Nam2017 PDF

This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2717048, IEEE Access
S.S. NAM et al.: NEW RESULTS ON ORDERED STATISTICS OVER NAKAGAMI-M FADING 1
New Closed-Form Results on Ordered

Statistics of Partial Sums of Gamma Random
Variables and its Application to Performance
Evaluation in the Presence of Nakagami Fading
Sung Sik Nam, Member, IEEE,, Young-Chai Ko, Senior Member, IEEE and
Mohamed-Slim Alouini, Fellow Member, IEEE
Abstract—Complex wireless transmission systems re- joint statistics). Later, with the advent of complex trans-
quire multi-dimensional joint statistical techniques for mission systems, more complicated multi-dimensional
performance evaluation. Here, we first present the exact joint statistical techniques became necessary [10]–[17].
closed-form results on order statistics of any arbitrary
partial sums of Gamma random variables with the closed- Some previous results have been helpful in the accurate
form results of core functions specialized for independent quantification of performance versus complexity among
and identically distributed Nakagami-m fading channels different transmission design options. Other results, how-
based on a moment generating function-based unified ever, such as the joint distribution functions of linear
analytical framework. These both exact closed-form results functions of ordered random variables (RVs) are not
have never been published in the literature. In addition, as
a feasible application example in which our new offered helpful due to their high complexity. Comprehensive
derived closed-form results can be applied is presented. analysis of how both conventional and new order statis-
In particular, we analyze the outage performance of the tics results help in obtaining the desired statistics of the
finger replacement schemes over Nakagami fading channels received output signal-to-noise ratio (SNR) in wireless
as an application of our method. Note that these analy- transmission systems has not yet been reported.
sis results are directly applicable to several applications,
such as millimeter-wave communication systems in which Recently, [18] and [19] introduced new results to
an antenna diversity scheme operates using an finger determine the joint statistics of partial sums of ordered
replacement schemes-like combining scheme, and other exponential RVs. In [19], a successive conditioning
fading scenarios. Note also that the statistical results can approach was used to convert dependent ordered RVs
provide potential solutions for ordered statistics in any into independent unordered RVs. Obtaining distribution
other research topics based on Gamma distributions or
other advanced wireless communications research topics functions, including the probability density function
in the presence of Nakagami fading. (PDF), the cumulative distribution function (CDF), and
the moment generating function (MGF), is now possi-
Index Terms—Fading channels, outage performance, or-
der statistics, partial sums, Nakagami-m fading. ble with this framework and related results. However,
this approach requires some case-specific manipulations,
which may not always be generalizable. In [18], we
I. I NTRODUCTION introduced a unified analytical framework to determine
Order statistics have played a critical role in the design the joint statistics of partial sums of ordered RVs using
and analysis of many emerging wireless transmission an MGF-based approach. With our proposed approach,
techniques, such as advanced diversity combining, chan- the joint statistics of any arbitrary partial sums of ordered
nel adaptive transmission, and multiuser scheduling [2]– statistics in terms of MGF and PDF, especially in the
[17]. Previous order statistics results in [5]–[9] were presence of Rayleigh fading, can be derived systemati-
obtained based on conventional or slightly modified cally.
statistical theories (e.g., simple one or two dimensional On another front, the Nakagami-m distribution often
gives the best fit to urban [20] and indoor [21] multipath
This is an extended version of a paper that was presented at the In- propagation of wireless transmission. Most importantly,
ternational Conference on Communications in China (ICCC), 2015 [1].
S. S. Nam is in Korea University, Korea (Email: [email protected]). Nakagami fading captures a wide range of multipath
Y.-C. Ko is with Korea University, Korea (Email: [email protected]). channels via the fading parameter, m, including the
M.-S. Alouini is with King Abdullah University of Science and Tech- Rayleigh distribution (m = 1) as a special case [22].
nology (KAUST), Computer, Electrical, and Mathematical Sciences
and Engineering Division (CEMSE), Thuwal, Saudi Arabia (Email: In addition, when m > 1, the Nakagami-m distri-
[email protected]). bution closely approximates the Rice distribution [22]
2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2017.2717048, IEEE Access
by one-to-one mapping between the Rician factor and of the number of Rake fingers, a significant im-
the Nakagami fading parameter. Some analytical results provement is expected because the channel impulse
on Nakagami-m fading assumptions based on order response is completely decayed in a very short time
statistics can be found in [6], [23]–[25]. However, in period compared with the typical RAKE receiver
most cases, fundamental one- or trivial two-dimensional based systems (i.e., carrier frequencies below 10
joint statistical results are provided. These results do not GHz) [27], [28]. Therefore, a larger number of
lend themselves to more sophisticated performance eval- fingers are required while there exist the limited
uation. Thus far, no exact closed-form results, even sim- number of fingers in the mobile unit. This can
plified results, of complicated multi-dimensional joint point to very clear conclusion that it is more neces-
statistics under Nakagami fading conditions are available sary to apply the low complexity and low power
in the literature. The primary goal of this paper is thus consumption finger management schemes with a
to provide new exact closed-form results on the order minimal amount of additional network resources
statistics of any arbitrary partial sums of Gamma random for RAKE reception in the SHO region with mul-
variables, we present a feasible performance evaluation tiple base stations (BSs) to achieve the required
example, in which we apply closed-form results under performance. Here, for mmWave communication
independent and identically distributed (i.i.d.) Nakagami- systems, Nakagami assumption is more proper than
m fading conditions to the MGF-based unified approach Rayleigh assumption because it is not always possi-
in [18]. ble to satisfy Rayleigh criterion [29], [30]. However,
in [26], the author has investigated and analyzed
the performance over i.i.d. Rayleigh fading envi-
A. Main Contributions ronments with multiple BSs based on the statistical
The main contributions and points of difference be- derivation approach used in [19]. In [19], the re-
tween the previous works and this work are briefly quired joint statistics of ordered RVs were obtained
summarized as follows: by applying the conditional PDF based approach
In [18], some closed-form results for Rayleigh proposed. However, this approach is limited to
fading assumptions were provided using a unified when assuming Rayleigh fading from path to path
MGF-based approach. Especially, with the newly and does not allow for similar simplifications for
provided MGF-based unified framework and related Nakagami case. Therefore, we address this mathe-
core functions specialized for Rayleigh fading, the matical issue by providing a general comprehensive
joint statistic closed-form results of any arbitrary analysis framework for outage performance analysis
partial sums of ordered statistics were derived in the presence of Nakagami fading by adopting
systematically. Although [18] provides new useful the MGF-based unified approach in [18] instead
closed-form results on ordered statistics, deriving of [19]. More specifically, we slightly modify
the closed-form results over Nakagami fading chan- the performance analysis framework used in [26]
nel is another challenge. Therefore, in this paper, to make it suitable for these newly derived joint
we provide some new closed-form results of core statistical results.
functions specialized for Nakagami fading and then Note that the slightly modified analytical framework
with these results, some exact closed-form results suitable for the derived statistical results can also be
on ordered statistics of partial sums of Gamma ran- configured to be directly applicable to other various
dom variables are newly provided. These both exact fading scenarios while the analytical framework in [26],
closed-form results have never been published in the the conditional PDF based approach, and related results
literature and may stimulate researchers to find new were limited only to i.i.d. Rayleigh fading assumptions.
results in the general order statistics theory. Note also that our derived statistical results are much
As a feasible application example in which our simpler than the original multiple-fold integral forms
derived joint statistic closed-form results can be ap- based on the conventional MGF based approaches.
plied, we consider the outage performance analysis
of the finger replacement schemes (FRS) proposed
II. S YSTEM M ODELS AND S TATISTICAL A NALYSIS
in [26] by extending channel model to Nakagami-
OF THE O UTAGE P ERFORMANCE OF A PPLICATION
m fading channels. It is very noticeable that the
E XAMPLE
FRS in [26] can also apply to the new “trendy”
applications such as millimeter-wave (mmWave) Here, we consider the full scanning method in [26]
communication systems in which an antenna diver- in the presence of Nakagami fading. Using the system
sity scheme operates using an FRS-like combining model assumptions in [26], we assume that L base
scheme. In mmWave systems, with an increase stations (BSs) are active and that there are a total of N(L)
L
resolvable paths which is defined as N(L) =
P
Nn SNR falls below the target SNR (i.e., 0 < x < γT ) and
n=1 ii) when the combined SNR is greater than or equal to
where Nn is the number of resolvable paths from the n- the target SNR, γT , (i.e., x ≥ γT ), the outage probability
th BS. [26] assumed that in the soft handover (SHO) can be rewritten as
region, for RAKE reception, only Nc out of N(n)
FγF (x)
(1 ≤ n ≤ L) paths are used. Without loss of generality,
Pr [Y + max {W1 , W2 , · · · , WL } < x] , 0 < x < γT ;
8
N1 is defined as the number of resolvable paths from the >
Pr [γT ≤ Y + W1 < x]
<
serving BS while N2 , N3 , · · · , NL are defined as those =
: + Pr [Y + W1 < γT , γT ≤ Y
>
from the target BSs. In the SHO region, the receiver + max {W1 , W2 , · · · , WL } < x] , x ≥ γT .
is assumed at first to rely only on N1 resolvable paths (1)
and, as such, to start with Nc /N1 -generalized selection Here, Y and Wn for n = 1 are correlated while Y and
combining (GSC) [5] which combines the strongest Nc Wn ( n = 2, · · · , L ) are independent. Thus, by adopting
resolvable paths among the N1 available ones. These the proposed mathematical approach in [18] instead of
schemes are based on the comparison of blocks consist- applying [19], we can evaluate key statistics in (1) as
ing of Ns (< Nc < Nn ) paths from each BS. Pr [γT ≤ Y + W1 < x] = FY +W1 (x) − FY +W1 (γT ) , (2)
If we let Y be the sum of the Nc − Ns strongest paths Pr [Y + max {W1 , W2 , · · · , WL } < x]
NcP−Ns
from the serving BS, Y = γi:N1 , and Wn be the Z x Z x−y YL
(3)
i=1 = fY,W1 (y, w1 ) FWn (x − y)dw1 dy,
sum of the Ns smallest paths from the serving BS for 0 0 n=2
n = 1 and be the sum of the Ns strongest paths from the
Nc
P and
target BS for n = 2, · · · , L, Wn = γi:Nn for Pr [Y +W1 < γT , γT ≤ Y +max {W1 , W2 , · · · , WL } < x]
i=Nc −Ns +1
Ns Z γT Z γT −y L
Y
P
n = 1 and Wn = γi:Nn for n = 2, · · · , L, then after = fY,W1 (y, w1 ) FWn (x − y)dw1 dy.
i=1 0 0 n=2
GSC, the received output SNR becomes Y +W1 , where (4)
γi:Nn (i = 1, 2, · · · , Nn ) is the i-th order statistics out of
Nn SNRs of paths from the n-th BS by arranging Nn III. J OINT S TATISTICS OF PARTIAL S UMS OF
nonnegative i.i.d. RVs, {γj }N j=1 , where γj is the SNR of
n
O RDERED R ANDOM VARIABLE OVER I.I.D.
the j-th path from the n-th BS, such that γi:N1 ≥ γi:N2 ≥ NAKAGAMI -m FADING
· · · ≥ γi:Nn . Based on [26], the receiver compares the
A. Main Approach
output SNR, Y +W1 , with a certain target SNR at the
beginning of every time slot. Then, if the sum of the For the Nakagami-m fading case, the instantaneous
Nc − Ns strongest paths from the serving BS and the SNR, γ, has the PDF given by [2, eq. (2.55)]
Ns smallest paths from the serving BS, Y +W1 is greater „ «m m−1
m γ
„
m
«
than or equal to the target SNR, a one-way SHO is used p (γ)= exp − γ , γ ≥ 0, (5)
γ̄ Γ(m) γ̄
and no finger replacement is needed. On the other hand
(i.e., Y +W1 falls below the target SNR), the receiver where Γ(·) denotes the gamma function [31, eq.
attempts a two-way SHO by starting to scan additional (8.310.1)] and γ̄ is the common average faded SNR.
paths from other target BSs. Note that the major difficulty lies in deriving the required
joint statistics of ordered RVs. In [26], by applying the
To show the validity of our derivations, we con- conditional PDF based approach proposed in [19], the
sider outage performance. We modify the mathematical required joint statistics were obtained, especially with
analysis framework in [26] to make it suitable for our an assumption of i.i.d. Rayleigh fading. However, our
newly derived joint statistical results. This framework concern is Nakagami-m fading which includes a wide
to determine outage performance is configured to be range of multipath channels via the fading parameter,
directly applicable to other fading scenarios with the help m, [22]. In this setting, we cannot directly adopt the
of the unified MGF-based approach in [18] rather than proposed method in [19]. Hence, we borrow the con-
the approaches in [26] and [19]. Based on the mode of cept of the unified MGF-based systematical framework
operation in Sec. [26, II-B], an overall outage probability proposed in [18].
is declared when the final combined SNR, γF , falls
below a predetermined threshold, x. Based on it, we can
define the outage probability as FγF (x) = Pr [γF < x], B. Common Core Functions and Relations
where γF = Y + W1 for Y + W1 ≥ γT and γF = For mathematical tractability, let us consider integer-
Y +max {W1 , W2 , · · · , WL} for Y +W1 < γT . Then, by order fading parameters (i.e., m takes positive integer
separately considering two cases i) when the combined values). Even with integer fading parameter values,
„ «n·m „ n
«−n·m X ! „ „ « «
m m n m
[c (γ, −si )]n = + si (−1)k exp − + si k · γ
γ̄ γ̄ k=0
k γ̄
! N (m)
!
X N (m) „ m «N (m)−j j (6)
X k γ N (m)
× si ,
n1 , n2 , · · · , nm m−1
Q nl+1 j=0 j γ̄
n1 ,n2 ,··· ,nm ≥0
n1 +n2 +···+nm =k
(l!)
l=0
„ «n·m „ «−n·m „ „ « «
m m m
[e (γ, −si )]n = + si exp − + si n · γ
γ̄ γ̄ γ̄
N (m)
! !
X n γ N (m) X N (m) „ m «N (m)−j j (7)
× si ,
n1 , n2 , · · · , nm m−1 j γ̄
(l!)nl+1 j=0
Q
n1 ,n2 ,··· ,nm ≥0
n1 +n2 +···+nm =n
l=0
and
„ «n·m „ «−n·m „ „ « n
«X !
nm m m n
[µ (γa , γb , −si )] = + si exp − + si γ a · n (−1)n−h
γ̄ γ̄ γ̄ h
h=0
! !
X X h n−h
×
n1 , n1 , · · · , nm n0 1 , n0 2 , · · · , n0 m
n1 ,n2 ,··· ,nm ≥0 n0 1 ,n0 2 ,··· ,n0 m ≥0 (8)
n1 +n2 +···+nm =h n0 +n0 +···+n0 =n−h
1 2 m
N (m) M (m)
! !
γa N (m) γb M (m) X X N (m) M (m) „ m «N (m)+M (m)−j−q j+q
× m−1 · m−1 si ,
0 j q γ̄
(l!)nl+1 (k!)n k+1 j=0 q=0
Q Q
l=0 k=0
Pm−1 Pm−1
where N (m) = l=0 l · nl+1 and M (m) = k=0 k · mk+1 .
closed-form results of partial sums of ordered RV over In the special case of the Rayleigh fading channel
Nakagami-m fading remain an open problem. Many (m = 1), the results are given in [18]. With (6)-(8) and
previous studies [32]–[39] focused on performance anal- the unified framework for Rayleigh fading assumptions
ysis over Nakagami fading channels with the integer in [18], we can obtain the generic MGF expressions in
fading parameter. These works showed that the integer a compact form as well as the desired PDF expressions
fading parameter is sufficient to model a wide range of through an inverse LT (see Appendix B).
fading conditions and can cover most cases of interest in In what follows, we show how our results can be
n
practice (e.g., for many practical channels, 1 ≤ m ≤ 15, greatly simplified. Let Z1 =
P
γi:N and Z2 =
[32]). i=1
N
Here, we first observe three common core functions
P
γi:N for example. Then, the original second-order
of i.i.d. Nakagami distributions:
R γ i) a mixture of a CDF
i=n+1
and an MGF, c (γ, λ) = 0 dx p (x) exp (λx), ii) a MGF expression of Z = [Z1 , Z2 ] can be written as an
mixture of an exceedance distribution function (EDF) N -fold integral expression
R∞
and an MGF, e (γ, λ) = γ dx p (x) exp (λx), and M GFZ (λ1 , λ2 ) = E {exp (λ1 Z1 + λ2 Z2 )}
Rγ
iii) an interval MGF, µ (γ, λ) = γab dx p (x) exp (λx), N!
Z∞
where γ is real and λ can be complex [18, Sec. III- = dγ1:N p (γ1:N ) exp (λ1 γ1:N )
(N − n)! (n − 1)!
A]. We further consider the n-th power of these com- 0
n γn−1:N
mon core functions for arbitrary n, such as [c (γ, λ)] , Z
n n
[e (γ, λ)] , and [µ (γ, λ)] . The closed-form results of ··· dγn:N p (γn:N ) exp (λ1 γn:N )
these functions will play a important role to simplify the 0 (9)
γZn:N
derivation of joint MGFs in later sections.
× dγn+1:N p (γn+1:N ) exp (λ2 γn+1:N )
As shown in Appendix A, each function can be
0
expressed in a finite summation form, enabling us to γK−1:N
apply an inverse Laplace transform (LT) with the MGF
Z
··· dγN :N p (γN :N ) exp (λ2 γN :N ).
expressions in deriving the closed-form expressions of
0
the final PDF. The resulting n-th power of common core
functions are as shown in top of this page. Following from (9) and simplifying the N -fold integral
expression with the help of the interchange of multiple as

integrals technique and simplified results given in [18, ( „ «−(n−1)·m „ „ « «)
m j0 m
eqs. (10) and (12)], the 2-dimensional joint PDF of Z = L−1
s1 s1 +s1 exp − +s1 ·n·γn:N
γ̄ γ̄
[Z1 , Z2 ] can be expressed, more specifically by applying 0
(z1 −n·γn:N)(n−1)·m−j −1
„ «
the PDF of the RV of interest, specifically (5), into the m
= exp − ·n·γn:N
simplified form [18, eq. (25)], as γ̄ ((n − 1) m)!
„ «
m
× 1 F̃1 (n − 1) m, (n − 1) m − j 0 , − (z1 − n · γn:N )
fZ (z1 , z2 ) = L−1
S1 ,S2 {M GFZ (−S1 , −S2 )}
γ̄
N! × U (z1 − n · γn:N) .
= (14)
(N − n)! (n − 1)!
„ «m Similarly, with (6), the second inverse LT term in (10)
γn:N m−1
Z ∞ „ «
m m (10)
× dγn:N exp − · γn:N can also be written as
0 γ̄ Γ (m) γ̄ n o
× Ls1 exp(−s1 γn:N )[e (γn:N , −s1)]n−1
−1 ˘ Ls2 −1 [c (γn:N , −s2 )]N −n
¯
n o N −n
× L−1 [c (γn:N , −s2 )]N −n . X X “N − n”“ k ”
s2 = (−1)k
k=0
k n1 , n2 , · · · , nm
n1 ,n2 ,··· ,nm ≥0
n1 +n2 +···+nm =k
Then, adapting (7) to (10) yields the first inverse LT term „«(N −n)·m NX (m) “ „ «
m N (m)” m N (m)−j
as ×
γ̄ j=0
j γ̄
( «)
L−1 exp (−s1 γn:N ) [e (γn:N , −s1 )]n−1
˘ ¯ „ «−(N−n)·m „ „ «
s1 m m
!„ « ×Ls2−1 s2 j +s2 exp − +s2 ·k·γn:N ,
X n−1 m
(n−1)·m γ̄ γ̄
= (15)
n0 1 , n0 2 , · · · , n0 m γ̄
0 0 0
n 1 ,n 2 ,··· ,n m ≥0 and in (15) the inverse LT term, Ls2−1 {·}, can be
n0 1 +n0 2 +···+n0 m =n−1
0 obtained as
γn:N N (m)
„ «
m ( „ «−(N −n)·m „ „ « «)
× exp − · γn:N m−1 j m m
γ̄ Q 0 n0 0 L−1
s2 s2 + s2 exp − + s2 · k · γn:N
(l !) l +1 γ̄ γ̄
l0 =0 8 “ ”
(z2 −k·γn:N )(N−n)·m−1
N 0 (m)
!„ « 0
N (m)−j 0
>
> exp − m γ̄
·k·γn:N ((N −n)·
X N 0 (m) m
>
>
> “ ” m−1)!
× >
> × exp − m
> ·(z 2 − k·γ n:N ) U (z2 − k·γn:N ) ,
j0 γ̄ >
>
> γ̄
j 0 =0 < for“j = 0
>
( «) ”h j
„ «−(n−1)·m „ „ « = exp − m · k·γ d g(z2 −k·γn:N )
−1 j0 m m > γ̄ n:N dz2 j
× Ls1 s1 +s1 exp − +s1 ·n·γn:N , >
> #
γ̄ γ̄ >
>
> j−1
P (k 0
) ( j−k 0
−1)
>
> + g (0) δ (z2 − k·γn:N ) U (z2 − k·γn:N ) ,
(11) >
>
> k0 =0
P m−1 >
:
where N 0 (m) = `0 =0 `0 · n0 `0 +1 and, with the help of for j > 0
(38) in Appendix B, (16)
where
t(N −n)·m−1 exp (−at)
g(t) = , (17)
( „ «−(n−1)·m „ „ « «)
−1 j 0 m m ((N − n) · m − 1)!
Ls1 s1 +s1 exp − +s1 ·n·γn:N
γ̄ γ̄
or equivalently for n > m, the inverse LT term in (15)
γn:N )(n−1)·m−1
8 “ ”
exp − m γ̄ ·n·γn:N (z1 −n·((n−1)· can be obtained as
” m)!
>
>
> ( «)
> “ „ «−(N −n)·m „ „ «
> × exp − m ·(z1 − n·γn:N ) U (z1 − n·γn:N ) ,
>
m m
j
L−1
> γ̄
>
> s2 s2 + s2 exp − + s2 · k · γn:N
> for j 0 = 0 γ̄ γ̄
>
>
>
> “ ” » j0 . „ «
< m d g(z1 −n·γn:N ) m
= exp − γ̄ ·n·γn:N dz1 j
0 = exp − · k · γn:N (z2 − k · γn:N)(N −n)·m−j−1
> # γ̄
>
j 0 −1 „ «
m
> P (k 0 ) ( j 0 −k0 −1)
>
+ g (0) δ (z1 − n·γn:N ) × 1 F̃1 (N − n) m, (N − n) m − j, − (z2 − k · γn:N)
>
>
γ̄
>
>
>
> k0 =0
× U (z2 − k · γn:N) .
>
: ×U (z0 1 − n·γn:N ) ,
>
>
(18)
>
for j > 0
(12)
where IV. C LOSED -F ORM E XPRESSIONS FOR K EY J OINT
t(n−1)·m−1 exp (−at) S TATISTICS FOR FRS
g(t) = , (13)
((n − 1) · m − 1)! In this section, we investigate the following key joint
statistics for outage performance evaluation: fY,W1 (·, ·),
or equivalently (12) can be also simplified when n > m FY +W1 (·), and FWn (·) for 2 ≤ n ≤ L in Sec. II.
α X t1 “ ” „ «t1 −α−1
A. Two-dimensional Joint PDF of Two Adjacent Partial X t1 (−1)t1 −t2 α! m
+ (Ns − 1)
Sums of Ordered RVs, fY,W1 (x, y) t1 =0 t2 =0
t2 t1 ! γ̄
t t1 −t2 „ «
In this case, the target 2-dimensional joint PDF of Y = y (h + 1)
2 m Ns − 1
× exp − · y
NcP
−Ns Nc
P (Ns − (h + 1))t1 γ̄ Ns − (h + 1)
γi:N1 and W1 = γi:N1 can be obtained  „
m
«ff
(Ns −1) (h+1) −β−t1 +t2 −1
i=1 i=Nc−Ns+1 × k−
Nc−Ns−1 γ̄ Ns −(h+1)
P 2
with the 4-dimensional joint PDF of Z1 = γi:N1 ,  „ «ff
i=1 x y
× 4 1−U −
c−1
NP Nc −Ns Ns − (h + 1)
Z2 = γNc −Ns :N1 , Z3 = γi:N1 , and Z4 = γNc :N1 , „ „ « «
i=Nc−Ns+1 m (Ns − 1) (h + 1)
× γ β + t1 − t2 + 1, k− a
where the order statistics of N1 resolvable paths can be γ̄ Ns − (h + 1)
3
viewed as „
m 1
„
(Ns − 1) (h + 1)
« «
Y −γ β +t1 −t2 +1, · k− y 5,
z }| { γ̄ Ns Ns − (h + 1)
γ1:N1 , · · · , γNc −Ns −1:N1 , γNc −Ns :N1 ,
| {z } | {z } (24)
Z1 Z2
where γ (·, ·) is the lower incomplete gamma function
W
z }|1 { [40, eq. (8.352.1)].
γNc −Ns +1:N1 , · · · , γNc −1:N1 , γNc :N1 , γNc +1:N1 , · · · , γN1 :N1 .
| {z } | {z }
Z3 Z4
(19) B. One-dimensional CDF of the Nc /N1 -GSC Output
In (19), Z1 , Z2 , Z3 , and Z4 have the following con- SNR, FY +W1 (x)
ditions, such that Z4 < Z2 , Z1 > (Nc − Ns − 1)Z2 For convenience, we let Z 0 =Y +W1 . Then, the target
Nc
and (Ns − 1) Z4 < Z3 < (Ns − 1) Z2 . Based on these P
CDF of Z 0 = γi:N1 with the 2-dimensional joint PDF
conditions and with the help of a function of a marginal i=1
NP
c −1
PDF, the joint PDF of Y and W1 , fY,W1 , can be obtained
of Z1 = γi:N1 and Z2 = γNc :N1 can be obtained as
by integrating out z2 and z4 as i=1
Z x Z z
fY,W1 (x, y) Nc
Z yZ x
FY +W1 (x) = fZ1 ,Z2 (z − z2 , z2 ) dz2 dz. (25)
Ns Nc −Ns (20) 0 0
= fZ1 ,Z2 ,Z3 ,Z4(x−z2 , z2 , y−z4 , z4) dz2 dz4 .
0 y
Ns
Here, by applying a similar approach used in (11) and
(21), and adopting the generic form in [18, eq. (44)]
Here, by adopting the unified MGF approach proposed with (5) and the related common core functions given in
in [18], we obtain the 4-dimensional joint PDF in (20), (6) and (7), we can obtain the target 2-dimensional joint
fZ1 ,Z2 ,Z3,Z4 (z1 , z2 , z3 , z4 ), for i.i.d. Nakagami-m fad- PDF, fZ1 ,Z2 (z1 , z2 ), in (26) at the end of page for the
ing assumption after applying (6), (7), (8), and (5) to i.i.d. Nakagami-m fading assumption.
the generic 2-dimensional PDF form in [18, eq. (42)] as Substituting (26) into (25) and then, with the help
given in (21) at the end of page. Then, we substitute of [31, eqs. (8.352.6), (3.381.1)] and then after some
(21) into (20) and then, after re-arranging and some re-arranging and some mathematical simplification, (25)
simplification, (20) can be expressed as shown in (22). can be expressed as provided in (27).
Following the detailed derivations in Appendix C, we
can obtain the closed-form expressions of the double C. One-dimensional CDF of the Sums of the Ns
integral term in (22) as follows, Strongest Paths from Each Target BS, FWn (x)
i) For h = Ns − 1, In this case, by applying a function of a marginal PDF
„ «−α−1 „ «−β−1 „ « s−1
NP
m m m k with the 2-dimensional joint PDF of Z10 = γi:Nn and
(Ns −1) k γ β +1, · y
γ̄ γ̄ γ̄ Ns i=1
2 3
„ « „ « (23) Z20 = γNs :Nn , the target one-dimensional CDF of Wn =
m Ns −1 m Ns −1 5 Ns
× γ α+1, ·
4 x −γ α+1, · y . P
γ̄ Nc −Ns γ̄ Ns γi:Nn can be derived as
i=1
ii) For 0 ≤ h ≤ Ns − 2, Z x Z z
Ns
fZ10 ,Z20 z − z20 , z20 dz20 dz.
` ´
2 FWn (x) = (28)
„ «−α−1 „ «  „ «ff 0 0
m m −β−14 x y
(Ns −1) k 1−U − The closed-form expression of (28) can be easily ob-
γ̄ γ̄ Nc −Ns Ns −(h+1)
„ « „ « ff tained by replacing Nc and N1 with Ns and Nn in
m m Ns −1
× γ β + 1, ka γ α+1, · x −1 (27), respectively. The closed-form result of the integral
γ̄ γ̄ Nc −Ns
„ « „ « ff
3 form in (22) can be obtained by separately considering
m k m Ns − 1 i) h = Ns −1 and ii) 0 ≤ h ≤ Ns −2 as shown in (23) and
−γ β + 1, · y γ α + 1, · y −1 5
γ̄ Ns γ̄ Ns (24), respectively.
V. D ISCUSSIONS AND C ONCLUSIONS funded by the Korea government(MSIP) (No. 2014-

In this work, we provided new exact closed-form order 00552, Next Generation WLAN System with High Effi-
cient Performance).
statistics of partial sums of Gamma random variables
by deriving the closed-form results of common core
functions specialized for Nakagami fading. In addition, A PPENDICES
we analyzed the outage performance of FRS proposed A PPENDIX A
in [26] operating over Nakagami-m fading channels as D ERIVATION OF C LOSED -F ORM E XPRESSIONS OF
a feasible application example. T HREE C OMMON C ORE F UNCTIONS
If the closed-form results of the joint statistics, espe-
In Sec. III, with (5), we can write the mixture of
cially in order statistics, are not available, the numerical
a CDF and an MGF for the i.i.d. Nakagami fading
estimation of multi-fold integral expressions (e.g., the
assumption as
N -fold integrals given in (9)) are required. However,
estimating them accurately as N increases is difficult „ «m Z γ „ „ « «
even with conventional mathematical tools. When the m 1 m−1 m
c (γ, −si ) = γ exp − +si x dx.
N is large, estimating the analytical results is almost γ̄ Γ(m) 0 γ̄
(29)
impossible. However, with closed-form results derived
Then, with the help of [31, eq. (3.381.1) and eq.
here, probabilistic analysis is numerically possible with
(8.352.1)], the closed-form result of (29) can be obtained
conventional mathematical tools. Note that the closed-
as
form expressions seems to be complicated and they
can be further summarized in a functionalized shape. c (γ, −si )
„ «m „ «−m „ „ « «
However, to demonstrate how we obtained the derived m 1 m m
= + si γ m, + si γ
results and the feasibility of applying the derived results γ̄ Γ (m) γ̄ γ̄
„ «m „ «−m
to an application, we maintain them in a minimally or m m
= + si
simplified form. With these results, the user can directly γ̄ γ̄
2 h“ ” il 3
change/apply the obtained results in the form desired 6
„ „
m
« « m−1
X
m
γ̄
+ si γ
7
by the user. The closed-form expressions may appear to × 41 − exp − + si γ 5.
γ̄ l=0
l!
be complicated, but the numerical results can be easily
(30)
obtained. On the other hand, it is almost impossible
With the summand expression in (30), by applying
to obtain numerical results physically with conventional
the binomial theorem, the n-th power of c (γ, −si ) for
mathematical tools due to estimation difficulties.
arbitrary n can be obtained as
Further, as a validation of our analytical formula for
the outage probability, in Fig. 1, we cross-verified the [c (γ,−si )]n
„ «nm„ «−nm
analytical results and the simulation results obtained via =
m m
+si
Monte-Carlo simulation. Fig. 1 showed that the derived γ̄ γ̄
2 “ ”l 3k
analytical results match the simulation results. As a n “ ” „ „ « « m−1 m
X γ̄ +si γ l 7
X n m
(−1)k exp −
6
result, we believe that we can accurately predict the × +si kγ 4 5 .
k γ̄ l!
performance with them. k=0 l=0
Note that closed-form results on ordered statistics of (31)

partial sums of ordered random variables over Nakagami Here, with the help of the multinomial theorem, we
fading remained in an open problem, even with integer obtain the following relationship:
fading parameter values. Note also that although derived 2 “ ”l 3k
m−1 m +si γl
closed-form results limited to integer fading parameter 6X γ̄ 7
4 5
values, they can still covers most cases of interest l=0
l!
in practice. Therefore, in the view of contribution to X “ k ”
ordered statistics, these new statistical results can provide = M0n1 M1n2 · · · Mm−1
nm
,
n1 ,n1 ,··· ,nm ≥0
n1 , n1 , · · · , nm
the potential solution of both other ordered statistics in n1 +n1 +···+nm =k
the presence of Nakagami fading in advanced wireless h il (32)
m
communications research and any other research topics where Ml = γ̄ +si γ /l! and some mathematical
based on Gamma distributions. manipulations give us
N(m)
!
γ N(m) X N (m) „m«N(m)−j j
ACKNOWLEDGMENT M0n1 M1n2 nm
· · · Mm−1 = m−1 si ,
j γ̄
(l!)nl+1
Q
j=0
This work was supported by Institute for Information l=0
& communications Technology Promotion(IITP) grant (33)
Pm−1
where N (m) = l=0 l · nl+1 . Thus, after successive and
substitution from (33) to (31), we can get (6). exp(−a·t) `
g 000 (t)= (−1)3 a3 ·tn−1 − 3a2 (n−1) tn−2
Similarly, with the help of [31, eq. (3.381.3) and (n−1)!
´
eq. (8.352.2)] and then by applying the binomial the- +3a (n−1)(n−2)tn−3 −(n−1) (n−2)(n−3) tn−4 .
orem, the closed-form expressions of the n-th power (41)
of e(γ, −si ) and µ(γa , γb , −si ) for arbitrary n can be As a result, from (39)-(41), we can now obtain the m-th
obtained as shown in (7) and (8), respectively. derivative of g (t) for arbitrary m as
Qk
Xm “m” (n−l) n−1−k
g (m)(t)= (−1)k+m am−k l=0 t exp (−at) .
A PPENDIX B k=0 k n!
(42)
I NVERSE LT PAIR AND R ELATED U SEFUL F UNCTION
The following inverse LT is useful for the derivation
A PPENDIX C
of final PDF closed-form expressions from MGF expres-
D ERIVATION OF THE C LOSED - FORM E XPRESSION OF
sions in Sec. III
(22)
sm
 ff
L−1
s exp (−b (a + s)) . (34) To obtain the closed-form expression of (22), we need
(a + s)n
m
to calculate the following double integral term
s −1
Here, let F (s) = (a+ s)n , then, Ls {F (s)} = f(t) and we
Z yZ
Ns
x
Nc−Ns
„
m
« „
m
«
can obtain the following inverse LT pair for b > 0: z2α z4β exp − (Ns −1) z2 exp − kz4 U (z2 −z4)
0 y γ̄ γ̄
Ns
L.T
L−1
s {F (s) exp (−b (a + s))} ←→ exp (−ba) f (t − b) U (t − b) . ×U (x−(Nc −Ns) z2)U (y−((h+1) z4 +(Ns −(h+1)) z2)) dz2 dz4 .
(35) (43)
1 −1
In (35), let G(s) = (a+s) n and Ls {G(s)} = g(t), then Based on the given conditions associated with pa-
F (s) = sm G (s) and we can obtain the inverse LT rameters z2 and z4 , we need to consider two different
pair of F (s) by applying classical inverse LT pairs and shaded regions shown in Fig. 2. More specifically, the
properties as overall intersection region depends on the intersection
h+1 y
m−1 point between z2 = − Ns −(h+1) · z4 + Ns −(h+1) (for
L.T dm g (t) X (k)
F (s) = smG(s) ←→ f (t)= m
+ g (0) δ (m−k−1) (t), 0 ≤ h ≤ Ns − 1) and the z2 -axis. For case i),
dt k=0 the z2 -coordinate term for the intersection point of
(36) h+1 y
n−1
exp(−at) z2 = − Ns −(h+1) · z4 + Ns −(h+1) and z2 -axis (i.e.,
where g (t)= t (n−1)! . Therefore, the inverse LT pair y
in (34) can be obtained as cz2 = Ns −(h+1) , where cz2 represents the z2 -axis value
m
of the intersection.) is located between Nys and Nc −N x
.
L−1
s {s G (s) exp (−b (a + s))}
s
" # Therefore, the intersection becomes the shaded region

m−1 h+1
L.T dm g (t−b) X (k) (m−k−1) filled with a blue line under z2 = − Ns −(h+1) · z4 +
←→ exp(−ba) + g (0)δ (t−b) U (t−b) . y
dtm k=0 . For case ii), the z2 -coordinate term for the
Ns −(h+1)
(37) intersection point is located over Nc −N x
on the z -axis.
Pm−1 2
Here, for m > 0, k=0 g(k)(0)δ (m−k−1) (t−b)U (t−b) ≈ 0. s
Therefore, the intersection region becomes the shaded
Thus, (37) can be finally simplified as region filled with a red line under both z2 = Nc −N x
s
m h+1 y
L−1
s {s G (s) exp (−b (a + s))} and z2 = − Ns −(h+1) · z4 + Ns −(h+1) . Specifically,
L.T dm g (t − b) (38) for z4 < ez4 (where ez4 represents the z4 -axis value
←→ exp (−ba) U (t − b) . x
dtm of the intersection point between z2 = Nc −N and
s
With (38), we still need to derive the m-th derivative of h+1 y y
z2 = − Ns −(h+1) · z4 + Ns −(h+1) and ez4 = (h+1) −
exp(−a·t)tn−1
g (t). In this paper, we assume that g(t) = Ns −(h+1)
(n−1)! (Nc −Ns ) x), the intersection becomes the shaded region
and we derive the derivative of g (t) for a special case under z2 = Nc −N x
. Otherwise, the intersection becomes
s
and then we can extend this result to the general case. h+1
the shaded region under z2 = − Ns −(h+1) y
·z4 + Ns −(h+1) .
More specifically, after i) differentiating g(t) based on Note that these two cases are dominated by the relation-
the product rule one time, two times, and three times ship among parameters (Nc and Ns ).
and then ii) rearranging and simplifying them, the first, Based on the above observations, the valid integration
second, and third derivative of g (t) can be written, for case i) is 0 ≤ z4 ≤ Nys and Nys ≤ z2 ≤ − Ns −(h+1)
h+1
·
respectively, as y
z4 + Ns −(h+1) . For case ii), we need to consider the
exp(−a·t) following two cases: a) h = Ns − 1 and b) 0 ≤ h ≤
g0 (t) = (−1) a·tn−1 − (n−1) tn−2 ,
` ´
(39)
(n−1)! Ns − 2. As a result, for case ii)-a), the valid integration
exp(−a·t) ` is 0 ≤ z4 ≤ Nys and Nys ≤ z2 ≤ Nc −N x
. Otherwise,
g 00 (t) = (−1)2 a2 ·tn−1 − 2a (n − 1) tn−2 s
(n − 1)! (40) based on the above observations, we need to consider
´
+ (n − 1) (n − 2) tn−3 , two regions separately. More specifically, for the shaded
x
region under z2 = Nc −N s
, the valid integration region [6] Y. Ma and C. C. Chai, “Unified error probability analysis for
is 0 ≤ z4 ≤ ez4 and Nys ≤ z2 ≤ Nc −N x
. Otherwise, generalized slection combining in nakagami fading channels,”
s IEEE J. Select. Areas Commun., vol. 18, no. 11, pp. 2198–2210,
y
the valid integration is ez4 ≤ z4 ≤ Ns and Nys ≤ z2 ≤ Nov. 2000.
h+1 y
− Ns −(h+1) ·z4 + Ns −(h+1) . [7] M. Z. Win and J. H. Winters, “Virtual branch analysis of sysmbol
As a result, we can rewrite (43) as error probability for hybrid selection/maximal-ratio combining
Rayleigh fading,” IEEE Trans. Commun., vol. 49, no. 11, pp.
a) For h = Ns − 1, 1926–1934, Nov. 2001.
Z y Z x „ « „ « [8] A. Annamalai and C. Tellambura, “Analysis of hybrid
Ns Nc−Ns m m selection/maximal-ratio diversity combiners with Gaussian er-
z2αz4β exp − (Ns −1) z2 exp − kz4 dz2dz4 . rors,” IEEE Trans. Wireless Commun., vol. 1, no. 3, pp. 498–511,
0 y γ̄ γ̄
Ns July 2002.
(44) [9] R. K. Mallik, D. Gupta, and Q. T. Zhang, “Minimum selection
GSC in independent Rayleigh fading,” IEEE Trans. Veh. Technol.,
b) For 0 ≤ h < Ns − 1, vol. 54, no. 3, pp. 1013–1021, May 2005.
„ « [10] Y. C. Ko, H. C. Yang, and M.-S. Alouini, “Adaptive modulation
x y and diversity combining based on output-threshold MRC,” IEEE
U − Trans. Wireless Commun., vol. 6, no. 10, pp. 3727–3737, Oct.
Nc −Ns Ns −(h+1)
h+1 2007.
− ·z
Z y Z Ns−(h+1) 4 „ « „ « [11] S. Choi, M.-S. Alouini, K. A. Qaraqe, and H.-C. Yang, “Finger
Ns +
y m m
× z2αz4β exp − (Ns −1)z2 exp − kz4 dz2dz4
Ns−(h+1) replacement method for RAKE receivers in the soft handover
0 y γ̄ γ̄ region,” IEEE Trans. Wireless Commun., vol. 7, no. 4, pp. 1152–
Ns
» „ «– 1156, Apr. 2008.
x y
+ 1−U − [12] P. Lu, H.-C. Yang, and Y.-C. Ko, “Sum-rate analysis of MIMO
Nc −Ns Ns −(h+1)
(Z Z x „ « „ « broadcast channel with random unitary beamforming,” in Proc.
ez4 N −N m m
c s α β of IEEE Wireless Commun. and Networking Conf. (WCNC’08),
× z2 z4 exp − (Ns −1) z2 exp − kz4 dz2dz4 Las Vegas, Nevada, Mar. 2008.
0 y γ̄ γ̄
Ns
h+1 9 [13] Z. Bouida, N. Belhaj, M.-S. Alouini, and K. A. Qaraqe, “Mini-
− z4
Z y Z Ns−(hy+1) „ « „ « >
= mum selection GSC with down-link power control,” IEEE Trans.
Ns +
Ns−(h+1) z αz β exp −
m m Wireless Commun., vol. 7, no. 7, pp. 2492–2501, July 2008.
+ 2 4 (Ns −1)z2 exp − kz4 dz2dz4 .
ez4 N y γ̄ γ̄ >
; [14] H. C. Yang and M. S. Alouini, Order Statistics in Wireless
s
Communications, 1st ed. New York, NY: Cambridge University
(45) Press, 2011.
With (44) and (45), we need to determine four double- [15] G. Chen, Y. Gong, and J. Chambers, “Study of relay selection in
integral terms over z2 and z4 . For the double-integral a multi-cell cognitive network,” IEEE Wireless Commun. Lett.,
vol. 2, no. 4, pp. 435 – 438, Aug. 2013.
term in (44) and the second double-integral term in (45),
[16] D. B. Smith and D. Miniutti, “Cooperative selection combining
we can simply obtain the closed-form expression by in body area networks: switching rates in gamma fading,” IEEE
simply adopting [31, eq. (3.381.1)] for each integration Wireless Commun. Lett., vol. 1, no. 4, pp. 284–287, Aug. 2012.
over z2 and z4 separately as shown in (23) and (24). [17] S. S. Nam, M. S. Alouini, and M. O. Hasna, “Joint statistics
of partial sums of ordered exponential variates and performance
For the first and third double-integral terms in (45), of GSC RAKE receivers over Rayleigh fading channel,” IEEE
we can obtain the closed-form expression of the inner Trans. Commun., vol. 59, no. 8, pp. 2241–2253, Aug. 2011.
integral term as the function of the exponential and the [18] S. S. Nam, M.-S. Alouini, and H.-C. Yang, “A MGF-based unified
framework to determine the joint statistics of partial sums of
incomplete Gamma function, γ (·, ·) [31, eq. (3.381.1)]. ordered random variables,” IEEE Trans. Inform. Theory, vol. 56,
By rearranging the incomplete Gamma function as the no. 8, pp. 5655–5672, Nov. 2010.
summation form with the help of [31, eq. (8.352.6)] [19] H.-C. Yang, “New results on ordered statistics and analysis
and then applying [31, eq. (3.381.1)], we can obtain the of minimum-selection generalized selection combining (GSC),”
IEEE Trans. Wireless Commun., vol. 5, no. 7, pp. 1876–1885,
closed-form expression as given in (24). July 2006.
[20] H. Suzuki, “A statistical model for urban multipath propagation,”
IEEE Trans. on Commun., vol. COM-25, no. 7, pp. 673–680, July
R EFERENCES 1977.
[21] A. U. Sheikh, M. Handforth, and M. Abdi, “Indoor mobile radio
[1] S. S. Nam, S. H. Cho, D. H. Yim, and S. Choi, “More results channel at 946 MHz: Measurements and modeling,” in Proc.
on the joint statistics of partial sums of ordered random variates IEEE Veh. Technol. Conf. VTC’93, Secaucus, NJ, May 1993, pp.
with applications,” in IEEE Symposium on Communication and 73–76.
Control Theory (ICCC’15), Shenzhen, China, Nov. 2015. [22] M. Nakagami, “The m-distribution- A general formula of inten-
[2] G. L. Stüber, Principles of Mobile Communications, 2nd ed. sity distribution of rapid fading,” in Statistical Methods in Radio
Norwell, MA: Kluwer Academic Publishers, 2001. Wave Propagation. Pergamon Press, Oxford, U. K., 1960, pp.
[3] W. C. Jakes, Microwave Mobile Communications, 2nd ed. Pis- 3–36.
cataway, NJ: IEEE Press, 1994. [23] M. -S. Alouini and M. K. Simon, “Application of the Dirichlet
[4] M. K. Simon and M. -S. Alouini, Digital communications over transformation to the performance evaluation of generalized
generalized fading channels, 2nd ed. New York, NY: John Wiley selection combining over Nakagami-m fading channels,” Int. J.
& Sons, 2004. of Commun. and Networks, vol. 1, no. 1, pp. 5–13, Mar. 1999,
[5] M.-S. Alouini and M. K. Simon, “An MGF-based performance see also Proc. of the IEEE Vehicular Technology Conference
analysis of generalized selective combining over Rayleigh fading (VTC’99-Fall), pp. 953-957, Amsterdam, The Netherlands, Sept.
channels,” IEEE Trans. Commun., vol. 48, no. 3, pp. 401–415, 1999.
Mar. 2000. [24] ——, “Performance of coherent receivers with hybrid SC/MRC
over Nakagami-m fading channels,” IEEE Trans. Veh. Technol., Line: Simulation
vol. VT-48, no. 4, pp. 1155–1164, July 1999. 1 Marker: Analytical
[25] A. Annamalai and C. Tellambura, “Error rates for hybrid
Outage Probabilty, FγF (x)

SC/MRC systems on Nakagami channels,” in IEEE Wireless m=1
Commun. and Networking Conf. (WCNC’00), Chicago, IL, Sept. 0.8
2000, pp. 227–231. m=2
[26] S. Choi, M.-S. Alouini, K. A. Qaraqe, and H.-C. Yang, “Finger m=4
replacement schemes for RAKE receivers in the soft handover 0.6
region with multiple base stations,” IEEE Trans. Veh. Technol.,
vol. 57, no. 4, pp. 2114–2122, July 2008.
[27] D. Cassioli, “60 GHz UWB channel measurement and model,” 0.4
in Proc. of IEEE International Conference on Ultra-Wideband
(ICUWB’12), New York, USA, Sept. 2012, pp. 145–149.
[28] C. Rinaldi, N. Rendevski, and D. Cassioli, “Performance eval- 0.2
uation of UWB signaling at mm waves,” in Proc. of IEEE
International Conference on Ultra-Wideband (ICUWB’14), Paris,
France, Sept. 2014, pp. 379–384. 0
[29] T. Bai and R. W. Heath, “Coverage and rate analysis for mil- 0 5 10 15
x
limeter wave cellular networks,” IEEE Trans. Wireless Commun.,
vol. 14, no. 2, pp. 1100–1114, Feb. 2015.
[30] S. Lenin, “Performance analysis of millimeter wave systems with Fig. 1. Outage probability of finger replacement schemes for RAKE
amplify-and-forward relay in Nakagami-m fading,,” in Proc. of
reception in the soft handover region over i.i.d. Nakagami-m fading
4th IRF International Conference, Cochin, India, Apr. 2015, pp. channels when L = 4, N1 = · · · = N4 = 5, Nc = 3, Ns = 2,
33–35.
γT = 3, and γ̄ = 1.
[31] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and
Products, 6th ed. San Diego, CA: Academic Press, 2000.
[32] W. R. Braun and U. Dersch, “A physical mobile radio channel
model,” IEEE Trans. Veh. Technol., vol. 42, no. 2, pp. 472–482,

May 1991.

[33] A. Annamalai and C. Telumbura, “Error rates for Nakagami-

m fading multichannel reception of binary and M -ary signals,”

IEEE Trans. Commun., vol. 49, no. 1, pp. 58–68, Jan. 2001.
[34] S. Haghani and N. C. Beaulieu, “SER of M -ary NCFSK with

S + N selection combining in Nakagami fading for integer m,”
IEEE Trans. Wireless Commun., vol. 5, no. 11, pp. 3050–3055,
Nov. 2006. !
[35] R. K. Mallik, “Average of product of two Gaussian Q-functions
and its application to performance analysis in Nakagami fading,”
IEEE Trans. Commun., vol. 56, no. 8, pp. 1289–1299, Aug. 2008.
[36] C.-C. Hung, C.-T. Chiang, N.-Y. Yen, and R.-C. Wu, “Outage
probability of multiuser transmit antenna selection/maximal-ratio

combining systems over arbitrary Nakagami-m fading channels,”
IET Commun., vol. 4, no. 1, pp. 63–68, Jan. 2010.
[37] Y. Chen, R. Shi, and M. Long, “Performance analysis of amplify-
and-forward relaying with correlated links,” IEEE Trans. Veh.
Technol., vol. 62, no. 5, pp. 2344–2349, June 2013.

$
"

[38] F. J. Lopez-Martinez, D. Morales-Jimenez, E. Martos-Naya, and

J. F. Paris, “On the bivariate Nakagami-m cumulative distribution

function: closed-form expression and applications,” IEEE Trans. Fig. 2. Integral regions for Eq. (43)
Commun., vol. 61, no. 4, pp. 1404–1414, Apr. 2013.
[39] T. Chu, H. Phan, and H. Zepernick, “Hybrid interweave-underlay
spectrum access for cognitive cooperative radio networks,” IEEE
Trans. Commun., vol. 62, no. 7, pp. 2183–2197, July 2014.
[40] M. Abramowitz and I. A. Stegun, Handbook of Mathematical
Functions with Formulas, Graphs, and Mathematical Tables,
9th ed. New York, NY: Dover Publications, 1970.
fZ1 ,Z2 ,Z3 ,Z4 (z1 , z2 , z3 , z4 )

N1X
−Nc N 0 (m) Ns −1 N 00 (m) M (m)
X X X X X X X X
=
k=0
n1 ,n2 ,··· ,nm ≥0 n0 1 ,··· ,n0 m ≥0 j=0 h=0 n00 1 ,··· ,n00 m ≥0 n000 1 ,··· ,n000 m ≥0 j 0 =0 q=0
n1 +n2 +···+nm =k n0 +···+n0 =N −N −1 n00 1 +···+n00 m =h n000 1 +···+n000 m =Ns −1−h
1 m c s
“N −N ”“ k ”“ N −N −1 ”“N 0 (m)”“N −1”“ h ”“ N −1−h ”“N 00 (m)”
1 c c s s s
k 0 0
n1 , n2 , · · · , nm n 1 , · · · , n m j h n 1 , · · · , n m n00 1 , · · · , n000 m
00 00 j0
“ ” N(m)+N 0 (m)+N 00 (m)
+M (m)+Nc ·m−j−j 0 −q
“M (m)” F (−1)k+Ns −1−h m γ̄
q m−1
Q m−1
Q 0 m−1
Q 00 m−1
Q 000
(Nc −Ns −1)! (Ns −1)!Γ(m)2 (`1 !)n`1 +1 (`2 !)n `2 +1 (`3 !)n `3 +1 (`4 !)n `4 +1 (21)
`1 =0 `2 =0 `3 =0 `4 =0
„ « „ «
N 0 (m)+M (m)+m−1 m m
M (m)+N 00 (m)+m−1
× z2 exp − z2 z4 exp −
(1 + k)z4 U (z2 − z4 )
γ̄ γ̄
8 9
> „ „ « «>
< sj1 m =
× L−1
s1 “ ”(Nc −Ns −1)m exp − + s1 (Nc − Ns − 1) z2
>
: m γ̄ >
γ̄
+ s1 ;
8 9
> 0 „ „ « «>
< sj3 +q m =
× L−1
s3 “ ”(Ns −1)m exp − + s3 (h · z4 + (Ns − 1 − h) z2 ) ,
>
: m γ̄ >
γ̄
+s 3
;
N1 ! Pm−1
where F = (N1 −Nc )!
, N 00 (m) = `00 =0 `00 · n00 `00 +1 ,
8 9
> „ „ « «>
< sj1 m =
L−1
s1 “ ” exp − + s1 (Nc − Ns − 1) z2
: m + s (Nc −Ns −1)m
> γ̄ >
;
γ̄ 1
0
0 Qk
(−1)j+k ((Nc − Ns − 1) m − k1 )
j
X “ j ”„ m «j−k0 k1 =0 0
= (z1 − (Nc − Ns − 1) z2 )(Nc −Ns −1)m−1−k
0 k0 γ̄ ((Nc − Ns − 1) m)!
k =0
„ « „ «
m m
× exp − z1 exp − (Nc − Ns − 1) z2 U (z1 − (Nc − Ns − 1) z2 ) ,
γ̄ γ̄
and („ «−(Ns −1)·m „ „ « «)
m 0 +q m
L−1
s3 + s3 (s3 )j exp − + s3 (h · z4 + (Nc − Ns − 1) z2 )
γ̄ γ̄
00
k
j 0 +q+k00 Q
0 (−1) ((Ns − 1) · m − k2 )
jX +q “j 0 + q ”„ m «j 0 +q−k00 k2 =0
=
k00 γ̄ ((Ns − 1) · m)!
k00 =0
„ «
00 m
× (z3 − (h · z4 + (Ns − 1 − h) z2 ))(Ns −1)·m−1−k exp − z3 U (z3 − (h · z4 + (Ns − 1 − h) z2 )) .
γ̄
N1−Nc N 0(m) Ns−1 N 00(m)M(m)

j j +q 0
X X X X X X X X X X X
fY,W1 (x, y) =
k=0 n1 ,n2 ,··· ,nm ≥0 n0 1 ,··· ,n0 m ≥0 j=0 h=0 n00 ,··· ,n00 m ≥0
1 n000 1 ,··· ,n000 m ≥0 j 0 =0 q=0 k0 =0 k00 =0
n1+···+nm=k
n0 1+···+n0m=Nc−Ns−1 n00 1+···+n00 m=h n000 1+···+n000 m=Ns−1−h
(Nc−Ns−1)m (Ns−1)m (Ns−1)m

1−k0
−X −X k00−p2
1−k00 −1−X “N −N ”“ ”“ N −N −1 ”“N 0 (m)”“N −1”“
1 c k c s s h ”
k 0 0
n1 , · · · , nm n 1 , · · · , n m j h 00 00
n 1, · · · , n m
p1 =0 p2 =0 p3 =0
“ Ns −1−h ”“N 00 (m)”“M (m) ”“ j ”“j 0 +q”“(Nc −Ns −1) m−1−k 0”“(Ns −1) m−1−k 00”“(Ns −1) m−1−k 00 −p2 ”
000 000
n 1, · · · , n m j0 q k0 k 00 p1 p2 p3
k0 k 00 (Ns −1)m−1
(22)
00 −p −p
((Ns −1) m−k2 )(Nc − Ns − 1)p1 hp3 (Ns − 1 − h) −k
Q Q
F ((Nc −Ns −1) m−k1 ) 2 3
k1 =0 k2 =0
×
m−1 m−1 m−1 m−1
nl +1 n0 l +1 Q n00 n000 l +1
(l3 !) l3 +1 (Nc −Ns −1)! (Ns −1)!Γ(m) 2 ((Ns −1) m)! ((Nc −Ns −1) m)!
Q Q Q
(l1 !) 1 (l2 !) 2 (l4 !) 4
l1 =0 l2 =0 l3 =0 l4 =0
(Nc−Ns−1)m
« N(m)+N 0 (m)+N 00 (m) −1−k0−p1 p2 “ k+(Ns−1−h)+j+k0
(Nc −Ns −1) m−1−k 0 −p1 ”“p2 ” p4 p5
„ „ «
m +M (m)+Nc m−k0 −k00 X X 0 00 m
× x y (−1) +j +q+k −p4 −p5 exp − (x + y)
γ̄ p4 =0 p =0
p4 p5 γ̄
5
yZ x „ « „ «
m m
Z
Ns Nc−Ns α β
z2 z4 exp − (Ns −1) z2 exp − kz4 U (z2 −z4) U (x−(Nc −Ns) z2) U (y −((h+1) z4 +(Ns −(h+1)) z2)) dz2 dz4 ,
0 y γ̄ γ̄
Ns
where α = N 0 (m)+M (m)+(Nc − 1) m−k0 −k00 −3−p2 −p3 −p4 and β = N (m)+N 00 (m)+m−1+p3 +p2 −p4 .
N1X
−Nc N (m) “
X X X N1 − Nc ”“ k ”“N (m)”
fZ1 ,Z2 (z1 , z2 ) =
k n0 1 , n0 2 , · · · , n0 m j
k=0 n0 1 ,n0 2 ,··· ,n0 m ≥0 n1 ,n2 ,··· ,nm ≥0 j=0
n0 1 +n0 2 +···+n0 m =k n1 +n2 +···+nm =Nc −1
“ ”N 0 (m)+N (m)+Nc ·m−j
F (−1)k m „
m
«
γ̄ N 0 (m)+N (m)+m−1 (26)
× !„ « z2 exp − (k + 1) z2
m−1
Q n0 `0 +1
m−1
Q n`+1
γ̄
(Nc − 1)! (m − 1)! (`0 !) (`!)
`0 =0 `=0
(„ «−(Nc −1)·m „ „ « «)
m j m
× Ls−1
1
+ s1 (s1 ) exp − + s1 (Nc − 1) z2 ,
γ̄ γ̄
where F = Nc ! and
(„ «−(Nc −1)·m „ „ « «)
m m
L−1
s1 + s1 (s1 )j exp − + s1 (Nc − 1) z2
γ̄ γ̄
k 0
Q
j “j ” „ «j−k0 ((Nc − 1) · m − `) „ «
X
j+k 0 m `=0 (Nc −1)·m−1−k0 m
= (−1) (z1 − (N c − 1) z2 ) exp − z1 U (z1 − (Nc − 1) z2 ) .
k0 γ̄ ((Nc − 1) · m)! γ̄
k0 =0
(Nc−1)m−1
0 −k0 −p
NX
1−Nc N (m)
j (Nc−X
1)·m+k “N −N ”“ ”
X X X X X 1 c k
FY +W1 (x) =
k n0 1 , · · · , n0 m
k=0 n0 1 ,··· ,n0 m ≥0 n1 ,··· ,nm ≥0 j=0 k0 =0 p=0 q=0
n0 1 +···+n0 m =k n1 +···+nm =Nc −1
“ Nc −1 ”“N (m)”“ j ”“(Nc − 1) m − 1 + k0 ”“(Nc − 1) m − 1 − k0 − p”
×
n1 , · · · , nm j k0 p q
k0
Q
F (−1)k+j+(Nc −1)m−q−1((Nc − 1) m − `)(Nc − 1)p (27)
`=0 ` 0 0
´
× !„ « N (m) + N (m) + m · Nc − k − q − 2 !
m−1
P 0 n0 0 m−1
Q
(Nc − 1)! ((Nc − 1) m)!Γ (m) (` !) ` +1 (`!)n`+1
`0 =0 `=0
2 8 0 93
N (m)+N(m)+mNc
„ « >>
< −k0−q−2 t−N 0 (m)−N(m)
„ «t„ «−q−t−1 „ „ « «>
>
=7
6 m X 1 −mNc+k0+q+1
1 k m k
×6 γ q + 1, · x − (k) 1+ γ q+t+1, 1+ x 7.
4 γ̄ >
> t=0
t! N c N c γ̄ N c >
>
5
: ;
Sung Sik Nam (S’05–M’09) received the B.S. and M.S. degrees in
Electronic Engineering from Hanyang University, Korea, in 1998 and
2000, respectively. Also he received the M.S. degree in Electrical
Engineering from University of Southern California, USA, in 2003,
and the Ph.D. degree at Texas A&M University, College Station, USA,
in 2009. From 1998 to 1999, he worked as a researcher at ETRI,
Korea. From 2003 through 2004, he worked as a manager at the Korea
Telecom Corporation, Korea. From 2009 to 2010 and from 2011 to
2013, he was with Hanyang University and Sungkyunkwan University,
Korea, respectively. From 2013 to 2016, he was with Hanyang Uni-
versity, Korea. Since 2017, he has been in Korea University, Korea.
His research interests include the design and performance analysis of
wireless communication system, diversity techniques, power control,
multiuser scheduling, cooperative communications, energy harvesting,
and wireless optical communication.
Young-Chai Ko (S’97–M’01–SM’06) received the B.Sc. degree in

electrical and telecommunication engineering from the Hanyang Uni-
versity, Seoul, Korea and the M.S.E.E. and Ph.D. degrees in electrical
engineering from the University of Minnesota, Minneapolis, MN in
1999 and 2001, respectively. He was with Novatel Wireless as a
research scientist from January 2001 to March 2001. In March 2001, he
joined the Texas Instruments, Inc., wireless center, San Diego, CA, as a
senior engineer. He is now with the school of electrical engineering at
Korea university as a professor. His current research interests include
the design and evaluations of multi-user cellular system, MODEM
architecture, mm-wave and Tera Hz wireless systems.
Mohamed-Slim Alouini was born in Tunis, Tunisia. He received the

Ph.D. degree in electrical engineering from the California Institute of
Technology (Caltech), Pasadena, CA, USA, in 1998. He was with the
department of Electrical and Computer Engineering of the University
of Minnesota, Minneapolis, MN, USA, then with the Electrical and
Computer Engineering Program at the Texas A&M University at Qatar,
Education City, Doha, Qatar. Since June 2009, he has been a Professor
of Electrical Engineering in the Division of Physical Sciences and
Engineering at KAUST, Saudi Arabia., where his current research
interests include the design and performance analysis of wireless
communication systems.

Nam2017 PDF

Uploaded by

Copyright:

Available Formats

Nam2017 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Nam2017 PDF

Uploaded by

Copyright:

Available Formats

This article has been accepted for publication in a future issue of this journal, but has not been

New Closed-Form Results on Ordered

expression with the help of the interchange of multiple as

V. D ISCUSSIONS AND C ONCLUSIONS funded by the Korea government(MSIP) (No. 2014-

Note that closed-form results on ordered statistics of (31)

" # Therefore, the intersection becomes the shaded region

Outage Probabilty, FγF (x)

[33] A. Annamalai and C. Telumbura, “Error rates for Nakagami- 

m fading multichannel reception of binary and M -ary signals,” 

IEEE Trans. Wireless Commun., vol. 5, no. 11, pp. 3050–3055,

[35] R. K. Mallik, “Average of product of two Gaussian Q-functions 

and its application to performance analysis in Nakagami fading,” 

probability of multiuser transmit antenna selection/maximal-ratio

Technol., vol. 62, no. 5, pp. 2344–2349, June 2013.

[38] F. J. Lopez-Martinez, D. Morales-Jimenez, E. Martos-Naya, and

J. F. Paris, “On the bivariate Nakagami-m cumulative distribution

fZ1 ,Z2 ,Z3 ,Z4 (z1 , z2 , z3 , z4 )

N1−Nc N 0(m) Ns−1 N 00(m)M(m)

(Nc−Ns−1)m (Ns−1)m (Ns−1)m

Young-Chai Ko (S’97–M’01–SM’06) received the B.Sc. degree in

Mohamed-Slim Alouini was born in Tunis, Tunisia. He received the

You might also like

[33] A. Annamalai and C. Telumbura, “Error rates for Nakagami-

m fading multichannel reception of binary and M -ary signals,”

[35] R. K. Mallik, “Average of product of two Gaussian Q-functions

and its application to performance analysis in Nakagami fading,”