RES/SEC 1
Introduction to the
Gathering Problem
Gathering -‐ Unlimited Visibility
Initially the robots are in arbitrary distinct positions.
RES/SEC 3
Gathering
Unlimited Visibility -‐ SYm
Istantaneous activities
n=2, the problem is unsolvable
Ando, Oasa, Suzuki, Yamashita
Siam Journal Of Computing, 1999
Gathering, n=2
Unlimited Visibility -‐ SYm
In fact, since the robots have no dimension…
…moving them towards each other is not useful…
…but it works if they can bump!
…and cannot bump...
RES/SEC 5
Gathering
Unlimited Visibility -‐ SYm
Istantaneous activities
n=2, the problem is unsolvable
n>2, they provide an oblivious algorithm that let the robots gather in finite time
Ando, Oasa, Suzuki, Yamashita
Siam Journal Of Computing, 1999
Gathering, ASYNC
• In spite of its apparent simplicity, this problem has been tackled in several studies
• In fact, several factors render this problem difficult to solve
– Major problems arise from symmetric configuraHons....
DifficulHes
RES/SEC 7
If at the beginning....
DifficulHes
Not symmetric....
General Algorithm!!
If at the beginning....
DifficulHes
RES/SEC 9
....after a while....
Impossible d'afficher l'image. Votre ordinateur manque peut-être de mémoire pour ouvrir l'image ou l'image est endommagée.
Redémarrez l'ordinateur, Impossible d'afficher l'image. Votre ordinateur manque peut- être de mémoire pour
DifficulHes
Symmetric…
go to c General Algorithm…
Gathering—easy soluHon
RES/SEC 11
Easy Solution: Weber Point (Weiszfeld, ‘36)!
ℜ
∑
∈
=
i p i
r p
WP arg min dist( , )
2
Given r1,…,rn:
1. It is unique (Weiszfeld, ‘36)
WP
Gathering—easy soluHon
2. WP is Weber Point of points on [ri,WP] (Weiszfeld, ‘36)
r4
r2 r1
r5
r
1. It is unique (Weiszfeld, ‘36)
Easy Solution: Weber Point (Weiszfeld, ‘36)!
ℜ
∑
∈
=
i p i
r p
WP arg min dist( , )
2
Given r1,…,rn:
WP
Gathering—easy soluHon
RES/SEC 13
r4
r2 r1
r5
r3
2. WP is Weber Point of points on [ri,WP] (Weiszfeld, ‘36)
1. It is unique (Weiszfeld, ‘36)
Easy Solution: Weber Point (Weiszfeld, ‘36)!
ℜ
∑
∈
=
i p i
r p
WP arg min dist( , )
2
Given r1,…,rn:
WP
Gathering—easy soluHon
r4
r2 r1
r5
r
2. WP is Weber Point of points on [ri,WP] (Weiszfeld, ‘36)
1. It is unique (Weiszfeld, ‘36)
Easy Solution: Weber Point (Weiszfeld, ‘36)!
ℜ
∑
∈
=
i p i
r p
WP arg min dist( , )
2
Given r1,…,rn:
WP
Gathering—easy soluHon
RES/SEC 15
2. WP is Weber Point of points on [ri,WP] (Weiszfeld, ‘36)
1. It is unique (Weiszfeld, ‘36)
Easy Solution: Weber Point (Weiszfeld, ‘36)!
ℜ
∑
∈
=
i p i
r p
WP arg min dist( , )
2
Given r1,…,rn:
WP
Gathering—easy soluHon
Easy Solution: Weber Point (Weiszfeld, ‘36)!
Algorithm:
1. Compute WP
2. Move Towards WP
Gathering
(Unlimited Visibility, no agreement)
RES/SEC 19
Gathering
(Unlimited Visibility, no agreement)
Gathering, n=3
(Unlimited Visibility, no agreement)
RES/SEC 21
r2
r1 r3
p p
r2
r3 r1
p
≥120° p
General Schema
• Use of MulHplicity DetecHon
– n=3,4 (and even with the use of Weber Point)
Is there 1 or more than 1 robot at a point?
>1
>1
>1
General Schema
RES/SEC 23
1. At the beginning, robots on distinct positions 2. Get a scenario where there is only one point p with multiplicity greater than one
3. All robots move towards p
MulHplicity DetecHon
p
P’
p
P’
MulHplicity DetecHon
RES/SEC 25
For n=2, the problem is not solvable (Suzuki et al., 1999)!
p
P’
It is possible to design an adversary that lets the robots occupy two distinct positions on the plane in a finite number of cycles…
MulHplicity DetecHon
Problem not solvable with n=2
No multiplicity detection
Let’s design the adversary
• We divide the robots in two sets
– n-‐1 robots (r1, ..., rn-‐1) are the blue robots
– There is one red robot (rn) (same direcHon but opposite orientaHon)
RES/SEC 27
E-‐configuraHons
E1-configuration
E-‐configuraHons
RES/SEC 29
E-‐configuraHons
E-‐configuraHons
RES/SEC 31
Time: 1
E-‐configuraHons
E-‐configuraHons
RES/SEC 33
Time: ....
E-‐configuraHons
E-‐configuraHons
RES/SEC 35
Time: t
E-‐configuraHons
In the example: E2-conf
E1-‐configuraHon
RES/SEC 37
E1-‐configuraHon
E1-‐configuraHon
RES/SEC 39
Therefore, for a while the configuration stays E1
E1-‐configuraHon
E1-‐configuraHon
RES/SEC 41
E1-‐configuraHon
E1-‐configuraHon
RES/SEC 43
E1-‐configuraHon
E1-‐configuraHon
RES/SEC 45
E1-‐configuraHon
Again in a E1
E2-‐configuraHon
RES/SEC 47
Finally....
Gathering
RES/SEC 49