Professional Documents
Culture Documents
protocols in a bidirectional relay network consisting of two Rayleigh fading channel.1 Also, we can accommodate path-
different end-sources and multiple relays. loss effect by choosing appropriate values for Ω0 , Ω1,𝑙 , and
The contributions of this paper can be summarized as Ω2,𝑙 . Specifically, let 𝑑𝑆1 ,𝑅 denote the distance between S1
follows. and R𝑙 , and 𝑑𝑆2 ,𝑅 , the distance between S2 and R𝑙 , both of
∙ We propose RS schemes for the ANC and TDBC proto- which are normalized by the distance between S1 and S2 .
cols to minimize the outage probabilities. Specifically, we Therefore, we have 𝑑𝑆1 ,𝑅 + 𝑑𝑆2 ,𝑅 = 1. Furthermore, we set
adopt the max-min criterion to maximize the minimum the path-loss exponent as four to model radio propagation in
mutual information of two opposite traffic flows. urban areas [26]. As a result, we set Ω0 = 1, Ω1,𝑙 = 𝑑−4 𝑆1 ,𝑅 ,
−4
∙ For the RS in the ANC protocol, we derive closed-form and Ω2,𝑙 = (1 − 𝑑𝑆1 ,𝑅 ) where 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿.
outage probability. In the ANC protocol, there are two opposite traffic flows:
∙ For the RS in the TDBC protocol, we derive outage one is from S1 via R𝑙 to S2 , and the other is S2 via R𝑙 to
probability in a one-integral form and its lower bound S1 . For the first traffic flow, R𝑙 and S2 are the receivers.
in closed-form. Thus, it is reasonable to assume that R𝑙 knows ℎ1,𝑙 , and
In particular, our work is different from [22]. In [22], a S2 knows {ℎ1,𝑙 , ℎ2,𝑙 : 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿} as in the traditional
single best relay was selected such that the sum of two mutual unidirectional relay network [27]–[29].2 Similarly, considering
information of two opposite traffic flows was maximized, the second traffic flow, it is reasonable to assume that R𝑙
whereas in our work, a single best relay is selected such knows ℎ2,𝑙 , and S1 knows {ℎ1,𝑙 , ℎ2,𝑙 : 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿}. In
that the minimum mutual information of two opposite traffic total, we assume that both end-sources know all the channel
flows is maximized. Note that the sum-rate maximization coefficients {ℎ1,𝑙 , ℎ2,𝑙 : 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿}, and that each relay R𝑙
criterion is useful when one focuses on the total traffic flows of knows the channel coefficients ℎ1,𝑙 and ℎ2,𝑙 . In a similar way,
bidirectional communication, whereas the max-min criterion for the TDBC protocol, we assume that both end-sources know
is useful when one focuses on the individual traffic flow of all the channel coefficients {ℎ0 , ℎ1,𝑙 , ℎ2,𝑙 : 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿}, and
bidirectional communication. As a result, the two criterions that each relay R𝑙 knows the channel coefficients ℎ1,𝑙 and
complement each other, not exclude each other. ℎ2,𝑙 . Compared with the ANC, the TDBC needs additional
The remainder of this paper is organized as follows. In channel information on ℎ0 at both end-sources because the
Section II, we describe the system model. In Section III, we TDBC utilizes the direct-path between the two end-sources,
consider RS schemes based on the max-min criterion for the while the ANC cannot utilize the direct-path due to the half-
ANC and TDBC protocols. In Section IV, we derive the outage duplex constraint even if such direct-path physically exists.
probability of the RS in the ANC protocol. In Section V, The channel information assumption adopted in this paper
we derive the outage probability of the RS in the TDBC has been made in almost all previous works3 on AF-based
protocol. In Section VI, we present simulation results. Finally, bidirectional protocols [4]–[9], [12].4
conclusions are drawn in Section VII.
Notation: We use 𝐴 := 𝐵 to denote that 𝐴, by definition, A. ANC Protocol with Multiple Relays
equals 𝐵, and we use 𝐴 =: 𝐵 to denote that 𝐵, by definition, The ANC protocol has widely been studied [4]–[9], [22].
equals 𝐴. Also, for a random variable 𝑋, 𝑓𝑋 (⋅) denotes its The system model for the ANC protocol with multiple relays
probability distribution function (PDF), and 𝐹𝑋 (⋅), its cumu- is depicted in Fig. 1(a). In this protocol, at the first time slot,
lative distribution function (CDF). Finally, 𝑥 ∼ 𝒞𝒩 (𝑚, Ω) S1 and S2 transmit simultaneously their symbols, 𝑥1 and 𝑥2 ,
indicates that 𝑥 is a circularly symmetric complex-valued over the multiple access channel, and the signal 𝑟𝑙ANC received
Gaussian random variable with mean 𝑚 and variance Ω. by the 𝑙-th relay R𝑙 is given by
√ √
𝑟𝑙ANC = ℰ𝑠 ℎ1,𝑙 𝑥1 + ℰ𝑠 ℎ2,𝑙 𝑥2 + 𝑛ANC 𝑙 , (1)
II. S YSTEM M ODEL
Consider a bidirectional relay network consisting of two 1 Also, it is an interesting topic to investigate a system for frequency-
different end-sources and 𝐿 relays, where each terminal has selective Rayleigh/Rician/Nakagami fading channels with timing errors [25].
2 Note that R is the receiver associated only with the path from S to R ;
𝑙 1 𝑙
a single antenna and operates in a half-duplex mode, and thus, R𝑙 knows ℎ1,𝑙 . On the other hand, S2 is the receiver associated with
the relays adopt the AF protocol. We use S1 , S2 , and R𝑙 all the pathes from S1 via {R𝑙 }𝐿 𝑙=1 to S2 ; thus, S2 knows {ℎ1,𝑙 , ℎ2,𝑙 : 𝑙 =
to denote the first source, the second source, and the 𝑙-th 1, ⋅ ⋅ ⋅ , 𝐿}.
3 In some works on AF-based bidirectional protocols, even a stronger
relay for 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿, respectively. We let 𝑥1 and 𝑥2 denote
channel assumption has been made. In [30], [31], it is assumed that each
the information-bearing symbols transmitted from S1 and S2 , relay R𝑙 knows all the channel coefficients even for other relays. That is,
respectively, and both symbols have unit average power. The each R𝑙 knows {ℎ1,𝑙 , ℎ2,𝑙 : 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿}. In this case, the relays may use
complex channel coefficient between S1 and S2 is denoted by beamforming techniques, but under the assumption of ideal frequency/time
synchronization across all the relays, which does not need to be assumed in
ℎ0 ; the complex channel coefficient between S1 and R𝑙 is de- the RS.
noted by ℎ1,𝑙 ; and the complex channel coefficient between S2 4 Due to the higher signaling overhead, the perfect instantaneous channel
and R𝑙 is denoted by ℎ2,𝑙 . We assume that all the channels are knowledge of all the links might be hard for a system with a large number
of relays. In practice, however, the number of relays should be restricted
reciprocal and they are modeled as follows: ℎ0 ∼ 𝒞𝒩 (0, Ω0 ), to a small number due to some reasons: the nature of indoor home/office
ℎ1,𝑙 ∼ 𝒞𝒩 (0, Ω1,𝑙 ), and ℎ2,𝑙 ∼ 𝒞𝒩 (0, Ω2,𝑙 ). Also, all the applications may need only a small number of relays; timing, frequency, and
channel coefficients are assumed to be fixed over two time phase synchronizations become very difficult for a large number of relays; and
the complexity of protocol becomes very high for a large number of relays.
slots for the ANC protocol and three time slots for the TDBC That is, assuming a small number of relays is practical and reasonable. As a
protocol. That is, our channel model is the frequency-flat result, the burden on total signaling overhead may not be too much.
3502 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 58, NO. 12, DECEMBER 2010
…
√
S1 Rl S2 for S1 and ℰ𝑠 ℰ𝑟 𝜌𝑙 ℎ22,𝑙 𝑥2 for S2 , from 𝑦𝑖,𝑙
ANC
in (2). Therefore,
ANC
S1 and S2 can obtain new interference-free signals 𝑦˜1,𝑙 and
…
ANC
𝑦˜2,𝑙 , respectively, as follows:
RL √ √
ANC
𝑦˜1,𝑙 = ℰ𝑠 ℰ𝑟 𝜌𝑙 ℎ1,𝑙 ℎ2,𝑙 𝑥2 + ℰ𝑟 𝜌𝑙 ℎ1,𝑙 𝑛ANC
𝑙
ANC
+ 𝑣1,𝑙 ,
(a) ANC protocol with multiple relays (5)
√ √
ANC ANC ANC
𝑦˜2,𝑙 = ℰ𝑠 ℰ𝑟 𝜌𝑙 ℎ1,𝑙 ℎ2,𝑙 𝑥1 + ℰ𝑟 𝜌𝑙 ℎ2,𝑙 𝑛𝑙 + 𝑣2,𝑙 .
R1
… (6)
Consequently, using the approximate 𝜌𝑙 in (4), the instanta-
Rl ANC
neous SNR 𝛾1,𝑙 ANC
of the signal 𝑦˜2,𝑙 in (6) is given by
…
0
10
PNC own symbol 𝑥𝑖 and all the channel coefficients, it can remove
ANC
the self-interference. Consequently, S1 and S2 can obtain new
TDBC TDBC
−1
interference-free signals 𝑦˜1,𝑙 and 𝑦˜2,𝑙 , respectively, as
10
follows:
√ √
TDBC
𝑦˜1,𝑙 = ℰ𝑠 ℰ𝑟 𝜁2,𝑙 ℎ1,𝑙 ℎ2,𝑙 𝑥2 + ℰ𝑟 𝜁1,𝑙 ℎ1,𝑙 𝑛TDBC
1,𝑙
−2 √
+ ℰ𝑟 𝜁2,𝑙 ℎ1,𝑙 𝑛TDBC TDBC
10
+ 𝑣 , (15)
Average BER
2,𝑙 1,𝑙
√ √
TDBC
𝑦˜2,𝑙 = ℰ𝑠 ℰ𝑟 𝜁1,𝑙 ℎ1,𝑙 ℎ2,𝑙 𝑥1 + ℰ𝑟 𝜁1,𝑙 ℎ2,𝑙 𝑛TDBC
1,𝑙
−3 √
TDBC TDBC
10
+ ℰ𝑟 𝜁2,𝑙 ℎ2,𝑙 𝑛2,𝑙 + 𝑣2,𝑙 . (16)
Then, using the approximate 𝜁𝑖,𝑙 in (13), the instantaneous
TDBC TDBC TDBC
−4
10 SNRs 𝛾1,𝑙 of the signal 𝑦˜2,𝑙 in (16) and 𝛾2,𝑙 of the
TDBC
signal 𝑦˜1,𝑙 in (15) are given by, respectively,
in bps/Hz for the whole network. Then the outage probability way, for the traffic flow from S2 to S1 , the mutual information
ANC TDBC
𝑃out,𝑙 (𝑅) of the 𝑙-th relay-path of the ANC protocol is given ℐ2,𝑙 is given by
by
[ ] TDBC 1
𝑅 𝑅 ℐ2,𝑙 = log2 (1 + 𝛾0TDBC + 𝛾2,𝑙
TDBC
). (25)
ANC
𝑃out,𝑙 (𝑅) = Pr ℐ1,𝑙ANC
< ANC
or ℐ2,𝑙 < 3
2 2 TDBC
[ ] Similar to (21), the outage probability 𝑃out,𝑙 (𝑅) of the
ANC ANC 𝑅 𝑙-th relay-path with the direct-path of the TDBC protocol is
= Pr min[ℐ1,𝑙 , ℐ2,𝑙 ]< . (21)
2 given by
Based on this outage probability for each relay-path, we [ ]
TDBC TDBC TDBC 𝑅
now consider RS in the ANC protocol with multiple relays. 𝑃out,𝑙 (𝑅) = Pr min[ℐ1,𝑙 , ℐ2,𝑙 ]< . (26)
2
The objective of RS is to choose a single relay which mini-
mizes the outage probability of the system. To this end, a relay Taking a step similar to (22), we select a single relay to
TDBC TDBC
must be selected such that min[ℐ1,𝑙 ANC ANC
, ℐ2,𝑙 ] is maximized maximize the value min[ℐ1,𝑙 , ℐ2,𝑙 ] over all the relays.
ˆ
over all the relays. That is, the index 𝑙 of the selected relay That is, the index ˜𝑙 of the selected relay must be determined
must be determined as follows:6 as follows:
ˆ𝑙 = arg max min[ℐ ANC , ℐ ANC ]. (22) ˜𝑙 = arg max min[ℐ TDBC , ℐ TDBC ]. (27)
1,𝑙 2,𝑙 1,𝑙 2,𝑙
𝑙=1,⋅⋅⋅ ,𝐿 𝑙=1,⋅⋅⋅ ,𝐿
The RS in the ANC protocol is implemented in a system as The RS in the TDBC protocol is implemented in a system
follows. At the beginning of every two-time-slot,7 the mutual as follows. At the beginning of every three-time-slot, the
TDBC
information {ℐ𝑖,𝑙ANC
: 𝑖 = 1, 2, 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿} is calculated mutual information {ℐ𝑖,𝑙 : 𝑖 = 1, 2, 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿} is
at both sources. Then one of the sources broadcasts selection calculated at both sources. Then one of the sources broadcasts
information to all the relays, and only the selected relay Rˆ𝑙 selection information to all the relays, and only the selected
helps bidirectional communication between S1 and S2 over relay R˜𝑙 helps bidirectional communication between S1 and
two time slots. Specifically, at the first time slot, S1 and S2 S2 over three time slots. Specifically, at the first time slot, S1
simultaneously broadcast their signals. At the second time broadcasts its signal; and at the second time slot, S2 broadcasts
slot, only the selected single relay Rˆ𝑙 among all the relays its signal. At the third time slot, only the selected single relay
multiplies the incoming signal, which is the sum of two signals R˜𝑙 among all the relays combines the two received signals and
transmitted from S1 and S2 , with an amplifying coefficient 𝜌ˆ𝑙 . broadcasts the combined signal to S1 and S2 .
Then Rˆ𝑙 broadcasts it to S1 and S2 . Using (26) with (27), the achieved system outage proba-
TDBC
Using (21) with (22), the achieved system outage probabil- bility 𝑃out (𝑅) of the RS in the TDBC protocol is given
ANC by
ity 𝑃out (𝑅) of the RS in the ANC protocol is given by
[ ] [ ]
ANC ANC ANC 𝑅 TDBC TDBC TDBC 𝑅
𝑃out (𝑅) = Pr max min[ℐ1,𝑙 , ℐ2,𝑙 ] < . (23) 𝑃out (𝑅) = Pr max min[ℐ1,𝑙 , ℐ2,𝑙 ]< .
𝑙=1,⋅⋅⋅ ,𝐿 2 𝑙=1,⋅⋅⋅ ,𝐿 2
(28)
Note that the selection rule in (22), which maximizes the
ANC ANC Note that the selection rule in (27), which maximizes the
minimum mutual information between ℐ1,𝑙 and ℐ2,𝑙 , also TDBC TDBC
minimum mutual information between ℐ1,𝑙 and ℐ2,𝑙 ,
minimizes the outage probability in (23). In Section IV, using
also minimizes the outage probability in (28). In Section V,
(23), we will derive the outage probability of the RS in the
using (28), we derive the outage probability of the RS in the
ANC protocol.
TDBC protocol.
( )
( ) 𝜉5 𝜖𝜉4 𝑤(𝜖𝜉4 + 𝜉5 )(𝜉1 + 𝜖(𝜉2 + 𝜉3 ))
Ψ1 𝑤; 𝜉1 , 𝜉2 , 𝜉3 , 𝜉4 , 𝜉5 = + exp −
𝜖𝜉4 + 𝜉5 𝜖𝜉4 + 𝜉5 𝜖𝜉1 𝜉3 Ω1,𝑙 Ω2,𝑙
( (𝜉 + 𝜉 ) )[ √ ( √ )
2 3 1 𝜉2 + 𝜉3 𝜉2 + 𝜉3
− exp −𝑤 + 2𝑤 𝐾1 2𝑤
𝜉1 𝜉3 𝜉4 𝜉3 𝜉5 𝜉1 𝜉32 Ω1,𝑙 Ω2,𝑙 𝜉1 𝜉32 Ω1,𝑙 Ω2,𝑙
∑∞ ( )𝑖+1 ( )]
(−1)𝑖 𝑤𝜖(𝜉2 + 𝜉3 ) 𝑤
− 𝐸𝑖+2 (33)
𝑖=0
𝑖! 𝜉1 𝜉3 𝜉5 𝜖𝜉3 𝜉4
In the following theorem, we derive the CDF of 𝒲𝑙 . Proof: Let 𝑈1 = max𝑙=1,⋅⋅⋅ ,𝐿 𝒲𝑙 . Since the CDF 𝐹𝑈1 (𝑤)
∏𝐿
Theorem 1: The CDF 𝐹𝒲𝑙 (𝑤) of 𝒲𝑙 is given by is given by 𝐹𝑈1 (𝑤) = 𝑙=1 𝐹𝒲𝑙 (𝑤), one can obtain (34). □
( )
𝐹𝒲𝑙 (𝑤) = Ψ1 𝑤; ℰ𝑠 , ℰ𝑠 , ℰ𝑟 , Ω1,𝑙 , Ω2,𝑙
( ) V. O UTAGE P ROBABILITY FOR RS IN TDBC P ROTOCOL
+ Ψ1 𝑤; ℰ𝑠 , ℰ𝑠 , ℰ𝑟 , Ω2,𝑙 , Ω1,𝑙 . (32)
( ) In this section, taking a step similar to (30), we first rewrite
In the above equation, Ψ1 𝑤; 𝜉1 , 𝜉2 , 𝜉3 , 𝜉4 , 𝜉5 is given at the the outage probability expression 𝑃outTDBC
(𝑅) in (28) for the
top of the next page. In the equation (33), 𝜖 = 𝜉1 (𝜉1 + 𝜉3 − RS in the TDBC protocol as follows:
𝜉2 )/(𝜉2 (𝜉2 + 𝜉3 − 𝜉1 )), 𝐾𝜈 (⋅) is the modified Bessel function [ (
of the second kind with order 𝜈, and 𝐸𝑖 (⋅) is the exponential TDBC
𝑃out (𝑅) = Pr log2 1 + 𝛾0TDBC
integral function for 𝑖 = 1, 2, ⋅ ⋅ ⋅ . ) ]
Proof: See Appendix A. □ + max min[𝛾1,𝑙 TDBC TDBC
, 𝛾2,𝑙 ] < 1.5𝑅 .
Note that Ψ1 (𝑤; 𝜉1 , 𝜉2 , 𝜉3 , 𝜉4 , 𝜉5 ) in (33) involves only stan- 𝑙=1,⋅⋅⋅ ,𝐿
8 For
∑∞ Proof: See Appendix B. □
a series 𝑖=0 𝑔𝑖 , we define ℒ := lim𝑖→∞ ∣𝑔𝑖+1 /𝑔𝑖 ∣. If ℒ < 1,
then the series converges absolutely; if ℒ > 1, then the series diverges; and Using (35) with the CDF 𝐹𝒱𝑙 (𝑤) in (37), we derive outage
if ℒ = 1, then the series is inclusive [33]. probability in the following theorem.
3506 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 58, NO. 12, DECEMBER 2010
TDBC TDBC
Theorem 4: The outage probability 𝑃out (𝑅) of the RS Theorem 5: The closed-form lower bound 𝑃out,LB (𝑅) of the
in the TDBC protocol is given by outage probability of the RS in the TDBC protocol is given
∫ 𝑅˜ (∏ 𝐿
) by
TDBC ˜ exp (−𝜂/(ℰ𝑠 Ω0 ))
𝑃out (𝑅) = 𝐹𝒱𝑙 (𝑅 − 𝜂) 𝑑𝜂, ( 𝑅˜ ) ∑ 𝐿 ∑𝐿 ∑𝐿
0 ℰ 𝑠 Ω0 TDBC
𝑃out,LB (𝑅) = 1 − exp − + (−1)𝑙 ⋅⋅⋅
𝑙=1
(39) ℰ 𝑠 Ω0 𝑚1 =1 𝑚𝑙 =1
𝑙=1
where 𝑅 ˜ = 21.5𝑅 − 1. !
TDBC 𝑚1 <⋅⋅⋅<𝑚𝑙
Proof: Let 𝑈2 = max𝑙=1,⋅⋅⋅ ,𝐿 𝒱𝑙 . Then 𝑃out (𝑅) in (35) ( ) ( )
˜ ∑𝑙 𝛽(𝑚𝑖 ) − exp − 𝑅/(ℰ
exp − 𝑅 ˜ 𝑠 Ω0 )
can be given by 𝑖=1
.
∑𝑙
TDBC
[ ( ) ] 1 − ℰ𝑠 Ω0 𝑖=1 𝛽(𝑚𝑖 )
𝑃out (𝑅) = Pr log2 1 + 𝛾0TDBC + 𝑈2 < 1.5𝑅 (44)
∫ 𝑅˜
[ ]
= Pr 𝑈2 < 𝑅 ˜ − 𝜂 𝑓𝛾 TDBC (𝜂)𝑑𝜂. (40) Proof: See Appendix D. □
0
TDBC
0 Note that since the lower bound 𝑃out,LB (𝑅) of (44) is given
∏ in closed-form, it does not require any numerical integration.
Since the CDF 𝐹𝑈2 (𝑤) is given by 𝐹𝑈2 (𝑤) = 𝐿 𝑙=1 𝐹𝒱𝑙 (𝑤) TDBC
and 𝛾0TDBC is an exponentially distributed random variable Also, note that 𝑃out,LB (𝑅) of (44) reduces to [12, eq. (26)]
with mean ℰ𝑠 Ω0 , one can obtain (39). □ when 𝐿 = 1 and 𝛼1,𝑙 = 𝛼2,𝑙 = 1/2 for 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿.
TDBC Therefore, the analysis in this subsection can be considered
Although the obtained outage probability 𝑃out (𝑅) in
as a generalization of that of [12].
(39) is very accurate, the final expression is given in a one-
integral form, which requires numerical integration. Therefore,
it should be also useful to derive a closed-form bound of the VI. S IMULATION R ESULTS
outage probability. In the next subsection, we derive a lower In this section, we first discuss the convergence speed of
bound of the outage probability in closed-form, which does the series expression in (33), which was analytically shown
not involve any numerical integration. to be convergent in Section IV. Then, for the proposed RS in
the ANC and TDBC protocols, we check the accuracy of the
B. Lower Bound of Outage Probability in Closed-Form obtained outage probabilities by comparing our analysis with
Monte Carlo simulations. Finally, we compare the RS in the
In the performance analysis of relay networks, the following ANC protocol with the RS in the TDBC protocol.
inequality has widely been used [13], [27]: 𝜃1 𝜃2 /(𝜃1 + 𝜃2 ) <
min[𝜃1 , 𝜃2 ], where 𝜃1 > 0 and 𝜃2 > 0. Using this in-
equality, the minimum value between 𝛾1,𝑙 TDBC TDBC
and 𝛾2,𝑙 , A. Convergence Speed of Series Expression in (33)
TDBC TDBC
min[𝛾1,𝑙 , 𝛾2,𝑙 ], is upper-bounded by In this subsection, we check the convergence speed of the
(𝑁 )
TDBC TDBC
series expression in (33). We let Ψ̂1 (𝑤; 𝜉1 , 𝜉2 , 𝜉3 , 𝜉4 , 𝜉5 )
min[𝛾1,𝑙 , 𝛾2,𝑙 ] denote a truncated version of Ψ1 (𝑤; 𝜉1 , 𝜉2 , 𝜉3 , 𝜉4 , 𝜉5 ) with
[ [ ] window size 𝑁 , i.e.,
𝛼1,𝑙 ℰ𝑠 ℰ𝑟 truncation ∑ ∑𝑁Ψ1 (𝑤; 𝜉1 , 𝜉2 , 𝜉3 , 𝜉4 , 𝜉5 ) of
< min min ℰ𝑠 𝑋1,𝑙 , 𝑋2,𝑙 , ∞
ℰ𝑠 + 𝛼2,𝑙 ℰ𝑟 (33) with 𝑖=0 replaced by 𝑖=0 . Then we define Ξ(𝑁 )
[ 𝛼 ℰ ℰ ]] as the truncation error normalized by Ψ1 (𝑤; 𝜉1 , 𝜉2 , 𝜉3 , 𝜉4 , 𝜉5 )
min
2,𝑙 𝑠 𝑟
𝑋1,𝑙 , ℰ𝑠 𝑋2,𝑙 as follows:
ℰ𝑠 + 𝛼1,𝑙 ℰ𝑟
[ ] Ψ1 (𝑤; 𝜉1 , 𝜉2 , 𝜉3 , 𝜉4 , 𝜉5 ) − Ψ̂(𝑁)
1 (𝑤; 𝜉1 , 𝜉2 , 𝜉3 , 𝜉4 , 𝜉5 )
𝛼1,𝑙 ℰ𝑠 ℰ𝑟 𝛼2,𝑙 ℰ𝑠 ℰ𝑟 Ξ(𝑁 ) :=
= min ℰ𝑠 𝑋1,𝑙 , 𝑋2,𝑙 , 𝑋1,𝑙 , ℰ𝑠 𝑋2,𝑙 Ψ1 (𝑤; 𝜉1 , 𝜉2 , 𝜉3 , 𝜉4 , 𝜉5 )
ℰ𝑠 + 𝛼2,𝑙 ℰ𝑟 ℰ𝑠 + 𝛼1,𝑙 ℰ𝑟 ( (𝜉 + 𝜉 )
2 3 1 )
U = exp −𝑤 +
=: 𝒱𝑙 . (41) 𝜉1 𝜉3 𝜉4 𝜉3 𝜉5
∑ ( )𝑖+1 ( )
∞ (−1)𝑖 𝑤𝜖(𝜉2 +𝜉3 )
In the following lemma, we derive the CDF 𝐹𝒱𝑙U (𝑤) of 𝒱𝑙U . 𝑖=𝑁+1 𝑖! 𝜉1 𝜉3 𝜉5
𝐸𝑖+2 𝜖𝜉𝑤𝜉
3 4
× .
Lemma 3: The CDF 𝐹𝒱𝑙U (𝑤) of 𝒱𝑙U is given by Ψ1 (𝑤; 𝜉1 , 𝜉2 , 𝜉3 , 𝜉4 , 𝜉5 )
(45)
𝐹𝒱𝑙U (𝑤) = 1 − exp (−𝑤𝛽(𝑙)) , (42)
Fig. 3 shows the normalized truncation error Ξ(𝑁 ) against
where truncation window size 𝑁 . We set 𝑤 = 2, 𝜉1 = 𝜉2 = 𝜉3 = ℰ,
⎧ and 𝜉4 = 𝜉5 = 1. For various ℰ values, ℰ = 10, 20, 30
𝛼2,𝑙 (ℰ𝑠 +𝛼2,𝑙 ℰ𝑟 )Ω1,𝑙 +𝛼1,𝑙 (ℰ𝑠 +𝛼1,𝑙 ℰ𝑟 )Ω2,𝑙
⎨ 𝛼1,𝑙 𝛼2,𝑙 ℰ𝑠 ℰ𝑟 Ω1,𝑙 Ω2,𝑙 , 𝐶𝑎𝑠𝑒 1, dB, we can see the normalized truncation error is very small
ℰ𝑠 Ω1,𝑙 +𝛼2,𝑙 ℰ𝑟 Ω1,𝑙 +𝛼1,𝑙 ℰ𝑟 Ω2,𝑙
𝛽(𝑙) = 𝛼1,𝑙 ℰ𝑠 ℰ𝑟 Ω1,𝑙 Ω2,𝑙 , 𝐶𝑎𝑠𝑒 2, even with a small value 𝑁 . For instance, the normalized
truncation error is smaller than 10−7 when ℰ = 20 dB and
⎩ ℰ𝑠 Ω2,𝑙 +𝛼2,𝑙 ℰ𝑟 Ω1,𝑙 +𝛼1,𝑙 ℰ𝑟 Ω2,𝑙
, 𝐶𝑎𝑠𝑒 3.
𝛼2,𝑙 ℰ𝑠 ℰ𝑟 Ω1,𝑙 Ω2,𝑙
𝑁 = 1. Therefore, even with a small value 𝑁 , substituting
(43) (𝑁 )
Ψ̂1 (𝑤; 𝜉1 , 𝜉2 , 𝜉3 , 𝜉4 , 𝜉5 ) into (32) and (37) gives very ac-
Proof: See Appendix C. □
curate closed-form approximations 𝐹ˆ𝒲𝑙 (𝑤) and 𝐹ˆ𝒱𝑙 (𝑤) to
(𝑁 ) (𝑁 )
Note that the CDF 𝐹𝒱𝑙U (𝑤) in (42) is expressed by the
actual 𝐹𝒲𝑙 (𝑤) and 𝐹𝒱𝑙 (𝑤), respectively. Using 𝐹ˆ𝒲𝑙 (𝑤) and
(𝑁 )
exponential function. Taking a step similar to (39) and using
𝐹ˆ𝒱𝑙 (𝑤), in the following, we calculate outage probabilities
(𝑁 )
the CDF 𝐹𝒱𝑙U (𝑤) in (42), we derive a closed-form lower bound
of outage probability in the following theorem. of (34) and (39).
JU and KIM: RELAY SELECTION WITH ANC AND TDBC PROTOCOLS IN BIDIRECTIONAL RELAY NETWORKS 3507
0
10
Ξ(N) for E = 10 dB 0 Analysis using eq. (34)
10
Ξ(N) for E = 20 dB Simulation with exact ρl in eq. (3)
Ξ(N) for E = 30 dB Simulation with approximate ρ in eq. (4)
l
−2
10
−1
10
−4 L=1
10
Normalized trucation error
−2
10
Outage probability
−6
10 −3
10 L=2
−8
10 −4
10
L=3
−10
10 −5
10
L=4
−12 −6
10 10
0 1 2 3 4 5 6 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
N d
S ,R
1
−3
L=1
probability against 10 log10 ℰ of the RS in the TDBC protocol,
10
where we set 𝐿 = 1, 2, 3, 4, 𝑅 = 1 bps/Hz, ℰ𝑠 = ℰ𝑟 = ℰ,
Ω0 = Ω1,𝑙 = Ω2,𝑙 = 1, and 𝛼1,𝑙 = 𝛼2,𝑙 = 1/2 for 𝑙 =
1, ⋅ ⋅ ⋅ , 𝐿. Also, we set 𝑁 = 2 in 𝐹ˆ𝒱𝑙 (𝑤) for 𝐹𝒱𝑙 (𝑤) of
−4 (𝑁 )
10
(39). From Fig. 6, we can see that (39) is very accurate even
−5
10
L=2 with 𝑁 = 2, and that (44) is tight. As in Fig. 5, we investigate
L=3
L=4
the effect of relay location. Fig. 7 shows the outage probability
against 𝑑𝑆1 ,𝑅 of the RS in the TDBC protocol, where we set
𝐿 = 1, 2, 3, 4, 𝑅 = 2 bps/Hz, ℰ𝑠 = ℰ𝑟 = 15 dB, Ω0 = 1,
−6
10
0 5 10 15 20 25 30 35 40 45
10 log
10
E
Ω1,𝑙 = 𝑑−4 𝑆1 ,𝑅 , Ω2,𝑙 = (1 − 𝑑𝑆1 ,𝑅 )
−4
, and 𝛼1,𝑙 = 𝛼2,𝑙 = 1/2
for 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿. Also, we set 𝑁 = 2 in 𝐹ˆ𝒱𝑙 (𝑤) for 𝐹𝒱𝑙 (𝑤)
(𝑁 )
Fig. 4. Outage probability against 10 log10 ℰ of the RS in the ANC protocol. of (39). Irrespective of relay location in Fig. 7, we can see
𝐿 = 1, 2, 3, 4. 𝑅 = 1 bps/Hz. ℰ𝑠 = ℰ𝑟 = ℰ. Ω1,𝑙 = Ω2,𝑙 = 1 where
(𝑁) that (39) is very accurate even with 𝑁 = 2, and (44) is tight.
𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿. 𝑁 = 1 in 𝐹ˆ𝒲 (𝑤) for 𝐹𝒲𝑙 (𝑤) of (34).
𝑙
very accurate even with 𝑁 = 1. In Fig. 5, we investigate the for 𝐹𝒱𝑙 (𝑤) of (39). With a simple modification of [12], for
effect of relay location. Fig. 5 shows the outage probability the RS with the ANC protocol and the RS with the TDBC
against 𝑑𝑆1 ,𝑅 of the RS in the ANC protocol, where we set protocol, it can be shown that the diversity orders are 𝐿 and
𝐿 = 1, 2, 3, 4, 𝑅 = 1.5 bps/Hz, ℰ𝑠 = ℰ𝑟 = 10 dB, Ω1,𝑙 = 𝐿 + 1, respectively, and the multiplexing gains are 1 and 2/3,
𝑑−4
𝑆1 ,𝑅 , and Ω2,𝑙 = (1 − 𝑑𝑆1 ,𝑅 )
−4
for 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿. Also, we respectively. Thus, in the high SNR regime, the RS with the
ˆ (𝑁 )
set 𝑁 = 1 in 𝐹𝒲𝑙 (𝑤) for 𝐹𝒲𝑙 (𝑤) of (34). Irrespective of TDBC protocol always outperforms the RS with the ANC
relay location in Fig. 5, we can see that (34) is very accurate protocol irrespective of the target rate 𝑅. In the low and
even with 𝑁 = 1. medium SNR regimes, however, one scheme outperforms the
3508 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 58, NO. 12, DECEMBER 2010
0
10
Analysis using eq. (39) RS with ANC: analysis using eq. (34)
2
Lower bound using eq. (44) 10 RS with ANC: simulation with exact ρl in eq. (3)
Simulation with exact ζi,l in eq. (12)
RS with ANC: simulation with approximate ρ in eq. (4)
l
−1
10 Simulation with approximate ζi,l in eq. (13) 1 RS with TDBC: analysis using eq. (39)
10
RS with TDBC: simulation with exact ζi,l in eq. (12)
RS with TDBC: simulation with approximate ζ in eq. (13)
0
i,l
10
−2
10
Outage probability
Outage probability
−1
10
−3
10
−2
10
−3
−4 10
10
R=1
−4 R=2
10
L=1 R=4
−5 L=3 R=3
10 L=2
−5
L=4 10
−6 −6
10 10
0 5 10 15 20 25 30 35 0 5 10 15 20 25 30 35
10 log10 E 10 log E
10
Fig. 6. Outage probability against 10 log10 ℰ of the RS in the TDBC pro- Fig. 8. Outage probabilities against 10 log10 ℰ of the RS with the ANC
tocol. 𝐿 = 1, 2, 3, 4. 𝑅 = 1 bps/Hz. ℰ𝑠 = ℰ𝑟 = ℰ. Ω0 = Ω1,𝑙 = Ω2,𝑙 = 1 protocol and the RS with the TDBC protocol. 𝐿 = 4. 𝑅 = 1, 2, 3, 4 bps/Hz.
(𝑁)
and 𝛼1,𝑙 = 𝛼2,𝑙 = 1/2 where 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿. 𝑁 = 2 in 𝐹ˆ𝒱 (𝑤) for ℰ𝑠 = ℰ𝑟 = ℰ. Ω0 = Ω1,𝑙 = Ω2,𝑙 = 1, and 𝛼1,𝑙 = 𝛼2,𝑙 = 1/2 for the RS
(𝑁)
with the TDBC protocol where 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿. 𝑁 = 2 in both 𝐹ˆ𝒲 (𝑤) for
𝑙
𝐹𝒱𝑙 (𝑤) of (39). 𝑙
(𝑁)
𝐹𝒲 (𝑤) of (34) and 𝐹ˆ 𝑙
(𝑤) for 𝐹𝒱 (𝑤) of (39). 𝒱𝑙 𝑙
0 1
10 10
Analysis using eq. (39) RS with ANC: analysis using eq. (34)
Lower bound using eq. (44) RS with ANC: simulation with exact ρ in eq. (3)
l
Simulation with exact ζi,l in eq. (12)
RS with ANC: simulation with approximate ρ in eq. (4)
0
l
10
−1 Simulation with approximate ζi,l in eq. (13) 10 RS with TDBC: analysis using eq. (39)
RS with TDBC: simulation with exact ζ in eq. (12)
i,l
RS with TDBC: simulation with approximate ζ in eq. (13)
L=1 i,l
−1
−2 10
10
Outage probability
Outage probability
R=3
−2
−3 L=2 10
10
−3 R=2
−4
10
10 L=3
−4
R=1
10
−5
10
L=4
−5
10
−6
10
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
d d
S ,R S ,R
1 1
Fig. 7. Outage probability against 𝑑𝑆1 ,𝑅 of the RS in the TDBC protocol. Fig. 9. Outage probabilities against 𝑑𝑆1 ,𝑅 of the RS with the ANC protocol
𝐿 = 1, 2, 3, 4. 𝑅 = 2 bps/Hz. ℰ𝑠 = ℰ𝑟 = 15 dB. Ω0 = 1, Ω1,𝑙 = 𝑑−4 𝑆1 ,𝑅 ,
and the RS with the TDBC protocol. 𝐿 = 2. 𝑅 = 1, 2, 3 bps/Hz. ℰ𝑠 = ℰ𝑟 =
Ω2,𝑙 = (1 − 𝑑𝑆1 ,𝑅 )−4 , and 𝛼1,𝑙 = 𝛼2,𝑙 = 1/2 where 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿. 15 dB. Ω0 = 1, Ω1,𝑙 = 𝑑−4 𝑆1 ,𝑅 , Ω2,𝑙 = (1 − 𝑑𝑆1 ,𝑅 )
−4 , and 𝛼
1,𝑙 = 𝛼2,𝑙 =
(𝑁)
𝑁 = 2 in 𝐹ˆ𝒱 (𝑤) for 𝐹𝒱𝑙 (𝑤) of (39). 1/2 for the RS with the TDBC protocol where 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿. 𝑁 = 2 in both
(𝑁) (𝑁)
𝑙
𝐹ˆ𝒲 (𝑤) for 𝐹𝒲𝑙 (𝑤) of (34) and 𝐹ˆ𝒱 (𝑤) for 𝐹𝒱𝑙 (𝑤) of (39).
𝑙 𝑙
VII. C ONCLUSIONS
other scheme depending on 𝑅 and SNR value. As in Figs. 5
and 7, we investigate the effect of relay location. Fig. 9 shows In this paper, we have studied RS for the ANC and TDBC
the outage probabilities against 𝑑𝑆1 ,𝑅 of the RS with the ANC protocols in a bidirectional relay network consisting of two
protocol and the RS with the TDBC protocol, where we set end-sources and multiple relays. A single best relay was
𝐿 = 2, 𝑅 = 1, 2, 3 bps/Hz, ℰ𝑠 = ℰ𝑟 = 15 dB, Ω0 = 1, selected depending on channel conditions to help bidirectional
Ω1,𝑙 = 𝑑−4
𝑆1 ,𝑅 , Ω2,𝑙 = (1 − 𝑑𝑆1 ,𝑅 )
−4
, and 𝛼1,𝑙 = 𝛼2,𝑙 = 1/2 communication between the two end-sources. Specifically, we
for the RS in the TDBC protocol with 𝑙 = 1, ⋅ ⋅ ⋅ , 𝐿. Also, we have selected a single best relay based on a max-min criterion
set 𝑁 = 2 in both 𝐹ˆ𝒲𝑙 (𝑤) for 𝐹𝒲𝑙 (𝑤) of (34) and 𝐹ˆ𝒱𝑙 (𝑤)
(𝑁 ) (𝑁 )
to minimize the outage probabilities of the ANC and TDBC
for 𝐹𝒱𝑙 (𝑤) of (39). Irrespective of relay location in Fig. 9, as protocols. Then, for the RS in the ANC protocol, we have
the target rate 𝑅 increases, we can see the RS with the ANC derived a closed-form expression of the outage probability.
protocol outperforms the RS with the TDBC protocol. Also, for the RS in the TDBC protocol, we have derived
JU and KIM: RELAY SELECTION WITH ANC AND TDBC PROTOCOLS IN BIDIRECTIONAL RELAY NETWORKS 3509
a one-integral form of the outage probability and its lower Then 𝑃1 in (A.7) can be calculated by
bound in closed-form. For future work, we will investigate ( )
∫ ℰ𝑤 1+ (ℰ𝑠ℰ+ℰ𝑟 ) ∫ 𝑡
RS with optimal power allocation when each terminal has a 𝑃1 =
𝑟 𝑠
𝑓𝑋1,𝑙 (𝑞)𝑓𝑋2,𝑙 (𝑡)𝑑𝑞𝑑𝑡
single antenna, and RS with precoding or beamforming when 𝑡=0 𝑞=0
each terminal has multiple antennas. ∫ ∞ ∫ (ℰ𝑠 +ℰ𝑟 )𝑡𝑤
ℰ𝑠 ℰ𝑟 𝑡−ℰ𝑠 𝑤
+ ( ) 𝑓𝑋1,𝑙 (𝑞)𝑓𝑋2,𝑙 (𝑡)𝑑𝑞𝑑𝑡
𝑡= ℰ𝑤𝑟 1+ (ℰ𝑠ℰ+ℰ 𝑟)
𝑞=0
A PPENDIX A 𝑠
( )
P ROOF OF T HEOREM 1 ∫ ∞ ∫ 𝑤
ℰ𝑟 1+
(ℰ𝑠 +ℰ𝑟 )
ℰ𝑠
( √ ) 𝑟 1,𝑙 Ω2,𝑙
[ (ℰ + ℰ )𝑋 𝑤 ]
min
𝑠 𝑟 2,𝑙
, 𝑋2,𝑙 ×𝐾1 2𝑤 ℰ𝑠 ℰℰ2𝑠Ω+ℰ 𝑟
. For the second part of the right-
𝑟 1,𝑙 Ω2,𝑙
ℰ𝑠 ℰ𝑟 𝑋2,𝑙 − ℰ𝑠 𝑤 hand side of (A.11), since we have no closed-form solution,
⎧ ( )
⎨ 𝑋2,𝑙 , 𝑤 𝑤 (ℰ𝑠 +ℰ𝑟 )
we tackle this problem by adopting
=
ℰ𝑟 < 𝑋 2,𝑙 ≤
( ℰ𝑟
1 +
)ℰ𝑠
, ∑∞a 𝑥series𝑖
expansion of
⎩ (ℰ +ℰ )𝑋 𝑤 𝑤 (ℰ +ℰ ) exponential function, exp(𝑥) = 𝑖=0 𝑖! . Substituting the
ℰ𝑠 ℰ𝑟 𝑋2,𝑙 −ℰ𝑠 𝑤 , 𝑋2,𝑙 > ℰ𝑟 1 + .
𝑠 𝑟 2,𝑙 𝑠 𝑟
ℰ𝑠 series expansion of exp(−𝑧/Ω2,𝑙 ) into the second part of the
(A.6) right-hand side of (A.11) yields
Referring to Fig. 10 and substituting (A.6) into (A.5) yield ∫ 𝑤(ℰℰ𝑠 ℰ+ℰ𝑟 ) ( )
[ ] 𝐴 𝑠 𝑟 𝑤2 𝑧
𝑤 ( (ℰ𝑠 + ℰ𝑟 ) ) exp − − 𝑑𝑧
𝑃1 = Pr 𝑋1,𝑙 < 𝑋2,𝑙 , 𝑋2,𝑙 ≤ 1+ Ω2,𝑙 0 ℰ𝑠 ℰ𝑟2 Ω1,𝑙 𝑧 Ω2,𝑙
ℰ𝑟 ℰ𝑠 ∫ 1∑ ∞ ( 𝑎 )
[ (−1)𝑖 𝑎𝑖+1 1 𝑧 ¯𝑖 2
(ℰ𝑠 + ℰ𝑟 )𝑋2,𝑙 𝑤 =𝐴 exp − 𝑑¯
𝑧 (A.12)
+ Pr 𝑋1,𝑙 < , 0 𝑖=0 𝑖! 𝑧
¯
ℰ𝑠 ℰ𝑟 𝑋2,𝑙 − ℰ𝑠 𝑤
] ∑∞
(−1)𝑖 𝑎𝑖+1
𝑤( (ℰ𝑠 + ℰ𝑟 ) ) = 𝐴 1
𝐸𝑖+2 (𝑎2 ), (A.13)
𝑋2,𝑙 > 1+ . (A.7) 𝑖!
ℰ𝑟 ℰ𝑠 𝑖=0
3510 IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 58, NO. 12, DECEMBER 2010
(C.1)
Fig. 10. 𝑃1 on the (𝑋1,𝑙 , 𝑋2,𝑙 ) plane.
For 𝐶𝑎𝑠𝑒 1 of (C.1), the CDF 𝐹𝒱𝑙U (𝑤) of 𝒱𝑙U is given at the
top of the next page. With a simple integration, one can obtain
where 𝑎1 = 𝑤(ℰ𝑠 + ℰ𝑟 )/(ℰ𝑠 ℰ𝑟 Ω2,𝑙 ) and 𝑎2 = 𝑤/(ℰ𝑟 Ω1,𝑙 ). (42) with the first case of (43). Taking steps similar to those
Finally, substituting (A.13) into (A.11) yields 𝑃13 , which in used from (C.2) to (C.3), one can obtain (42) with the second
turn is substituted into 𝑃1 of (A.8) with 𝑃11 − 𝑃12 of (A.9). and third cases of (43).
Taking steps similar to those used from (A.4) to (A.13), one
can obtain 𝑃2 . Then, using (A.3) with 𝑃1 and 𝑃2 , one can A PPENDIX D
obtain the final expression of the CDF 𝐹𝒲𝑙 (𝑤), which is given P ROOF OF T HEOREM 5
in (32). Taking a step similar to (39), one can obtain the lower bound
TDBC
𝑃out,LB (𝑅) as follows:
∫ 𝑅˜ (∏ 𝐿
)
A PPENDIX B TDBC ˜ exp (−𝜂/(ℰ𝑠 Ω0 ))
𝑃out,LB (𝑅) = 𝐹𝒱𝑙U (𝑅 − 𝜂) 𝑑𝜂.
P ROOF OF T HEOREM 3 0 ℰ 𝑠 Ω0
𝑙=1
∏𝐿 ( ) (D.1)
TDBC TDBC
From (17) and (18), the difference 𝛾1,𝑙 − 𝛾2,𝑙 is Using the product identity 𝑙=1 1 − 𝑑(𝑙) = 1 +
∑𝐿 ∑𝐿 ∏𝑙
given at the top of the next page. Depending on the values 𝑙=1 (−1)
𝑙
𝑚1 =1,⋅⋅⋅ ,𝑚𝑙 =1 𝑖=1 𝑑(𝑚𝑖 ), one can rewrite the
(ℰ𝑠 , ℰ𝑟 , 𝛼1,𝑙 , 𝑋1,𝑙 , 𝑋2,𝑙 ), 𝒱𝑙 in (36) is given by ∏𝐿 𝑚1 <⋅⋅⋅<𝑚𝑙
expression 𝑙=1 𝐹𝒱𝑙𝑈 (𝑤) in (D.1) as follows:
⎧ TDBC TDBC
⎨ min[𝛾1,𝑙 , 𝛾2,𝑙 ], 𝐶𝑎𝑠𝑒 1, 𝐿
∏
TDBC
𝒱𝑙 = 𝛾1,𝑙 , 𝐶𝑎𝑠𝑒 2, (B.2) 𝐹𝒱𝑙U (𝑤)
⎩ TDBC
𝛾2,𝑙 , 𝐶𝑎𝑠𝑒 3. 𝑙=1
𝐿 𝐿 𝐿
( 𝑙
)
∑ ∑ ∑ ∑
𝑙
For 𝐶𝑎𝑠𝑒 1 of (B.2), the CDF 𝐹𝒱𝑙 (𝑤) of 𝒱𝑙 is given by =1+ (−1) ⋅⋅⋅ exp −𝑤 𝛽(𝑚𝑖 ) .
𝑙=1 𝑚1 =1 𝑚𝑙 =1 𝑖=1
[ TDBC ] !
𝐹𝒱𝑙 (𝑤) = Pr 𝛾2,𝑙 < 𝑤, ℰ𝑒𝑞 𝑋1,𝑙 < 𝑋2,𝑙 𝑚1 <⋅⋅⋅<𝑚𝑙
[ TDBC ] (D.2)
+ Pr 𝛾1,𝑙 < 𝑤, ℰ𝑒𝑞 𝑋1,𝑙 ≥ 𝑋2,𝑙 , (B.3)
Then, substituting (D.2) into (D.1) and taking a simple inte-
where ℰ𝑒𝑞 = (1−𝛼1,𝑙 )(ℰ𝑠 −(2𝛼1,𝑙 −1)ℰ𝑟 )/(𝛼1,𝑙 (ℰ𝑠 +(2𝛼1,𝑙 − gration, one can obtain the final closed-form result in (44).
1)ℰ𝑟 )). Then, taking steps similar to those used from (A.3) to
(A.13), one can obtain the first case of (37). For 𝐶𝑎𝑠𝑒 2 of R EFERENCES
(B.2), taking steps similar to the proof of [23, Lemma 1], one [1] A. Sendonaris, E. Erkip, and B. Aazhang, “User cooperation
can obtain the CDF 𝐹𝒱𝑙 (𝑤) of 𝒱𝑙 as follows: diversity—part I, II,” IEEE Trans. Commun., vol. 51, pp. 1927–1948,
Nov. 2003.
[2] J. N. Laneman, D. N. C. Tse, and G. W. Wornell, “Cooperative di-
𝐹𝒱𝑙 (𝑤) versity in wireless networks: efficient protocols and outage behavior,”
[ ] IEEE Trans. Inf. Theory, vol. 50, pp. 3062–3080, Dec. 2004.
𝛼1,𝑙 ℰ𝑠 ℰ𝑟 𝑋1,𝑙 𝑋2,𝑙 [3] B. Rankov and A. Wittneben, “Spectral efficient protocols for half-
= Pr <𝑤
(ℰ + 𝛼2,𝑙 ℰ𝑟 )𝑋1,𝑙 + 𝛼1,𝑙 ℰ𝑟 𝑋2,𝑙 duplex fading relay chnnels,” IEEE J. Sel. Areas Commun., vol. 25,
( 𝑠 ) pp. 379–389, Feb. 2007.
= Ψ2 𝑤; (ℰ𝑠 + 𝛼2,𝑙 ℰ𝑟 )Ω1,𝑙 , 𝛼1,𝑙 ℰ𝑟 Ω2,𝑙 , 1 + 𝛼2,𝑙 ℰ𝑟 /ℰ𝑠 , [4] S. Katti, S. Gollakota, and D. Katabi, “Embracing wireless interfer-
(B.4) ence: analog network coding,” in Proc. ACM SIGCOMM, 2007, pp.
397–408.
[5] P. Popovski and H. Yomo, “Wireless network coding by amplify-and-
where Ψ2 (𝑤; ⋅, ⋅, ⋅) is given in (38). Finally, for 𝐶𝑎𝑠𝑒 3 forward for bi-directional traffic flows,” IEEE Commun. Lett., vol. 11,
of (B.2), taking a step similar to (B.4), one can solve pp. 16–18, Jan. 2007.
the( CDF 𝐹𝒱𝑙 (𝑤) of 𝒱𝑙 , which is given by 𝐹)𝒱𝑙 (𝑤) = [6] X. Tang and Y. Hua, “Optimal design of non-regenerative MIMO
wireless relays,” IEEE Trans. Wireless Commun., vol. 6, pp. 1398–
Ψ2 𝑤; 𝛼2,𝑙 ℰ𝑟 Ω1,𝑙 , (ℰ𝑠 + 𝛼1,𝑙 ℰ𝑟 )Ω2,𝑙 , 1 + 𝛼1,𝑙 ℰ𝑟 /ℰ𝑠 . 1407, Apr. 2007.
[7] R. Zhang, Y.-C. Liang, C. C. Chai, and S. Cui, “Optimal beamforming
for two-way multi-antenna relay channel with analogue network
coding,” IEEE J. Sel. Areas Commun., vol. 27, pp. 699–712, June
A PPENDIX C 2009.
JU and KIM: RELAY SELECTION WITH ANC AND TDBC PROTOCOLS IN BIDIRECTIONAL RELAY NETWORKS 3511
( )
ℰ𝑠 ℰ𝑟 𝑋1,𝑙 𝑋2,𝑙 𝛼1,𝑙 (ℰ𝑠 + (2𝛼1,𝑙 − 1)ℰ𝑟 )𝑋2,𝑙 − (1 − 𝛼1,𝑙 )(ℰ𝑠 − (2𝛼1,𝑙 − 1)ℰ𝑟 )𝑋1,𝑙
TDBC TDBC
𝛾1,𝑙 − 𝛾2,𝑙 = ( )( ) (B.1)
(ℰ𝑠 + (1 − 𝛼1,𝑙 )ℰ𝑟 )𝑋1,𝑙 + 𝛼1,𝑙 ℰ𝑟 𝑋2,𝑙 (1 − 𝛼1,𝑙 )ℰ𝑟 𝑋1,𝑙 + (ℰ𝑠 + 𝛼1,𝑙 ℰ𝑟 )𝑋2,𝑙
[ ]
(1 − 𝛼1,𝑙 )ℰ𝑠 ℰ𝑟 𝛼1,𝑙 (ℰ𝑠 + 𝛼1,𝑙 ℰ𝑟 )
𝐹𝒱𝑙U (𝑤) = Pr 𝑋1,𝑙 < 𝑤, 𝑋1,𝑙 < 𝑋2,𝑙
ℰ + 𝛼1,𝑙 ℰ𝑟 (1 − 𝛼1,𝑙 )(ℰ𝑠 + (1 − 𝛼1,𝑙 )ℰ𝑟 )
[𝑠 ]
𝛼1,𝑙 ℰ𝑠 ℰ𝑟 𝛼1,𝑙 (ℰ𝑠 + 𝛼1,𝑙 ℰ𝑟 )
+ Pr 𝑋2,𝑙 < 𝑤, 𝑋1,𝑙 ≥ 𝑋2,𝑙 (C.2)
ℰ𝑠 + (1 − 𝛼1,𝑙 )ℰ𝑟 (1 − 𝛼1,𝑙 )(ℰ𝑠 + (1 − 𝛼1,𝑙 )ℰ𝑟 )
∫ ℰ𝑠 +𝛼1,𝑙 ℰ𝑟 𝑤 ∫ ∞
(1−𝛼1,𝑙 )ℰ𝑠 ℰ𝑟
= (1−𝛼 )(ℰ +(1−𝛼 )ℰ )
𝑓𝑋1,𝑙 (𝑞)𝑓𝑋2,𝑙 (𝑡)𝑑𝑡𝑑𝑞
1,𝑙 𝑠 1,𝑙 𝑟
𝑞=0 𝑡= 𝛼1,𝑙 (ℰ𝑠 +𝛼1,𝑙 ℰ𝑟 )
𝑞
[8] Y. Han, S. H. Ting, C. K. Ho, and W. H. Chin, “Performance bounds [27] P. A. Anghel and M. Kaveh, “Exact symbol error probability of a
for two-way amplify-and-forward relaying,” IEEE Trans. Wireless cooperative network in a Rayleigh-fading environment,” IEEE Trans.
Commun., vol. 8, pp. 432–439, Jan. 2009. Wireless Commun., vol. 3, pp. 1416–1421, Sep. 2004.
[9] Q. Li, S. H. Ting, A. Pandharipande, and Y. Han, “Adaptive two-way [28] M. O. Hasna and M.-S. Alouini, “End-to-end performance of trans-
relaying and outage analysis ” IEEE Trans. Wireless Commun., vol. mission systems with relays over Rayleigh-fading channels,” IEEE
8, pp. 3288–3299, June 2009. Trans. Wireless Commun., vol. 2, pp. 1126–1131, Nov. 2003.
[10] S. J. Kim, P. Mitra, and V. Tarokh, “Performance bounds for bidi- [29] A. Ribeiro, C. Xiaodong, and G. B. Giannakis, “Symbol error
rectional coded cooperation protocols,” IEEE Trans. Inf. Theory, vol. probabilities for general cooperative links,” IEEE Trans. Wireless
54, pp. 5235–5241, Nov. 2008. Commun., vol. 4, pp. 1264–1273, May 2005.
[11] S. J. Kim, N. Devroye, P. Mitra, and V. Tarokh, “Achievable rate [30] I. Hammerstom, M. Kuhn, C. Esli, J. Zhao, A. Wittneben, and G.
regions for bi-directional relaying,” IEEE Trans. Inf. Theory, to be Bauch, “MIMO two-way relaying with transmit CSI at the relay,” in
published. Proc. SPAWC, June 2007, pp. 1–5.
[12] Z. Yi, M. Ju, and I.-M. Kim, “Outage probability and optimum [31] N. Lee, H. J. Yang, and J. Chun, “Achievable sum-rate maximizing
combining for time division broadcast protocol,” to be published. AF relay beamforming scheme in two-way relay channels,” in Proc.
[13] A. Bletsas, A. Khisti, D. P. Reed, and A. Lippman, “A simple IEEE ICC, May 2008, pp. 300–305.
cooperative diversity method based on network path selection,” IEEE [32] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series and
J. Sel. Areas Commun., vol. 24, pp. 659–672, Mar. 2006. Products, 6th edition. San Diego, CA: Academic Press, 2000.
[14] A. Bletsas, H. Shin, and M. Z. Win, “Cooperative communications [33] W. Rudin, Principles of Mathematical Analysis, 3rd edition. New
with outage-optimal opportunistic relaying,” IEEE Trans. Wireless York: McGraw-Hill, 1976.
Commun., vol. 6, pp. 3450–3460, Sep. 2007.
[15] Y. Zhao, R. Adve, and T. J. Lim, “Improving amplify-and-forward MinChul Ju received the B.S. degree in electrical
relay networks: optimal power allocation versus selection,” IEEE engineering from Pohang University of Science and
Trans. Wireless Commun., vol. 6, pp. 3114–3123, Aug. 2007. Technology (POSTECH), Pohang, Korea, in 1997,
[16] E. Beres and R. Adve, “Selection cooperation in multi-source cooper- the M.S. degree in electrical engineering from Ko-
ative networks,” IEEE Trans. Wireless Commun., vol. 7, pp. 118–127, rea Advanced Institute of Science and Technology
Jan. 2008. (KAIST), Taejon, Korea, in 1999, and the Ph.D.
[17] D. S. Michalopoulos and G. K. Karagiannidis, “Performance analysis degree in electrical and computer engineering from
of single relay selection in Rayleigh fading,” IEEE Trans. Wireless Queen’s University, ON, Canada, in 2010. From
Commun., vol. 7, pp. 3718–3724, Oct. 2008. 1999 to 2005, he was a Researcher at Korea Elec-
[18] R. Tannious and A. Nosratinia, “Spectrally-efficient relay selection tronics Technology Institute (KETI), Korea, where
with limited feedback,” IEEE J. Sel. Areas Commun., vol. 26, pp. he was involved in many projects related with
1419–1428, Oct. 2008. WPAN systems such as Bluetooth, IEEE802.15.3, IEEE802.11, and HomeRF.
[19] Y. Jing and H. Jafarkhani, “Single and multiple relay selection In 2010, he returned to the Wireless Network Research Center at KETI, where
schemes and their achievable diversity order,” IEEE Trans. Wireless he is currently a Senior Researcher. His research interests are in the areas
Commun., vol. 7, pp. 1414–1423, Mar. 2009. of opportunistic transmissions, cooperative networks, and synchronization in
[20] I. Krikidis, J. S. Thompson, S. McLaughlin, and N. Goertz, “Max- communication.
min relay selection for legacy amplify-and-forward systems with
interference,” IEEE Trans. Wireless Commun., vol. 8, pp. 3016–3027, Il-Min Kim received the B.S. degree in electronics
June 2009. engineering from Yonsei University, Seoul, Korea, in
[21] A. Bletsas, A. G. Dimitriou, and J. N. Sahalos, “Interference-limited 1996, and the M.S. and Ph.D. degrees in electrical
opportunistic relaying with reactive sensing,” IEEE Trans. Wireless engineering from the Korea Advanced Institute of
Commun., vol. 9, pp. 14–20, Jan. 2010. Science and Technology (KAIST), Taejon, Korea, in
[22] K. S. Hwang, Y. C. Ko, and M.-S. Alouini, “Performance bounds for 1998 and 2001, respectively. From October 2001 to
two-way amplify-and-forward relaying based on relay path selection,” August 2002 he was with the Department of Elec-
in Proc. IEEE VTC, Apr. 2009, pp. 1–5. trical Engineering and Computer Science at MIT,
[23] D. Tse, P. Viswanath, and L. Zheng, “Diversity-multiplexing tradeoff Cambridge, USA, and from September 2002 to June
in multiple access channels,” IEEE Trans. Inf. Theory, vol. 50, pp. 2003 he was with the Department of Electrical En-
1859–1874, Sep. 2004. gineering at Harvard University, Cambridge, USA,
[24] D. Tse and P. Viswanath, Fundamentals of Wireless Communications. as a Postdoctoral Research Fellow. In July 2003, he joined the Department
Cambridge, MA: Cambridge University Press, 2005. of Electrical and Computer Engineering at Queen’s University, Kingston,
[25] J. Wang and J. Chen, “Performance of wideband CDMA systems with Canada, where he is currently an Associate Professor. His research interests
complex spreading and imperfect channel estimation,” IEEE J. Sel. include cooperative diversity networks, bidirectional communications, CoMP,
Areas Commun., vol. 19, pp. 152–163, Jan. 2001. femto cells, and green communications. He is currently serving as an Editor
[26] T. S. Rappaport, Wireless Communications: Principle and Practice. for the IEEE T RANSACTIONS ON W IRELESS C OMMUNICATIONS and the
Upper Saddle River, NJ: Prentice Hall, 2002. Journal of Communications and Networks (JCN).