Upload
nicholas-newman
View
214
Download
0
Tags:
Embed Size (px)
Citation preview
1
Energy-aware stage illumination.
Written by:
Friedrich Eisenbrand
Stefan Funke
Andreas Karrenbauer
Domagoj Matijevic
Presented By: Yossi Maimon.
2
Illumination Problem
Input:
Given a stage and a set of light sources.
Target:
To illuminate the stage such that each point
on the stage receives a sufficient amount of
light (one unit) while minimizing the overall
Power assignment.
3
Illuminate Vs Guarding
• Illuminate is a different version of guarding
problem.
• Energy from a light source is decreasing
quadratically with the distance .
• Add up: A point can accumulate energy from
different light sources.
6
Article Contribution
• A polynomial-time solution based on
convex programming.
• A approximate solution based on a
discretization and linear programming.
• A purely combinatorial O(1) approximate
solution with running time 2( )O n
(1 )
7
Convex programming
2
min
. .
: 1( , )
0
ss S
s
s
Objective
x
s txp L
d p sx
S – light source.
X - The energy of light source.
P – Point on the stage.
8
Convex programming (cont)
( , )S K ( , )S K All point in distance from K
All point in distance from K contained in the convex
K-Convex body
The constraints are not bounded so we will look for lighting LP and combinatorial
Target:To determine if a point is in K.
9
Pruning light sources•Under the assumption that each light source can be assigned arbitrary high power.
•Only light sources whose Voronoi cells are intersect the stage can be part of
optimal solution.
Let s be a light source whose Voronoi cell does not intersect the stage,
be the first neighbors to the left and right whose
Voronoi cells intersect the stage.
,l rs s
10
A approximation scheme(1 )
Guard: a point on the stage that receive a sufficient amount of energy
Goal:
•Discrete the problem by using a finite number of guard.
•Solve the linear programming only for the guards
•power up all light sources
•In the end: each point on the stage that isn’t a guard will
Receive a enough light.
, ( ) min ( , )s Sp L ens p d p s Definition:
11
approximation (cont)(1 )
good
, : ( , ) * ( )G Lp L g G d p g ens p
Set of guards
Construct:
Assume |S|=1.
Let p0 be the closest point to s.
Add p0 to G.
Build p-1 and p+1 in
02 ( )ens p
1( , ) 2 ( )i i id p p ens p
12
approximation (cont 1)(1 )
log| | ( )
DGs O
D denote the length of L.
The constraints is depend on the length of the stage.
Numbers of guards:
Several light sources:
•For each light source |Gs| will be computed.
•Union all the sets.
• , : ( , ) ( )p L g G d p g ens p
| | ( log[ ])n D
Gs On
13
approximation (cont 2)(1 )
Powering:
Powering every light source in
Ensures that every point receive enough light.
(1 6 )
Summery:
(1 ) opts s
s S s Sx x
The light source energy can be found by solving LP with
constraint and n variables.( log )n DOn
14
A simple O(1) approximationAlgorithm
Restricting the problem to O(n) guards.
Transfer back to the original problem in O(1) in terms of quality.Lemma:
4*Xv is power assignment to all point on the stage.
A 4 approximation can be solve by LP with n+1 constraint.
Independent to the length of the stage.
15
1. Compute for each guard p the ens(p).
2. Sort the guards in decreasing order.
3. For i=1…n
if has not been remove yet, remove all guards at distance
4. Return the guards as Gp.
simple O(1)-Pruning guards
ip
| * | *max( ( )* ( )), 0i j i jg g C ens g ens g C
jp * ( )iens p
,
| | ( )
i i j
i j j
p Gv p Gp p Gp
p p ens p
2 2(1 ) * 4(1 )Xv Xp
16
simple O(1)
1. Compute the set of guards Gv (via the Voronoi diagram of S).
2. Prune the set of guards Gv with pruning constant to obtain Gp, |Gp|=m.
3. Let Gp be ordered such that
4. For all i=1..m
1 2( ) ( ) .... ( )mens p ens p ens p
1221
max{0,| * | *(1 )}| * |
i ji i i j
i j
xx p s
p s
Running time: 2( )O n
No guard gets more then a constant amount of energy.
17
simple O(1) – (cont)Definitions:
will be the amount of light in ipiP
4 1 6 11i i i iP P P P
i
i
i
P
P
P
Energy from light sources where j<i
Energy from light source
Energy from light sources where j>i is
is
is
18
Open problemsArt gallery illumination with fix number of light source.
Given polygon with n vertices and k fix number of light sources,
determine the position and power to each light source
such that each point (on edge or interior) has at least 1 unit of energy.
Another variant is to restrict the position of the light source
only on vertices or edges.
19
Open problems (cont)Stage illuminations with obstacles:
The same problem only this time with obstacle.
The pruning light sources nor the disretization can be applied immediately.
20
ResultsPerformance according to the Analysis.
D-The length of the stage.
|Gv|-Number of light sources.
21
Actual resultAdaptive power up.
Using a refined power up strategy the algorithms achieve result closest to
The optimal.