0% found this document useful (0 votes)
43 views25 pages

Shape Optimization Among Convex Sets

This document discusses numerical shape optimization among convex sets. The goal is to find the optimal shape given a shape functional and constraints that guarantee an optimal set exists. Challenges include ensuring perturbations remain admissible and handling non-local constraints. Previous works have used support functions, level sets, and explicit parameterizations. The author proposes using a spectral decomposition of the support function or direct parameterization to represent shapes. Numerical optimization is then applied to minimize the objective subject to convexity and other constraints. Examples are provided to demonstrate the approach.

Uploaded by

Beni Bogosel
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
43 views25 pages

Shape Optimization Among Convex Sets

This document discusses numerical shape optimization among convex sets. The goal is to find the optimal shape given a shape functional and constraints that guarantee an optimal set exists. Challenges include ensuring perturbations remain admissible and handling non-local constraints. Previous works have used support functions, level sets, and explicit parameterizations. The author proposes using a spectral decomposition of the support function or direct parameterization to represent shapes. Numerical optimization is then applied to minimize the objective subject to convexity and other constraints. Examples are provided to demonstrate the approach.

Uploaded by

Beni Bogosel
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 25

Numerical shape optimization among convex sets

ANR SHAPO – Roscoff

Beniamin Bogosel

CMAP, École Polytechnique

14/06/2022

Beniamin Bogosel Numerical shape optimization among convex sets 1/25


Context

min J(ω)
ω∈A

Q: existence, regularity, find the optimal shape, qualitative results, numerics


Problems of interest: considering constraints like:
convexity
diameter
inclusion
constant width
minimal width
or any combination...
Handle arbitrary shape functionals: geometric, PDE constraints, eigenvalues

Beniamin Bogosel Numerical shape optimization among convex sets 2/25


Examples
1. Blaschke-Lebesgue Theorem. Among planar shapes of constant width the
Reuleaux triangle minimizes the area.

2. Blaschke-Lebesgue in 3D (open). The three dimensional body of constant width


with the minimal volume is one of the Meissner bodies.

Beniamin Bogosel Numerical shape optimization among convex sets 3/25


Numerical optimization in this context

Purpose
given a functional and a set of constraints that guarantee existence of an optimal
set: what the optimal shape looks like?
confirm/disprove conjectures, give new directions for theoretical study
Questions
how to discretize shapes?
level-set, phase field, explicit parametrization, polygonal approximation
handle all constraints of interest
evolve of the shape? shape derivative? gradient?

Beniamin Bogosel Numerical shape optimization among convex sets 4/25


Possible numerical difficulties
? not all small perturbations are admissible: flat part of a convex boundary
ω

? some constraints are non-local: diameter or constant width

Beniamin Bogosel Numerical shape optimization among convex sets 5/25


Previous works on the subject

[Lachand-Robert, Oudet, 05] intersection of half-spaces


[Oudet et al. 07, 13] width constraints
[Bayen, Henrion 12] support function in 2D, Fourier coefficients, quadratic functionals
[Merigot, Oudet, 14] projection on the convexity constraint
[Bartels, Wachsmuth 18] convexity constraint: mesh deformation with research of
admissible perturbations

Beniamin Bogosel Numerical shape optimization among convex sets 6/25


Support function
? [Schneider, Convex bodies]
? replace the shape variable ω ⊂ Rd with the function variable p : Sd−1 → R

p(θ) = max(x · θ)
x∈ω
? geometric interpretation: distance to the tangent bounding ω in the direction θ
? can naturally handle width constraints: look at p(θ) + p(−θ)!
? Convexity constraint (in 2D): functional inequality p + p00 ≥ 0

p(θ2 )

p(θ1 )
p(θ3 )

Beniamin Bogosel Numerical shape optimization among convex sets 7/25


Examples of support functions

Beniamin Bogosel Numerical shape optimization among convex sets 8/25


Explicit parametrization: support function

Advantages
exact representation of the shape
precise information on geometric elements
can handle efficiently the constraints of interest
convexity
constant width
inclusion
diameter inequalities

Objective
Given a general shape functional, its shape derivative and a set of constraints, find an
approximation of the optimal shape!

Beniamin Bogosel Numerical shape optimization among convex sets 9/25


Parametrization

? r(θ) = (cos θ, sin θ), t(θ) = r0 (θ) = (− sin θ, cos θ) p(θ2 )


? p denotes the support function
? if ω is strictly convex: a parametrization of ∂Ω is p(θ1 )
given by p(θ3 )
x(θ) = p(θ)r(θ) + p0 (θ)t(θ).

Options
The choice of p and p0 gives a parametrization of ω:
spectral decomposition: direct access to p and p0 - limited to strictly convex sets!
direct choice of values for some angle discretization: how to choose p0 rigorously?

Beniamin Bogosel Numerical shape optimization among convex sets 10/25


Spectral decomposition
? decompose p using Fourier series in 2D, spherical harmonics in 3D
? [Bayen, Henrion], Semidefinite programming for optimizing convex bodies under width
constraints]: completely rigorous, limited applicability: quadratic functions
? [Bogosel, Henrot, Lucardesi, Minimization of the eigenvalues of the Dirichlet-Laplacian with a
diameter constraint]: general 2D framework
? [Antunes, Bogosel, Parametric Shape Optimization using the Support Function]: 2D, 3D

? F: a shape functional with shape derivative F 0 (Ω)(V ) = ∂Ω f V · n dσ


R

? find sensitivity of F with respect to the parameters defining p

Handling the constraints:


convexity 2D: p + p00 ≥ 0 or p(θi ) + p00 (θi ) ≥ 0 for a discretization of [0, 2π]
convexity 3D: Gaussian curvature positive on a finite family of points on the sphere
constant width: p(θ) + p(−θ) = w – some coefficients are 0.
diameter inequalities, inclusion, etc.
Discrete problem: minimization under pointwise and linear inequality constraints.
Beniamin Bogosel Numerical shape optimization among convex sets 11/25
Numerical Optimization

start from randomized coefficients: avoid local minima


compute the gradient: derivatives with respect to each parameter
compute gradients of the (eventual) non-linear constraints
use the Hessian matrix when available, LBFGS otherwise.
use a constrained optimization algorithm: fmincon in Matlab or IPOPT in FreeFEM

Beniamin Bogosel Numerical shape optimization among convex sets 12/25


Examples
min λk (Ω) with diameter Meissner’s conjecture Minimal volume rotors
constraint

[Antunes, B.]
[B., Henrot, Lucardesi]

Beniamin Bogosel Numerical shape optimization among convex sets 13/25


Segments in the boundary
? [Schneider, Convex bodies]: singularities in the support function of a convex set
correspond to segments in the boundary
? do not assume that the support function is smooth!
? on the smooth parts, the choice of p0 (θ) should be consistent!

Initial try: Take pi = p(θi ), θi = 2πi


N , i = 0, ..., N − 1 as parameters. Take h = 2π/N
and use finite differences approximations:
pi+1 − pi−1 pi+1 + pi−1 − 2pi
p0 (θi ) ≈ p00 (θi ) = .
2h h2
? even number of discretization points: diameter inequalities can be expressed with

wi ≤ pi + pi+N/2 ≤ Wi .

? diameter constraint: upper bound D for all directions, lower bound D for one
direction.
Beniamin Bogosel Numerical shape optimization among convex sets 14/25
Example

[Al Sayed, B., Henrot, Nacry, Maximization of the Steklov Eigenvalues with a Diameter
Constraint]
? Maximize the Steklov eigenvalues under diameter constraint:

−∆u = 0 in Ω
∂n u = σk (Ω)u on ∂Ω
? Example: with and without convexity constraint

? applying the same method to other functionals works very well. Why?
Beniamin Bogosel Numerical shape optimization among convex sets 15/25
Discrete configuration

? pi are the values of the support functions at θi = 2πi


N .
? Denote qi the approximations for the derivatives
? Denote %i the approximations for the radii of curvatire: p(θi ) + p00 (θi ) ≈ %i
A3 A2
? choosing qi determines a polygon A0 ...AN −1 . Is q3
this polygon convex assuming %i ≥ 0?
q2
? when using finite differences: the answer is NO
p3
pi+1 − pi−1 pi+1 + pi−1 − 2pi p2
qi = , %i = pi +
2h h2 A1
p1 q1

1
Area(∆A1 A2 A3 ) = [6%1 %2 + 12%1 %3 + 6%2 %3 + (p1 − p3 )(%1 − %3 )] h3 + O(h5 )
48

Beniamin Bogosel Numerical shape optimization among convex sets 16/25


Path towards a rigorous idea

? p(θ) 7→ p(θ) + a cos θ + b sin θ corresponds to a translation with vector (a, b).
? using first order finite differences:
translation does not commute with the discretization process
discrete curvature radii are not preserved under translations
? Choose the simplest option which fixes this:
pi+1 − pi−1 pi+1 + pi−1 − 2pi
p0 (θi ) ≈ , p(θi ) + p00 (θi ) ≈ %i = pi + .
2 sin h 2 − 2 cos h
? observe that the formulas are still consistent as h → 0.
? discrete shapes obtained are convex provided %i ≥ 0!

Area(∆A1 A2 A3 ) = [%2 (%1 + %3 ) + 2%1 %3 cos h] sin2 (h/2) tan(h/2).

Beniamin Bogosel Numerical shape optimization among convex sets 17/25


Further geometric arguments

? This is the only natural discretization


X(pi+1 , θi+1 )

pi+1
Ai
pi+1 −pi−1
pi 2 sin h
O
pi+1 +pi−1 X(pi , θi )
2 cos h

pi−1

X(pi−1 , θi−1 )

? pi+1 + pi−1 − 2pi cos h ≥ 0: Ai is on the correct side of the tangents at Ai−1 , Ai+1 .
−pi−1
? pi+1
2 sin h is the only tangential coordinate that always produces a point on the correct
side of the tangents at Ai−1 , Ai+1
? all convex shapes can be approximated arbitrarily well with the method proposed!

Beniamin Bogosel Numerical shape optimization among convex sets 18/25


Before going further: the gauge function
? if θ 7→ ρ(θ) is a radial function parametrizing Ω then
γ(θ) = 1/ρ(θ) is the gauge function of Ω.
? why? Ω is convex is characterized by γ + γ 00 ≥ 0.
? the gauge function of the convex body Ω is the support function of the polar body
Ω◦ = {y ∈ Rd : x · y ≤ 1, ∀x ∈ Ω}

Beniamin Bogosel Numerical shape optimization among convex sets 19/25


Numerical setting: gauge function

? discretize γ using γi = γ(θi ) = 1/ρ(θi ) as before.


Question: What is the correct discrete convexity constraint?
? straightforward finite differences for γ + γ 00 ≥ 0 are not ok!

(γi−1 + γi+1 − 2γi cos h) sin h


Area(∆Ai−1 Ai Ai+1 ) = .
2γi−1 γi γi+1
? the correct condition is: γi−1 + γi+1 − 2γi cos h ≥ 0
? the same as for the support function!

Beniamin Bogosel Numerical shape optimization among convex sets 20/25


Examples of application
Maximize λk (Ω) under Maximize λk (Ω) under
min λk (Ω): area and constant width constraint: minimal width constraint:
convexity constraints
[Henrot, Oudet] Conjecture Conjecture
[Antunes, Henrot] The Reuleaux triangle is The equliateral triangle is
optimal for 1 ≤ k ≤ 10. optimal for 1 ≤ k ≤ 10.

λ2 (Ω∗ ) = 37.9855
Obtain best known shape
without any effort!

Beniamin Bogosel Numerical shape optimization among convex sets 21/25


PDE constrained functional
? taken from [Bartels, Wachsmuth 18]
Z
min u such that − ∆u = fi , u ∈ H01 (Ω), Ω is convex
Ω Ω

Beniamin Bogosel Numerical shape optimization among convex sets 22/25


The Cheeger inequality

? Minimize λ1 (Ω)/h(Ω)2
? compute the Cheeger constant of Ω at every iteration
? compute the Cheeger set
? compute the associated shape derivatives
? the optimal shape found numerically is the square.

Beniamin Bogosel Numerical shape optimization among convex sets 23/25


Combining support and gauge functions
? one can parametrize K and its polar K ◦ using the same set of parameters and the
same set of constraints
? Mahler problem in 2D: minimize |K| · |K ◦ |

general: |K| · |K ◦ | = 6.750 symmetric: |K| · |K ◦ | = 8.000

Beniamin Bogosel Numerical shape optimization among convex sets 24/25


Conclusions, Perspectives

Preprint: [Bogosel, Numerical shape optimization among convex sets, 2022]


Arxiv: 2203.06981
spectral decomposition of the support function: works well for strictly convex
shapes in 2D and 3D
working directly with a set of values of the support function/gauge function:
rigorous and flexible framework in 2D: −→ extend to 3D
2D codes available online: https://github.com/bbogo/ConvexSets

Thank you for your attention!

Beniamin Bogosel Numerical shape optimization among convex sets 25/25

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy