocent01]: Green-Lagrange tensor allows large deformations Piola-Kirchhoff elasticity constitutive law
Bending: in progress… Twisting: not treated (need to extend the model to
a 4D model)
dss
sl
s
sl
s
tslrYtE
end
begin
stretch
)(
.1)(
/),(
8
.)( 0
22
0
Physical simulation of 1D B-spline Constraints by Lagrange multipliers (λi):
Direct integration into the mechanical system:
E
B
λ
A
0L
LM T
Physical simulation of 1D B-spline Constraints by Lagrange multipliers (λi):
Direct integration into the mechanical system:
λi links a scalar constraint g(s,t) to the DOFs
E
B
λ
A
0L
LM T
Physical simulation of 1D B-spline Constraints by Lagrange multipliers (λi):
Direct integration into the mechanical system:
λi links a scalar constraint g(s,t) to the DOFs L and E are determined via the Baumgarte
scheme:
E
B
λ
A
0L
LM T
012
2
g
tgt
g
Time step
=> Possible violation but no drift
Physical simulation of 1D B-spline
Resulting physical simulation: - 6 constraint equations - 33 DOF
Lack of DOF in some area
Geometric problem
Geometric subdivision of a B-spline Exact insertion in NUB-spline (Oslo algorithm):
NUBS of degree d
Knot vectors:
i
idjij bb~
,
sinon
~si
0
1 10,
jjjji
ttt
rji
iri
rjrirji
iri
irjrji tt
tt
tt
tt,1
11
1,
1,
~
~
~
01,
0,
1 )1( idiii
diii qqq
)(sbit it 1it
insertion
)(~sbit~ it
~1
~it 2
~it
The simplification of BSplines is often an approximation
Mechanical multiresolution
Insertion and suppression: Reallocate the data structure: pre-allocation Shift the pre-computed data and re-compute the
missing part (example: , ) Continuous stretching deformation energies:
Pre-computed terms: 4D array Sparse Symmetric
ds
sbq
sbsbsbsbB
n
jjj
qpmiimpq
3
1
)()0(
)()()()(
Μ 1Μ
Avoid multiple computationStorage in an 1D array
Mechanical multiresolution
Criteria for insertion: Geometric problem => geometric criteria Problem appears in high curvature area Fast curvature evaluation based on control points
Criteria for suppression: Segment rectilinear
1.
2
1
11
11
iiii
iiiii qqqq
qqqqC
Results
Adaptive resolution
Low resolution
High resolution
Geometric property: Multiple insertion at the same location decrease
locally the continuity => ‘Degree’ insertions
= C-1 local continuity = Cutting
Mechanical property: Dynamic cutting
without anythingspecial to handle
Side effect
Side effect
Example of multiple cutting:
Conclusion
Real-time adaptive 1D mechanical model Continuous model (in time and space)
=> Stable over time=> Can handle sliding constraint [Lenoir04]
Dynamic cutting appears as a side effect Future works:
Enhance the deformation energies:Better bending + Twisting (4D model, cosserat [Pai02])
Handle length constraint