PIMRC’08 1 - 16/09/2008
Hypergraph T-Coloring for
Automatic Frequency Planning problem in Wireless LAN
Alexandre Gondran - Oumaya Baala Hakim Mabed - Alexandre Caminada Cannes – 16 September 2008
19th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications
PIMRC’08 2 - 16/00/2008
Agenda
1. Channel Assignment in Wireless LAN based on SINR
2. Necessary condition:
Graph T-coloring Problem 3. Quasi equivalent condition:
Hypergraph T-coloring Problem 4. Results
5. Conclusions/Perspectives
PIMRC’08 3 - 16/00/2008
Channel Assignment in WLAN
Limit the interferences
which degrade Quality of Service network by limiting its capacity.
Not possible to avoid interferences
to spread as well as possible interferences over the whole area
SINR computation : long time
Signal-to-Interference-plus-Noise-Ratio
SINR constraints
PIMRC’08 4 - 16/00/2008
Channel Assignment in WLAN:
SINR constraints
SINR ≥ 20 dB SINR ≥ 25 dB
SINR ≥ 15 dB
Determine f
iSatisfied : ∀ k, SINR
k≥≥≥≥ s
kuser
Access Point (AP) f1
f2 f3
f4
fi frequency channel used by Access Point i
sk SINR threshold necessary to satisfied user k
PIMRC’08 5 - 16/00/2008
(
k k)
k
i k k
ik i i
i i
SINR p
p γ f f N
≠
= ∑ × − +
IEEE 802.11g
Determine f
iSatisfied : ∀ k, SINR
k(f
i) ≥≥≥≥ s
kChannel Assignment in WLAN:
SINR constraints
f
iPIMRC’08 6 - 16/00/2008
SINR constraints
(
kk)
k
i k
k k
ik i i
i i
SINR p s
p γ f f N
≠
= ≥
× − +
∑
| f
j− ≥ f
i| t
ijGraph T-coloring
1 1
max , ; ,
k k
i k i k
k k
ij k k
k ik jk
p p
N N
s s
t j i i j
p p
γ − γ −
− −
= = =
PIMRC’08 7 - 16/00/2008
- 60 dBm - 73 dBm
- 72 dBm f1 f3
f2
SINR ≥ 15dB
Necessary condition:
Graph T-coloring problem
Example 1
PIMRC’08 8 - 16/00/2008
2 2
Necessary condition:
Graph T-coloring problem
1 2
1 3
| | 2
15 | | 2
f f
SINR f f
− ≥
≥ ⇔
− ≥
f1 f3
f2
PIMRC’08 9 - 16/00/2008
SINR constraints
(
kk)
k
i k
k k
ik i i
i i
SINR p s
p γ f f N
≠
= ≥
× − +
∑
| f
j− ≥ f
i| t
ijGraph T-coloring
Theorem:
Yes, if
, : k
( )
k k
i k
k k
ik ii
i i
k SINR p s
p γ t N
≠
∀ = ≥
× +
∑
equivalence ?
PIMRC’08 10 - 16/00/2008
f1 f3
f2
- 60 dBm - 73 dBm
- 72 dBm - 63 dBm
Quasi equivalent condition:
Hypergraph T-coloring problem
Example 2
PIMRC’08 11 - 16/00/2008
2 3
1 2
1 3
| | 2
15 | | 3
f f
SINR f f
− ≥
≥ ⇒
− ≥
f1 f3
f2
Quasi equivalent condition:
Hypergraph T-coloring problem
PIMRC’08 12 - 16/00/2008
1 2
1 3
| | 2
15 | 3
f f
SINR f f
− ≥
≥ ⇔
− ≥
no equivalence
It’s necessary to add a new constraint linear n-ary constraints :
1 2
1 3
1 2 1 3
| | 2
15 | | 3
| | | | 6
f f
SINR f f
f f f f
− ≥
≥ ⇔ − ≥
− + − ≥
Quasi equivalent condition:
Hypergraph T-coloring problem
1 2 1 3
| f − f | + | f − f | ≥ 6
PIMRC’08 13 - 16/00/2008
SINR constraints
(
kk)
k
i k
k k
ik i i
i i
SINR p s
p γ f f N
≠
= ≥
× − +
∑
, , 1 min ( ) ( ) k
k k
i k
k ik jk ji ik ii
t i j k
k i i p t t p t p N
α γ γ s
≠
∀ ∀ ≠ = + + ≤ −
∑
| |
k k
k
ik i i ii
i i
f f
α α
≠
− ≥
∑
Hypergraph T-coloring
| f
j− ≥ f
i| t
ijGraph T-coloring
PIMRC’08 14 - 16/00/2008
SINR constraints
(
kk)
k
i k
k k
ik i i
i i
SINR p s
p γ f f N
≠
= ≥
× − +
∑
,
k,
ik1
k i i α
∀ ∀ ≠ =
| |
k k
k
ik i i ii
i i
f f
α α
≠
− ≥
∑
Hypergraph T-coloring
| f
j− ≥ f
i| t
ijGraph T-coloring
equivalence ?
Theorem:
Yes, if
PIMRC’08 15 - 16/00/2008
Results
HypergraphT-coloring Graph
T-coloring SINR
constraints
9 AP
15 AP
30 AP
40 AP
time (s)
time (s) time (s)
PIMRC’08 16 - 16/00/2008
Conclusions / Perspectives
• Automatic method to build a good T matrix for graph T-coloring problem.
• Define a new problem:
hypergraph T-coloring problem.
• Results equivalents for a better computation time.
• Benchmarks soon in Internet:
http://alexandre.gondran.free.fr
• Integrate this approach into a global WLAN planning process.
PIMRC’08 17 - 16/00/2008
SINR constraints
(
kk)
k
i k
k k
ik i i
i i
SINR p s
p γ f f N
≠
= ≥
× − +
∑
| |
k k
k
ik i i ii
i i
f f
α α
≠
− ≥
∑
Hypergraph T-coloring
| f
j− ≥ f
i| t
ijGraph T-coloring
In real problem, threshold sk are unknown Only throughput demand per user
(kilobit/s) are known
Remark
PIMRC’08 18 - 16/00/2008
SINR constraints
(
kk)
k
i k
k k
ik i i
i i
SINR p s
p γ f f N
≠
= ≥
× − +
∑
| |
k k
k
ik i i ii
i i
f f
α α
≠
− ≥
∑
Hypergraph T-coloring
| f
j− ≥ f
i| t
ijGraph T-coloring Capacity
constraints
CapacityAP ≥ Demand
In real problem, threshold sk are unknown Only throughput demand per user
(kilobit/s) are known
PIMRC’08 19 - 16/00/2008
SINR constraints
(
kk)
k
i k
k k
ik i i
i i
SINR p s
p γ f f N
≠
= ≥
× − +
∑
| |
k k
k
ik i i ii
i i
f f
α α
≠
− ≥
∑
Hypergraph T-coloring
| f
j− ≥ f
i| t
ijGraph T-coloring
We define a new algorithm which determine dynamically the best threshold sk to transform the
problem into graph and
hypergraph T-coloring problem Capacity
constraints
CapacityAP ≥ Demand
PIMRC’08 20 - 16/09/2008
Hypergraph T-Coloring for
Automatic Frequency Planning problem in Wireless LAN
Alexandre Gondran - Oumaya Baala Hakim Mabed - Alexandre Caminada Cannes – 16 September 2008
19th Annual IEEE International Symposium on Personal, Indoor and Mobile Radio Communications