Você está na página 1de 5

Simple and Effective Scheduling in Wireless

Networks under the Physical Interference Model


Dejun Yang Xi Fang Guoliang Xue Afsheen Irani Satyajayant Misra
AbstractIn this paper, we study the problem of maximizing
the number of concurrent requests and the problem of minimiz-
ing the number of time-slots needed to schedule all requests in
wireless networks under the physical interference model. It has
been proved that both problems are NP-complete in [7]. Thus
either approximation algorithms with guaranteed approximation
factors or effective heuristics with practically good performances
are desirable. We focus on the latter and present simple and
effective heuristic algorithms for these two problems. Extensive
experiments show that our algorithm for the rst problem
outperforms the best approximation algorithm by 62% 72%
on average, and our two algorithms for the second problem give
the best results among existing algorithms.
I. INTRODUCTION
Given a set of communication requests, represented by
links, in wireless networks, two fundamental questions are
usually asked: What is the maximum number of requests
that can be scheduled simultaneously? and What is the
minimum number of time-slots needed to schedule all the
requests?. Both scheduling questions are at the heart of
wireless communications, as they are crucial to determine the
capacity of networks. Extensive work has been done in attempt
to answer these two questions over the past decades.
In order to answer these questions, we have to take into
account two factors: the network topology and the interference
model. Network topology is the structure of the networks
including random networks, where nodes are randomly dis-
tributed, and arbitrary networks, which is a more general net-
work structure. Studies have shown that there is a signicant
gap between the capacities of random networks and worst-case
networks, which is special cases of arbitrary networks [8, 9].
As transmissions are using a common medium, mutual
interference among concurrently transmitting links is another
issue when studying scheduling problems in wireless net-
works. In order to characterize the interference in wireless
communications, two models have been commonly used, the
graph-based interference model and the physical interference
model. Although the scheduling problems under both models
have been proved to be NP-complete, the choice of the
interference model has a signicant impact on the difculty
of developing approximation algorithms with performances
close to optimal. In the graph-based model, e.g. the protocol
interference model [6], a node can successfully receive a
message from another node if and only if no node within
Yang, Fang, Xue are all with the CSE Dept. at Arizona State University,
Tempe, AZ 85287. Irani is with The Goldman Sachs Group, Inc. Misra is with
the CS Dept. at New Mexico State University, Las Cruces, NM 88003. Email:
{dejun.yang, xi.fang, xue, arani}@asu.edu, misra@cs.nmsu.edu. This
research was supported in part by NSF grants 0905603 and 0901451.
a certain range from it is also transmitting simultaneously.
To address the scheduling problems under this model, a
conict graph is always built as a key step while designing
algorithms. Unfortunately, this simplied interference model
abstracts away the accumulative nature of interference. Even if
a single transmitter far away from a receiver may not corrupt
the transmission, the accumulated interference from several
such nodes could still generate enough interference to prevent
the receiver from successfully decoding the received message.
Compared to the graph-based model, the physical interfer-
ence model has been recognized as a more realistic model
for interference in wireless communications. In this model,
a communication is successful if and only if the ratio of
the received signal to the total interference caused by all
other concurrently transmitting nodes plus noise at the receiver
is beyond a hardware-dened threshold. Unlike the graph-
based model, the conict relation between two links in the
physical interference model is not binary anymore. Thus the
methodology for the graph-based model cannot t this model.
In this paper, we study two scheduling problems in arbitrary
networks under the physical interference model. The rst
problem is to schedule a maximum number of links in one
time-slot. For this problem, we present an effective heuristic
algorithm with time complexity O(n
2
), where n is the number
of links. Simulations show that it improves the results returned
by the currently best approximation algorithm by 62%-72% on
average. The second problem is to schedule all the links in a
minimum number of time-slots. For this problem, we develop
two effective heuristic algorithms. One can achieve the same
performance as the previously best heuristic algorithm, but in
less amount of time. The other one improves the results on
the cost of acceptable extra time.
The organization of this paper is as follows. In Section II, we
briey review previous work related to the studied problems.
In Section III, we present our system model and formally
dene the One-Shot Scheduling and the Minimum-Length
Scheduling problems. Then we present effective heuristic
algorithms for the One-Shot Scheduling problem and the
Minimum-Length Scheduling problem in Section IV and Sec-
tion V, respectively. In Section VI, we evaluate the perfor-
mance of our algorithms by comparing them with existing
algorithms. Finally, we conclude this paper in Section VII.
II. RELATED WORK
In this section, we only focus on the previous work related
to the One-Shot Scheduling problem and the Minimum-Length
Scheduling problem under the physical interference model.
From the information theory perspective, Gupta and Kumar [6]
analyzed the overall capacity of wireless networks under the
physical interference model and gave a few of upper bounds.
The One-Shot Scheduling problem was rst studied and
proved to be NP-complete by Goussevskaia et al. [7]. More
specically, they considered a weighted version of this prob-
lem, whose objective is to nd a maximum-weight subset of
links to be scheduled in one time-slot. By assuming there is
no noise, they proposed the rst algorithm with a non-trivial
approximation factor O(p()), where p() is the diversity
of the network. Recently, Goussevskaia et al. [8] made the
rst effort to propose a constant approximation algorithm.
Unfortunately, as observed in [12, 13], their proofs only
work when the noise is 0. Another attempt was made by Xu
and Tang [13], who proposed another constant approximation
algorithm. However, Wan et al. [12] recently pointed out an
incorrect proof in their paper. As the real icebreaker, Wan et al.
developed the rst constant approximation algorithm, which in
addition has a signicantly smaller approximation factor.
The study of the Minimum-Length Scheduling problem
under the physical interference model can be traced back
to [2]. Like the One-Shot Scheduling problem, it has also
been proved to be NP-complete in [7]. Unlike the One-Shot
Scheduling problem, the existence of a constant approximation
algorithm for this problem is still open. In [2], Brar et al.
proposed a greedy scheduling algorithm with approximation
ratio of O(n
1
2
;(c)+c
(log n)
2
;(c)+c
), where c is an arbitrary
constant and (c) is a constant depending on path loss
exponent c. However, this result is only valid when nodes
are uniformly randomly distributed in a square. Goussevskaia
et al. [7] proposed an algorithm using similar techniques and
with same approximation ratio as the algorithm for the One-
Shot Scheduling problem in the same paper. In [8], the authors
applied One-Shot Scheduling algorithm in the same paper
iteratively to achieve a multiple time-slots scheduling.
A series of variants of the link scheduling problems have
also been intensely studied. In [9], Moscribroda and Wat-
tenhofer designed an efcient power assignment algorithm
to schedule a connected set of links in O(log
4
n) time-slots
under the physical interference model. Along the same line,
the same group improved the result to O(log
3
n) in [10] and
further to O(log
2
n) in [11]. In [3], Chafekar et al. developed
algorithms for the problem of joint power control and routing
to minimize the end-to-end latency. In [4], they proposed an
algorithm with approximation guarantee of O(log ) for a
similar problem without power control, where is the ratio
between the maximum and the minimum link lengths. With
power control, Andrews and Dinitz [1] proposed an approx-
imation algorithm with approximation ratio of O(log ) for
the One-Shot Scheduling problem. Fu et al. [5] designed an
approximation algorithm with O(log ) guarantee factor for
the Minimum-Length Scheduling problem with consecutive
transmission constraints.
III. SYSTEM MODEL AND PROBLEM FORMULATION
In our model, we consider a set = {|
1
, |
2
, , |
n
} of n
links (we use links and requests interchangeably). Each link
|
.
represents a transmission request from a sender :
.
to a
receiver :
.
. We assume there is a unit-trafc demand on each
link. For the case of multiple demands, say r, we can simply
replicate each link r times. All nodes are static and located
on a two-dimensional Euclidean plane. Let d(:
.
, :

) denote
the Euclidean distance between :
.
and :

. For simplicity, let


d
.
= d(:
.
, :

). Let 1
.
denote the xed transmission power
of sender :
.
. Then the received power from :
.
at receiver :

is given by 1
.
= 1
.
, max{d
o
.
, 1}, where c is the path loss
exponent ranging from 2 to 4 depending on the communication
medium. Here the max operator is used to account for the fact
that the log-distance path loss model holds only for distance
beyond a certain value.
We assume there is only one available channel. As nodes are
communicating with each other in a common medium, mutual
interference is inevitable among links transmitting simultane-
ously. Although two interference models, i.e. the graph-based
and the physical interference models, have been widely used,
we adopt the physical interference model to characterize the
interference in reality as closely as possible. In this model,
a message sent by a sender :
.
can be correctly received at
a receiver :
.
if and only if the Signal-to-Interference-plus-
Noise-Ratio (SINR) at :
.
is greater than or equal to a certain
threshold, or mathematically, the following condition holds:
1
..

l=l1
1
.
+
0
a, (1)
where is the set of links transmitting at the same time with
link |
.
and a 0 is the minimum SINR required for a message
to be successfully decoded. Let 1
.
denote the interference on
link |
.
caused by link |

, i.e., 1
.
= 1
.
.
In this paper, we study two scheduling problems in wireless
networks under the physical interference model.
One-Shot Scheduling: Given a set of links, One-Shot
Scheduling is to schedule a maximum number of links in one
time-slot such that Condition (1) holds for each link |
.
.
Minimum-Length Scheduling: Given a set of links,
Minimum-Length Scheduling is to partition into a minimum
number of subsets
1
,
2
, . . . ,
J
such that Condition (1)
holds when is replaced by
|
for 1 / 1.
Before giving algorithms for these two problems, we start
with some denitions. We use the same denitions Link
tolerance and Residual link tolerance as in [14], but mainly
focus on directed links in this paper.
Denition 3.1 (Link Tolerance): The tolerance t
.
of a
link |
.
indicates how much interference the link can endure
before the SINR falls below the threshold a. It can be
calculated by
t
.
=
1
..
a

0
. (2)
Denition 3.2 (Residual Link Tolerance): The residual
tolerance t
:
.
of a link |
.
indicates that when |
.
is scheduled
in , how much more interference the link can endure before
the SINR falls below the threshold a. It can be calculated by
t
:
.
= t
.

l
1
.
. (3)
Denition 3.3 (Feasible Link): A link |
.
is feasible with
respect to a set of links if, after the addition of |
.
to the link
set, Condition (1) is satised for all links |

{|
.
}.
IV. THE ONE SHOT PROBLEM
In this section, we design an effective heuristic algorithm
to solve the One-Shot Scheduling problem, whose objective is
to schedule a maximum number of links in one time-slot. It
has been proved that the One-Shot Scheduling problem is NP-
complete in [7]. Thus either an approximation algorithm with
provably good approximation ratio or an effective heuristic
algorithm is desirable. Here we focus on designing an effective
heuristic algorithm which can achieve signicantly better
results in practice than any other approximation algorithms.
We start with an intuitive idea. When we choose a link
for a time-slot, what we are really seeking for is one that is
most resistant to interference and least threatening to others in
that time-slot. This property is best characterized by the ratio
r
r
1
max
I

J1
, where is the set of links in the time-slot. Now
we are ready to present our heuristic algorithm, as listed in
Algorithm 1, for solving the One-Shot Scheduling problem.
Algorithm 1 Maximum Tolerance-to-Interference-Ratio
(MTIR)
Input: A set of links.
Output: A set of links that can transmit simultaneously.
1: ;
2: Compute t
.
for each |
.
and set t
:
.
t
.
;
3: Pick a link |
.
with maximum t
.
;
4: {|
.
}, {|
.
};
5: for all |

do
6: t
:

t
:

1
.
;
7: if t
:

< 0 then
8: {|

};
9: end if
10: end for
11: while = do
12: Pick a feasible link |
.
with maximum
r
r
1
max
I

J1
;
13: {|
.
}, {|
.
};
14: for all |

do t
:

t
:

1
.
; end for
15: for all |

do
16: t
:

t
:

1
.
;
17: if t
:

< 0 then
18: {|

};
19: end if
20: end for
21: end while
22: return .
The basic idea of Algorithm 1 is as follows: We rst
add a link with maximum link tolerance into the time-slot
(Line 3) and update the residual tolerance of other links
(Lines 5-10). We then greedily schedule a feasible link |
.
with
maximum ratio
r
r
1
max
I

J1
(Line 12). After |
.
is added to the
solution , we update the residual tolerance for each link in
and each unscheduled link (Lines 15-20). If an unscheduled
links residual tolerance drops below 0, we remove it from
(Line 18). This process is repeated until every link is either
removed or scheduled.
Next, we prove the correctness and the time complexity of
Algorithm 1.
Theorem 4.1: Given a set of links, Algorithm 1 com-
putes a set of feasible links, which can transmit simultane-
ously in one time-slot. In addition, the time complexity of
Algorithm 1 is O(n
2
), where n is the number of links in .
Proof: The feasibility of each link in can be guaranteed by
Lines 5-10, Line 12 and Lines 15-20. Next we analyze the
time complexity of Algorithm 1. Line 2 takes O(n) time to
initialize the tolerance for each link. It then takes O(n) time to
pick a link in Line 3 and update the residual tolerance for other
links. Obviously, the ni|c-|ooj is executed at most n times.
For each execution of the ni|c-|ooj, it takes O(n) time to
pick the feasible link |
.
and update the residual tolerance for
each link in and each unscheduled link. Therefore, the time
complexity of Algorithm 1 is O(n
2
).
V. THE MINIMUM-LENGTH SCHEDULING PROBLEM
In this section, we design two effective algorithms to solve
the Minimum-Length Scheduling problem, whose objective is
to schedule all the links into a minimum number of time-slots.
Similar to the One-Shot Scheduling problem, the Minimum-
Length Scheduling problem has also been proved to be NP-
complete in [7]. Again, we concentrate on developing heuristic
algorithms for solving this problem.
Our rst algorithm takes Algorithm MTIR as a building
block, which is illustrated in Algorithm 2.
Algorithm 2 Iterated Maximum Tolerance-to-Interference-
Ratio (IMTIR)
Input: A set of links.
Output: A set of subset of links such that all links in each
subset can transmit simultaneously.
1: , t 1;
2: while = do
3:
|
;
4: Apply Algorithm 1 to to get a set of links for t-th
time-slot;
5: ,
|
, {
|
};
6: t t + 1;
7: end while
8: return .
The basic idea of Algorithm 2 is as follows: We iteratively
compute a one-shot schedule using Algorithm 1. Each returned
set of links is scheduled in a separate time-slot. The remaining
links are used as input in the next iteration. This process is
repeated until all links in have been scheduled. We prove the
correctness and the time complexity in the following theorem.
Theorem 5.1: Given a set of links, Algorithm 2 com-
putes a schedule of links, in which links belonging to the
same time-slot can transmit simultaneously. In addition, the
time complexity of Algorithm 2 is O(n
3
), where n is the
number of links in .
Proof: The correctness of Algorithm 2 is guaranteed by the
correctness of Algorithm 1, which is proved by Theorem 4.1.
It is clear that the most time consuming part is the ni|c-
|ooj. As there are at most n time-slots, it can execute at most
n times. Considering the time complexity of Algorithm 1 is
O(n
2
), the time complexity of Algorithm 2 is O(n
3
).
Now, let us analyze the Minimum-Length Scheduling prob-
lem more carefully. Each time when we try to schedule a
link, we should pick one such that the residual tolerance of
the most vulnerable link is maximized after we add it in the
current time-slot. By using this criteria, we potentially increase
the chance of scheduling more links in one time-slot. After
formally dene the most vulnerable link, we give our second
algorithm in Algorithm 3.
Denition 5.1 (Bottleneck): A link |
b
is a bottleneck of a
set of links, if t
:
b
= :in
l1
t
:
.
. Let T() denote the
residual tolerance of a bottleneck of link set .
Algorithm 3 Maximum Bottleneck Tolerance (MBT)
Input: A set of links.
Output: A set of subsets of links such that all links in each
subset can transmit simultaneously.
1: , t 1;
2: Compute t
.
for each |
.
;
3: Sort the links in in ascending order based on their link
tolerances;
4: while = do
5:
|
;
6: Add the rst unscheduled link |
.
to
|
;
7:
}
feasible links with respect to
|
;
8: while
}
= do
9: Pick a link |
.

}
, such that |
.
=
o:p:or
l
!

j
T(
|
{|
|
});
10: If T(
|
{|
.
}) < 0 then break; end if
11:
|

|
{|
.
},
}

}
{|
.
};
12: Update the t
:

for each |


}
;
13: If t
:

< 0 then
}

}
{|

}; end if
14: end while
15:
|
, {
|
};
16: t t + 1;
17: end while
18: return .
The basic idea of Algorithm 3 is as follows: We rst
initialize the variables and compute the tolerance for each
link. Then we sort the links in ascending order of their link
tolerances in Line 3. Every time we start scheduling links
for a new time-slot, we pick the rst unscheduled link from
the sorted list. After we add the rst link in time-slot
|
,
we calculate a set
}
of feasible links with respect to
|
.
Next, we iteratively pick a link from
}
that will give the
maximum bottleneck tolerance if scheduled and add it to
|
.
If the maximum bottleneck tolerance is less than 0, we know
there is no feasible link in
}
and can stop scheduling links
for time-slot
|
. As soon as a link is scheduled, we need to
update the residual tolerance for each unscheduled link in
}
.
If the residual tolerance of a link falls below 0, we remove it
from feasible links set
}
. The above procedure is repeated
until there is no unscheduled link left.
Theorem 5.2: Given a set of links, Algorithm 3 com-
putes a schedule of links, in which links belonging to the
same time-slot can transmit simultaneously. In addition, the
time complexity of Algorithm 3 is O(n
4
), where n is the
number of links in .
Proof: The correctness of Algorithm 3 is guaranteed by
Lines 10 and 13. Obviously, the most time consuming part is
the ni|c-|ooj from Line 4 to Line 17. As there are at most
n time-slots, this ni|c-|ooj can execute at most n times.
Similarly, in each execution, the most time consuming part
is the ni|c-|ooj from Line 8 to Line 14. In each iteration,
picking a link that gives maximum bottleneck tolerance if
scheduled takes O(n
2
) time. Thus the inner ni|c-|ooj takes
at most O(n
3
) time. Finally, we prove that the time complexity
of Algorithm 3 is O(n
4
).
VI. NUMERICAL RESULTS
In this section, we evaluate the performance of our algo-
rithms by comparing them with the current existing algorithms.
A. Simulation Setup
For the One-Shot Scheduling problem, we implemented
Algorithm 1 (denoted by MTIR), and compared it with algo-
rithms in [12] and [7] (denoted by MISL-OS and GSINR-OS,
respectively). For the Minimum-Length Scheduling problem,
we implemented Algorithm 2 and Algorithm 3 (denoted by
IMTIR and MBT), and compared them with algorithms in
[12], [7], [2] and [14] (denoted by MISL-SS, GSINR-SS,
GPhysical and MCG, respectively). Although originally de-
signed for the one-shot scheduling problem, MISL-OS can
be easily extended to an algorithm (i.e. MISL-SS) for the
minimum-length scheduling problem by running MISL-OS
multiple times until all the links have been scheduled.
As in [14], we uniformly distributed links in a 1000 1000
square region. The length of each link is randomly chosen
between 1 and 30. The number of nodes (twice the number of
links) were chosen to be 1000, 2000, . . ., 10000. For each
value of numbers of nodes, we ran simulations ten times,
and averaged the results. The SINR threshold a was set
to 16, the environment noise
0
= 10
9
n, the path loss
exponent c = 4, and transmit power 1
.
= 0.2n. Note that
for this setting the approximation ratio of MISL-OS is 272.
In addition, the links in one slot scheduled by GSINR-OS
and GSINR-SS may be infeasible, since these two algorithms
require the environment noise to be 0, which however is not
a fair assumption [12]. Thus the results shown in Fig.1(a) and
1(b) can be considered as the upper bounds of the feasible
schedule obtained by GSINR-OS and GSINR-SS.
B. Simulation Results
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
100
200
300
400
500
600
number of nodes
n
u
m
b
e
r
o
f lin
k
s


MTIR
MISLOS
GSINROS
(a) Comparison of numbers of links
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
0
2
4
6
8
10
number of nodes
r
u
n
n
in
g
tim
e
(
s
e
c
)


MTIR
MISLOS
GSINROS
(b) Comparison of running times
Fig. 1. Results of one-shot scheduling problem
Fig. 1(a) and Fig. 1(b) show the average numbers of
links scheduled by MISL-OS, GSINR-OS and MTIR, and
the corresponding average running times with the increase in
the density of nodes. We observe that MTIR schedules more
links in one slot than MISL-OS and GSINR-OS by 62%-72%
and 888%-941%, respectively. This is expected, since MISL-
OS is a 272-approximation algorithm and GSINR-OS has an
approximation ratio of O(log p()). As shown in Fig. 1(b),
GSINR-OS runs very fast since it is a linear time algorithm.
Surprisingly, MISL-OS with a theoretical time complexity of
O(n
2
) also runs very fast. This is because it could abandon
a large number of links after each iteration, which may
signicantly reduce the number of links to be checked in the
next iterations. Although MTIR takes more time than GSINR-
OS and MISL-OS, for a network with even 10,000 nodes,
it can nish scheduling within 9 seconds. Considering the
signicant improvement on the number of links in one time-
slot, the sacriced running time is worthwhile.
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
20
40
60
80
100
120
number of nodes
n
u
m
b
e
r
o
f tim
e
s
lo
ts


MBT
IMTIR
GPhysical
MCG
MISLSS
(a) Comparison of numbers of time slots
1000 2000 3000 4000 5000 6000 7000 8000 9000 10000
10
20
30
40
50
60
number of nodes
r
u
n
n
in
g
tim
e
(
s
e
c
)


MBT
IMTIR
GPhysical
MCG
MISLSS
(b) Comparison of running times
Fig. 2. Results of link scheduling problem
Fig. 2(a) and Fig. 2(b) show the average schedule lengths
(i.e. number of time-slots) computed by IMTIR, MBT, MISL-
SS, GPhysical and MCG, and the corresponding average
running times. Since the average schedule lengths computed
by GSINR-SS are from 132 to 950, which is much longer
than the other algorithms, we do not plot its results for clarity
and concentrate on the other ve algorithms. We observe that
MISL-SS leads to the longest schedule length, which is almost
twice as long as the corresponding schedule lengths computed
by the heuristic algorithms (i.e. IMTIR, MBT, GPhysical
and MCG ). This is because its building block MISL-OS
algorithm is a 272-approximation algorithm and does not work
very well in practice. Compared to GPhysical, IMTIR and
MBT use up to 27% and 37% fewer time-slots, respectively.
This is expected, since more accurate information is used
by IMTIR and MBT (i.e. residual tolerance) for selecting
the next link. After each iteration, residual tolerances will be
updated, whereas the metric (i.e the interference number) used
by GPhysical becomes more and more stale as more and more
links are scheduled. IMTIR and MCG have the comparable
schedule lengths, while the actual running times of IMTIR are
less than MCG by 30%-40%, although IMTIR has a higher
time complexity. We now compare IMTIR and MBT. The
schedule lengths computed by IMTIR are longer than those
computed by MBT by 2%-31% This is expected, since MBT
aims to increase the residual tolerance of the bottleneck of a
link set. Intuitively, this link selection principle can lead to
more links scheduled in one slot. However, this principle also
requires a longer running time.
VII. CONCLUSION
In this paper, we studied the One-Shot Scheduling problem,
of which the objective is to maximize the number of requests
that can transmit simultaneously, and the Minimum-Length
Scheduling problem, of which the objective is to minimize the
number of time-slots to schedule all the requests. For the One-
Shot Scheduling problem, we designed an effective heuristic
algorithm, which improves the currently best approximation
algorithm by 62%72% on average. For the Minimum-Length
Scheduling problem, we developed two heuristic algorithms.
One of them produces as good results as the previously best
heuristic algorithm MCG does, but reduces the running time.
The other one improves the results returned by the MCG.
REFERENCES
[1] M. Andrews and M. Dinitz; Maximizing capacity in arbitrary wireless
networks in the SINR model: complexity and game theory; Infocom09,
pp. 13321340, 2009.
[2] G. Brar , D. M. Blough and P. Santi; Computationally efcient schedul-
ing with the physical interference model for throughput improvement
in wireless mesh networks;ACM Mobicom06, pp. 213, 2006.
[3] D. Chafekar, V.S.A. Kumar, M.V. Marathe, S. Parthasarathy and A.
Srinivasan; Cross-layer latency minimization in wireless networks with
SINR constraints; Mobihoc07, pp. 110119, 2007.
[4] D. Chafekar, V.S.A. Kumar, M.V. Marathe, S. Parthasarathy and
A. Srinivasan; Approximation algorithms for computing capacity of
wireless networks with SINR constraints; Infocom08, pp. 1166-1174,
2008.
[5] L. Fu, S.C. Liew and J. Huang; Power controlled scheduling with con-
secutive transmission constraints: complexity analysis and algorithm
design; Infocom09, pp. 15301538, 2009.
[6] P. Gupta and P.R. Kumar; The Capacity of Wireless Networks; IEEE
Trans. Info. Theory, Vol. 46, No. 2, pp.388-404, 2000.
[7] O. Goussevskaia, Y.A. Oswald and R. Wattenhofer; Complexity in
geometric SINR; MobiHoc 07, pp. 100109, 2007.
[8] O. Goussevskaia, R. Wattenhofer, M.M. Halldorsson and E. Welzl;
Capacity of arbitrary wireless networks; Infocom09, pp. 18721880,
2009.
[9] T. Moscribroda and R. Wattenhofer; The complexity of connectivity in
wireless networks; Infocom06, pp. 113, 2006.
[10] T. Moscribroda, R. Wattenhofer and A. Zollinger; Topology control
meets SINR: the scheduling complexity of arbitrary topologies; Mobi-
hoc06, pp. 310-321, 2006.
[11] T. Moscribroda, Y.A. Oswald and R. Wattenhofer; How optimal are
wireless scheduling protocols?; Infocom07, pp. 14331441, 2007.
[12] P.-J. Wan, X. Jia and F. Yao; Maximum independent set of links under
physical interference model; WASA09, pp. 169178, 2009.
[13] X.-H. Xu and S.-J. Tang; A constant approximation algorithm for link
scheduling in arbitrary networks under physical interference model;
FOWANC09, pp. 1320, 2009.
[14] D. Yang, X. Fang, N. Li and G. Xue; A simple greedy algorithm for
link scheduling with the physical interference model; Globecom09,
pp. 16, 2009.

Você também pode gostar