Beruflich Dokumente
Kultur Dokumente
Abstract—Energy-efficient transmission is an important aspect requirement and channel condition. This allows simultaneous
of wireless system design due to limited battery power in data transmission for several users. For simplicity, we investi-
mobile devices. We consider uplink energy-efficient transmission gate energy-efficient OFDMA communication with flat fading
in OFDMA systems since mobile stations are battery powered. channels in this paper.
We account for both circuit and transmit power when design- Consider uplink transmission in an OFDMA network with
ing energy-efficient communication mechanisms and emphasize one base station (BS) and multiple users, i.e. mobile stations.
energy efficiency over peak rates or throughput. Both link Denote N and K as the numbers of users and subchannels,
adaptation and resource allocation schemes are developed to respectively. Denote ci as the number of subchannels assigned
optimize the overall bits transmitted per Joule of energy, which to User i. Each subchannel will be assigned to one user
allows for maximum energy savings in a network. Our simulation exclusively at each frame slot. Hence,
results show that the proposed schemes significantly improve
N
X
energy efficiency.
ci ≤ K. (1)
Index Terms– energy efficiency, OFDMA, bits per Joule, link
i=1
adaptation, resource allocation
Denote ri as the achievable data rate on each subchannel by
I. I NTRODUCTION User i, then the data rate of User i is
Power efficiency is becoming increasingly important for
Ri = ri ci . (2)
wireless communication systems due to limited battery re-
sources in mobile devices. Unfortunately, battery technol- The BS allocates subchannels to improve the overall network
ogy has not progressed as fast as silicon technology [1]. energy efficiency, which is measured by the number of bits
Hence, recent energy-efficient management schemes [2]–[4] transmitted per Joule. Additionally, each user also adjusts
have focused on minimizing energy consumption rather than transmit power and modulation order for further optimization.
throughput maximization [5]. Additionally, orthogonal fre-
quency division multiple access (OFDMA) has emerged as III. O PTIMAL E NERGY-E FFICIENT L INK A DAPTATION
one of the prime multiple access schemes for next generation
multi-user broadband wireless networks. However, limited This section considers per link adaptation schemes that
research has been done for energy-efficient communication in will result in minimum energy consumption, or equivalently,
OFDMA systems. In this paper, we consider uplink energy- maximum energy-efficiency. Throughout this section, assume
efficient transmission in OFDMA systems to improve battery c subchannels are assigned. Since we focus on per link energy-
consumption at the mobiles. We account for both circuit and efficient optimization, subscripts indicating different users will
transmit power when designing link adaptation and resource be dropped subsequently.
allocation schemes, and emphasize energy efficiency over peak
rates or throughput. We initially focus on the case of flat-fading A. Energy-Efficient Transmission Rate
OFDMA channels, and defer the frequency selective case to
future work. Power consumption of a mobile station in transmission
The rest of the paper is organized as following. In Section mode consists of two parts. The first is circuit power, denoted
II, we briefly describe the system model. Then we develop as PC , which is independent of data rate and exists whenever
optimal energy-efficient link adaptation and network resource the user is in transmission mode. The second is transmit
allocation schemes in Sections III and IV respectively. Finally, power, PT (R), which depends on data transmission rate, R.
we conclude the paper in Section V. For example, we consider an additive white Gaussian noise
(AWGN) channel with signal bandwidth W , the achievable
II. S YSTEM D ESCRIPTION data rate is given by the Shannon capacity as
Multiple access is achieved in OFDMA by assigning sub- PT g
channels to individual users based on quality of service (QoS) R = W log(1 + ), (3)
No W
where g is the channel power gain, No is the power spectral and the overall data rate is
density. Hence, cbl
R = cr = . (9)
PT (R) = (e
R
W − 1)No W/g, (4) Ts + τ
Consequently, for a given data transmission rate, the number
which is monotonically increasing and strictly convex in R. of bits transmitted per symbol will be b = R(Tcls +τ ) .
In general, we assume PT (R) to be monotonically increasing The bit-error rate (BER) for coherently detected M-QAM
and strictly convex and PT (0) = 0. with Gray mapping over an AWGN channel is approximated
The overall power used for data transmission is by [8] µ ¶
1.5γ
P (R) = PC + PT (R). (5) Pe (γ) ≈ 0.2 exp − , (10)
M −1
The number of bits transmitted per Joule of energy, called
where γ is the signal-to-noise ratio (SNR). Denote g to be the
energy efficiency, is used as a performance measure, and it is
power gain of the channel. The SNR on each subchannel will
defined as
be
R R PT (R)g
U (R) = = . (6) γ= , (11)
P (R) PC + PT (R) cNo W
The network is optimized for the highest energy efficiency. where No is the power spectral density and W is the signal
Thus, the intended data rate is bandwidth in each subchannel. For a given BER, Pe , the
required SNR can be determined by (10). Consequently, the
R
R∗ = arg max U (R) = arg max . (7) required transmit power will be
R R PC + PT (R)
γcNo W
The optimal transmission data rate is given by the following PT (R) = = A(1 − 2BR ), (12)
g
theorem, which is proved in Appendix A.
where A = 2c ln(5P3ge )No W and B = Tscl+τ . Usually, 5Pe < 1
Theorem 1. If PT (R) is monotonically increasing and strictly for effective transmission, therefore, A < 0. It can be seen that
convex in R, there exists a unique globally optimal transmis- PT (R) is monotonically increasing and strictly convex in R.
sion data rate for (7) given by PC characterizes circuit power consumption in both the data
and signalling intervals. According to Theorem 1, the desired
PC + PT (R∗ )
R∗ = 0 , (8) data rate is ∗
PT (R∗ ) A(2BR − 1) − Pc
R∗ = . (13)
0 AB2BR∗ ln 2
where PT (·) is the first derivative of function PT (·).
R∗ (Ts +τ ) R∗ (Ts +τ )
Correspondingly, b∗ = cl and M ∗ = 2 cl .
Data Interval n Signaling Interval
B. Characteristics of Energy-Efficient Transmission
Theorems 2 and 3 describes the characteristics of energy-
efficient link transmission and are proved in Appendices B and
Sym Sym Sym Sym Sym Sym C respectively.
1 2 3 l-1 l 1
0.5
1.4
1.2 0.4
1
0.3
0.8
0.6 0.2
0.4
0.1
0.2
0
0 0 0.5 1 1.5 2 2.5 3 3.5 4
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Modulation Order (bits/symbol)
Data Rate (Mbits/s)
(a) Relationship of energy efficiency, distance, modulation and transmis- (b) Relationship of energy efficiency, modulation and subchannel assign-
sion data rate ment
Fig. 2: Energy-efficiency relationship of per link transmission
4
Proposed Proposed
2QAM Adaptive15dBm
4QAM Adaptive20dBm
3
8QAM Adaptive25dBm
3 Adaptive30dBm
Energy (Joule)
Energy (Joule)
1
1
0 0
0.3 0.5 0.7 0.9 1.1 1.3 1.5 0.3 0.5 0.7 0.9 1.1 1.3 1.5
Distance (km) Distance (km)
(a) Comparison with fixed modulation (normalized energy) (b) Comparison with adaptive modulation (normalized energy)
Fig. 3: Energy consumed for transmitting one megabit
Denote Wi (ci ) = log(Vi (ci )). Problem (20) is equivalent to Hence, Wi (ci ) is strictly concave in ci . Problem (21) is
strictly concave and unique globally optimal assignment exists.
X X
c∗ = arg max log(Vi (ci )) = arg max Wi (ci ), (21a) Similar to (18), the optimal assignment is given by
c c
i i 0
c∗i = Wi −1 (λ∗ ), (23)
subject to P 0
N
X where λ∗ satisfies i Wi −1 (λ∗ ) = K.
ci ≤ K. (21b)
i=1
25 6
OptEE
PropEE
OptEE
RREE
PropEE
RREE
5 RRTrad
PrpTrad15dBm
20 RRTrad PrpTrad25dBm
PropTrad33dBm
PropTrad15dBm
Energy Efficiency (Mbits/Joule)
PropTrad25dBm
PropTrad33dBm
4
Throughput (Mbits/s)
15
3
10
2
5 1
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
User Number User Number