Transcript
Page 1: [IEEE 2011 8th International Symposium on Wireless Communication Systems (ISWCS 2011) - Aachen, Germany (2011.11.6-2011.11.9)] 2011 8th International Symposium on Wireless Communication

Sum-Rate Analysis of the Two-Way Relay Channelin Spectrum-Sharing Environments

Najmeh MadaniKousha Communication Industries, Tehran, Iran

Email:[email protected]

Abstractβ€”This paper examines the two-way relay channel(TWRC) in a spectrum-sharing environment. Two secondaryusers, a primary user and a relay are the involved nodes inour model. The transmit power of the secondary users and therelay are adapted optimally to approach the maximum achievablesum-rate while keeping the interference level at the primaryuser below a threshold. Numerical simulations are conductedto demonstrate the performance of the proposed scenario.

I. INTRODUCTION

Relay-based networks have been widely perceived as greatassets to wireless communications in terms of coverage exten-sion, throughput improvement and saving network resources.In this context, two-way relaying which provides a bi-directioncommunication is of great importance. The two-way relaychannel (TWRC) is privileged as it helps in redeeming thespectral loss one-way relaying suffers from [1]. In a sim-ple TWRC model, two terminals exchange their messagesvia a relay terminal. The terminals operate in half-duplexmode. Protocols such as amplify-and-forward (AF), decode-and-forward (DF), and compress-and-forward (CF) have beenutilized for information transmission in TWRC [2].

The merits of TWRC in power saving and spectral efficiencymake it to be best adopted in spectrum-sharing networks. Inthese networks, secondary users use the spectrum of primaryusers in a way that the links of the primary users are not af-fected [3]. To protect the QoS of primary users, the secondaryusers have to be cautious about the transmit power. Applicationof the TWRC in spectrum-sharing networks has been recentlystudied in [4]. This work has concentrated on sum-rate opti-mization by relay beamforming and power allocation at thesecondary users while maintaining the received power at theprimary user below a certain level. The amplify-and-forwardprotocol has been considered in this research.

In this paper, we aim at the maximum achievable sum-rate of the TWRC using decode-and-forward protocol in aspectrum-sharing environment. The decode-and-forward sup-ports different coding schemes at the relay node, e.g., superpo-sition coding [1], [5], XOR coding [6], [5], and optimal coding[7], [8]. We follow the superposition scheme. However, theproposed solution is flexible and can accommodate other cod-ing strategies. It is assumed that the channel state information(CSI) is provided at the nodes. The sum-rate maximizationis performed with the help of CSI while complying powerconstraints. The constraints are introduced by the spectrum-sharing system and the transmit power limitations of the nodes.

We consider a Rayleigh flat fading environment.The rest of the paper is organized as follows. Section II

describes the system model and the problem formulation. Sec-tion III provides the sum-rate optimization strategy. Simulationresults are presented in section IV and finally conclusions aremade in section V.

II. SYSTEM MODEL

Two source nodes 𝑆1 and 𝑆2 exchange information withthe help of a relay node 𝑅. All nodes operate in half-duplexmode and information exchange is performed within two time-slots. The source nodes use the same spectrum band and shareit with a primary user. All the channels are assumed to beconstant during the two time-slots. In the first time-slot, 𝑆1 and𝑆2 transmit to 𝑅 simultaneously, so 𝑅 receives the followingsignal:

π‘¦π‘Ÿ [𝑛] =√

β„Ž1𝑠π‘₯1 [𝑛] +√

β„Ž2𝑠π‘₯2 [𝑛] + π‘§π‘Ÿ [𝑛] , (1)

where 𝑛 represents the time index, β„Ž1𝑠 and β„Ž2𝑠 are the channelpower gains from 𝑆1 to 𝑅 and from 𝑆2 to 𝑅, respectively, andπ‘§π‘Ÿ[𝑛] indicates the AWGN with the power spectral density of𝑁0. The transmit power of 𝑆1 and 𝑆2 are denoted by 𝑝1 and𝑝2 and are constrained to the predefined value of 𝑄𝑠:

𝑝1 ≀ 𝑄𝑠,

𝑝2 ≀ 𝑄𝑠. (2)

As spectrum-sharing entails, signal transmission betweensource nodes and relay should not harm the performance of theprimary user. Let 𝑄𝑝 indicates the peak power that the primaryuser can tolerate without QoS violation, thus the receivedinterference at the primary user in the first time-slot shouldbe limited as:

𝑝1β„Ž1𝑝 + 𝑝2β„Ž2𝑝 ≀ 𝑄𝑝, (3)

where β„Ž1𝑝 and β„Ž2𝑝 indicate the channel power gains from 𝑆1 tothe primary user, and from 𝑆2 to the primary user, respectively.The relay employs a two-phase decode-and-forward strategy[9]. It decodes the received signals, re-encode them andbroadcasts the signals to the source nodes in the second time-slot. The received signals at 𝑆1 and 𝑆2 are expressed as:

𝑦1 [𝑛] =√

β„Ž1𝑠��1 [𝑛] +√

β„Ž1𝑠��2 [𝑛] + 𝑧1 [𝑛] , (4)

𝑦2 [𝑛] =√

β„Ž2𝑠��1 [𝑛] +√

β„Ž2𝑠��2 [π‘˜] + 𝑧2 [𝑛] , (5)

2011 8th International Symposium on Wireless Communication Systems, Aachen

978-1-61284-402-2/11/$26.00 Β©2011 IEEE 101

Page 2: [IEEE 2011 8th International Symposium on Wireless Communication Systems (ISWCS 2011) - Aachen, Germany (2011.11.6-2011.11.9)] 2011 8th International Symposium on Wireless Communication

π‘„π‘β„Ž1𝑝

π‘„π‘β„Ž2𝑝

𝑃1

𝑃2

(a)

𝑄𝑠

π‘„π‘β„Ž1𝑝

𝑃1

𝑃2

(b)

𝑄𝑠

π‘„π‘β„Ž2𝑝

𝑃1

𝑃2

(c)

𝑄𝑠

𝑄𝑠

𝑃1

𝑃2

(d)

𝑄𝑠

𝑄𝑠

𝑃1

𝑃2

(e)

Fig. 1. Possible power regions for source nodes.

where 𝑧1[𝑛] and 𝑧2[𝑛] refer to the AWGN with the powerdensity of 𝑁0. οΏ½οΏ½1[𝑛] and οΏ½οΏ½2[𝑛] are the re-encoded signalscorresponding to the signal streams of 𝑆1 and 𝑆2, respectively.𝑅 divides the power between οΏ½οΏ½1[𝑛] and οΏ½οΏ½2[𝑛] according to thesuperposition coding. Let π‘„π‘Ÿ defines the maximum power that𝑅 can transmit, π‘π‘Ÿ1 denotes the assigned power to οΏ½οΏ½1[𝑛] andπ‘π‘Ÿ2 indicates the allocated power to οΏ½οΏ½2[𝑛]. It is obvious that:

π‘π‘Ÿ1 + π‘π‘Ÿ2 ≀ π‘„π‘Ÿ. (6)

To protect the primary user from harmful interference, thesignal transmission in the second time-slot should satisfy thefollowing constraint:

π‘π‘Ÿ1β„Žπ‘ + π‘π‘Ÿ2β„Žπ‘ ≀ 𝑄𝑝, (7)

where β„Žπ‘ is the channel power gain between 𝑅 and theprimary user. It is assumed that the channel power gainsare independent and identically distributed (i.i.d). It is furtherassumed that the source nodes and the relay are fully awareof codebooks and channel power gains. We continue with theassumption of 𝑁0𝐡 = 1, where 𝐡 is the available bandwidth.

In the first time-slot, when the source nodes transmit andthe relay receives, we have a multiple access channel (MAC).Using sequential decoding at 𝑅, the achievable rate region ischaracterized by the following relations [10]:

𝑅1𝑅 ≀ log (1 + 𝑝1β„Ž1𝑠) , (8)

𝑅2𝑅 ≀ log (1 + 𝑝2β„Ž2𝑠) , (9)

𝑅1𝑅 +𝑅2𝑅 ≀ log (1 + 𝑝1β„Ž1𝑠 + 𝑝2β„Ž2𝑠) . (10)

The second time-slot represents a broadcast channel wherethe achievable rates are derived by the fact that each nodeknows about it’s transmitted signal. Each source node can sub-tract it’s self-interference from the received signal, describedby (4) or (5). In this time-slot the rates satisfy:

𝑅𝑅1 ≀ log (1 + π‘π‘Ÿ1β„Ž1𝑠) , (11)

𝑅𝑅2 ≀ log (1 + π‘π‘Ÿ2β„Ž2𝑠) . (12)

The rates are expressed in terms of bits/sec/Hz. The achievablerate from 𝑆1 to 𝑆2 via 𝑅 is expressed as the minimum of therates during the two time-slots.

𝑅12 ≀ 1

2min {log (1 + 𝑝1β„Ž1𝑠) , log (1 + π‘π‘Ÿ2β„Ž2𝑠)} . (13)

Similarly, the achievable rate from 𝑆2 to 𝑆1 satisfies:

𝑅21 ≀ 1

2min {log (1 + 𝑝2β„Ž2𝑠) , log (1 + π‘π‘Ÿ1β„Ž1𝑠)} . (14)

The factor 1/2 follows because the information transmissionis performed in two time-slots. Our goal in this paper is tomaximize the achievable sum-rate of the modeled TWRCby power allocation at the source nodes and the relay. Thisoptimization problem which is subjected to the constraints (2),(3), (6) and (7) can be formulated as follows:

π‘…π‘ π‘’π‘šπ‘šπ‘Žπ‘₯ = max

𝑝1,𝑝2,π‘π‘Ÿ1,π‘π‘Ÿ2

min

{𝑅12 +𝑅21,

1

2log(1 + 𝑝1β„Ž1𝑠 + 𝑝2β„Ž2𝑠)

}.

(15)

III. SUM-RATE ANALYSIS

The constraints introduced in the previous section definetransmit power regions for the nodes. As Fig.1 demonstrates,(2) and (3) shape the power regions of the source nodes. Weproceed with the Fig.1a. The solution will be extended nextto cover all the cases. The selected region corresponds to thefollowing conditions:

β„Ž1𝑝 >𝑄𝑝

𝑄𝑠,

β„Ž2𝑝 >𝑄𝑝

𝑄𝑠. (16)

For the case of 𝑅, the transmit power should satisfy (6) and(7). We can brief these two conditions as:

π‘π‘Ÿ1 + π‘π‘Ÿ2 ≀ 𝐴, 𝐴 =

{π‘„π‘Ÿ β„Žπ‘ <

𝑄𝑝

π‘„π‘Ÿπ‘„π‘

β„Žπ‘β„Žπ‘ >

𝑄𝑝

π‘„π‘Ÿ

. (17)

In the broadcast phase, the sum-rate is:

𝑅12 +𝑅21 =1

2{log(1 + π‘π‘Ÿ1β„Ž1𝑠) + log(1 + π‘π‘Ÿ2β„Ž2𝑠)} . (18)

The maximum value it can reach while satisfying (17) can beeasily calculated and is as follows:

π‘…π΅πΆπ‘šπ‘Žπ‘₯ =

1

2log(1+

𝐴2

4β„Ž1π‘ β„Ž2𝑠+

𝐴

2(β„Ž1𝑠+β„Ž2𝑠)+

(β„Ž2𝑠 βˆ’ β„Ž1𝑠)2

4β„Ž1π‘ β„Ž2𝑠).

(19)This maximum occurs when π‘βˆ—π‘Ÿ1 = 𝐴

2 + 12

(1

β„Ž2π‘ βˆ’ 1

β„Ž1𝑠

)and

π‘βˆ—π‘Ÿ2 = 𝐴2 +

12

(1

β„Ž1π‘ βˆ’ 1

β„Ž2𝑠

)and is valid as long as βˆ’π΄ < 1

β„Ž2π‘ βˆ’

1β„Ž1𝑠

< 𝐴. To have π‘…π΅πΆπ‘šπ‘Žπ‘₯ as the maximum achievable sum-rate

of the modeled TWRC, the source nodes should transmit withthe power levels that ensure:

𝑝1β„Ž1𝑠 β‰₯ π‘βˆ—π‘Ÿ2β„Ž2𝑠 equivalently 𝑝1 β‰₯ π΄β„Ž2𝑠

2β„Ž1𝑠+

(β„Ž2𝑠 βˆ’ β„Ž1𝑠)

2β„Ž21𝑠︸ οΈ·οΈ· οΈΈ

π‘βˆ—1

,

(20)

102

Page 3: [IEEE 2011 8th International Symposium on Wireless Communication Systems (ISWCS 2011) - Aachen, Germany (2011.11.6-2011.11.9)] 2011 8th International Symposium on Wireless Communication

P1

P2

𝛼

𝛽

π‘₯1

π‘₯2

(π‘βˆ—1, 𝑝2)

(𝑝1, π‘βˆ—2)

(a)

𝛼

𝛽

π‘₯1

π‘₯2

(π‘βˆ—1, 𝑝2)

(𝑝1, π‘βˆ—2)

𝑃1

𝑃2

(b)

𝛼

𝛽

π‘₯1

π‘₯2

(π‘βˆ—1, 𝑝2)

(𝑝1, π‘βˆ—2)

𝑃1

𝑃2

(c)

Fig. 2. Shaded areas indicate the power levels of the source nodes which result in a maximized broadcast sum-rate.

and

𝑝2β„Ž2𝑠 β‰₯ π‘βˆ—π‘Ÿ1β„Ž1𝑠 equivalently 𝑝2 β‰₯ π΄β„Ž1𝑠

2β„Ž2𝑠+

(β„Ž1𝑠 βˆ’ β„Ž2𝑠)

2β„Ž22𝑠︸ οΈ·οΈ· οΈΈ

π‘βˆ—2

.

(21)Meanwhile, the transmit power of the source nodes shouldresult in a MAC sum-rate greater than 𝑅𝐡𝐢

π‘šπ‘Žπ‘₯:

𝑝1β„Ž1𝑠 + 𝑝2β„Ž2𝑠 β‰₯ 𝐴2

4β„Ž1π‘ β„Ž2𝑠 +

𝐴

2(β„Ž1𝑠 + β„Ž2𝑠) +

(β„Ž2𝑠 βˆ’ β„Ž1𝑠)2

4β„Ž1π‘ β„Ž2𝑠.

(22)Fig.2 illustrates (20), (21), and (22) with the power region

of the source nodes in the 𝑃1 βˆ’ 𝑃2 plane. Line 1 and Line2 which represent boundary points of (3) and (22) have thefollowing equations, respectively:

Line 1:𝑝1π‘₯1

+𝑝2π‘₯2

= 1, (23)

where π‘₯1 =𝑄𝑝

β„Ž1𝑝and π‘₯2 =

𝑄𝑝

β„Ž2𝑝.

Line 2:𝑝1𝛼

+𝑝2𝛽

= 1, (24)

where 𝛼 = π΄β„Ž2𝑠

β„Ž1𝑠+ (β„Ž2π‘ βˆ’β„Ž1π‘ βˆ’π΄β„Ž1π‘ β„Ž2𝑠)

2

4β„Ž21π‘ β„Ž2𝑠

and 𝛽 = 𝐴 +(β„Ž2π‘ βˆ’β„Ž1π‘ βˆ’π΄β„Ž1π‘ β„Ž2𝑠)

2

4β„Ž1π‘ β„Ž22𝑠

. The shaded parts in Fig.2a, Fig.2b, andFig.2c indicate the regions where (20), (21), and (22) aresatisfied. Here, the broadcast sum-rate is the controlling rateand 𝑅𝐡𝐢

π‘šπ‘Žπ‘₯ is the response of our optimization problem definedin (15).

Line 1 and Line 2 might be positioned in a way thatno points in the power region validate (20), (21), and (22)concurrently. In this case, 𝑅𝐡𝐢

π‘šπ‘Žπ‘₯ is not supported anymore.Fig.3a demonstrates a possible scenario. In Fig.3a, the area ofthe power region which fulfills (22) and (21) dose not qualify(20). Hear, 𝑅12+𝑅21 = 1

2{log(1+𝑝1β„Ž1𝑠)+log(1+π‘π‘Ÿ1β„Žβ„Ž1𝑠)}is the controlling sum-rate. To improve the sum-rate, π‘π‘Ÿ1 andπ‘π‘Ÿ2 are selected as follows:

π‘π‘Ÿ1 = π΄βˆ’ π‘π‘Ÿ2, π‘π‘Ÿ2 = 𝑝1π‘ β„Ž1𝑠

β„Ž2𝑠. (25)

It can be verified that the controlling sum-rate is ascendingin terms of 𝑝1 as long as 𝑝1 ≀ π‘βˆ—1. This is on the contrary

to the MAC sum-rate which is descending as 𝑝1 increases.The downward sum-rate of the MAC phase is concluded fromthe fact that π‘₯2

π‘₯1> β„Ž1𝑠

β„Ž2𝑠and 𝑝2 = π‘₯2(1 βˆ’ 𝑝1

π‘₯1) is adopted. The

optimum value for the controlling sum-rate is achieved whenit equals the MAC sum-rate:

1+𝑝1β„Ž1𝑠+π‘₯2(1βˆ’ 𝑝1π‘₯1

)β„Ž2𝑠 = (1+𝑝1β„Ž1𝑠)(1+(π΄βˆ’π‘1 β„Ž1𝑠

β„Ž2𝑠))β„Ž1𝑠).

(26)The power level of 𝑆1 which comes from the above equationis:

𝑝1 =π΄β„Ž2𝑠

2β„Ž1𝑠+

π‘₯2β„Ž22𝑠

2π‘₯1β„Ž31𝑠

βˆ’ 1

2β„Ž1π‘ βˆ’βˆš

(β„Ž21𝑠 βˆ’ π‘₯2β„Ž2

2𝑠

π‘₯1βˆ’π΄β„Ž2

1π‘ β„Ž2𝑠)2 βˆ’ 4β„Ž31𝑠(π‘₯2β„Ž2

2𝑠 βˆ’π΄β„Ž1π‘ β„Ž2𝑠)

2β„Ž31𝑠

,

(27)

and the maximum achievable sum rate is:

π‘…π‘ π‘’π‘šπ‘šπ‘Žπ‘₯ =

1

2log(1 + (

π΄β„Ž21π‘ β„Ž2𝑠 +

π‘₯2β„Ž22𝑠

π‘₯1βˆ’ β„Ž2

1𝑠

2π‘₯1β„Ž31𝑠

βˆ’βˆš(β„Ž2

1𝑠 βˆ’ π‘₯2β„Ž22𝑠

π‘₯1βˆ’π΄β„Ž2

1π‘ β„Ž2𝑠)2 βˆ’ 4β„Ž31𝑠(π‘₯2β„Ž2

2𝑠 βˆ’π΄β„Ž1π‘ β„Ž2𝑠)

2π‘₯1β„Ž31𝑠

)

(π‘₯1β„Ž1𝑠 βˆ’ π‘₯2β„Ž2𝑠) + π‘₯2β„Ž2𝑠). (28)

Fig.3b presents another case where π‘…π΅πΆπ‘šπ‘Žπ‘₯ is not supported.

The points of the power region which fulfill (22) and (20)do not satisfy (21). The controlling sum-rate is 𝑅12 +𝑅21 =12{log(1+𝑝2β„Ž2𝑠)+log(1+(π΄βˆ’ 𝑝2β„Ž2𝑠

β„Ž1𝑠)β„Ž2𝑠)} and it’s maximum

is obtained similar to the previous case:

π‘…π‘ π‘’π‘šπ‘šπ‘Žπ‘₯ =

1

2log(1 + (

π΄β„Ž22π‘ β„Ž1𝑠 +

π‘₯1β„Ž21𝑠

π‘₯2βˆ’ β„Ž2

2𝑠

2π‘₯2β„Ž32𝑠

βˆ’βˆš(β„Ž2

2𝑠 βˆ’ π‘₯1β„Ž21𝑠

π‘₯2βˆ’π΄β„Ž2

2π‘ β„Ž1𝑠)2 βˆ’ 4β„Ž32𝑠(π‘₯1β„Ž2

1𝑠 βˆ’π΄β„Ž2π‘ β„Ž1𝑠)

2π‘₯2β„Ž32𝑠

)

(π‘₯2β„Ž2𝑠 βˆ’ π‘₯1β„Ž1𝑠) + π‘₯1β„Ž1𝑠). (29)

103

Page 4: [IEEE 2011 8th International Symposium on Wireless Communication Systems (ISWCS 2011) - Aachen, Germany (2011.11.6-2011.11.9)] 2011 8th International Symposium on Wireless Communication

𝛼

𝛽

π‘₯1

π‘₯2

(π‘βˆ—1, 𝑝2)(π‘βˆ—

1, 𝑝2)(π‘βˆ—

1, 𝑝2)

(𝑝1, π‘βˆ—2)(𝑝1, π‘βˆ—

2)(𝑝1, π‘βˆ—

2)

𝑃1

𝑃2

(a)

𝛼

𝛽

π‘₯1

π‘₯2

(π‘βˆ—1, 𝑝2)

(𝑝1, π‘βˆ—2)

𝑃1

𝑃2

(b)

𝛼

𝛽

π‘₯1

π‘₯2(π‘βˆ—

1, 𝑝2)

(𝑝1, π‘βˆ—2)

𝑃1

𝑃2

(c)

Fig. 3. No points in the power region support the maximized broadcast sum-rate.

To reach this rate, the power level of 𝑆2 should be:

𝑝2 =π΄β„Ž1𝑠

2β„Ž2𝑠+

π‘₯1β„Ž21𝑠

2π‘₯2β„Ž32𝑠

βˆ’ 1

2β„Ž2π‘ βˆ’βˆš

(β„Ž22𝑠 βˆ’ π‘₯1β„Ž2

1𝑠

π‘₯2βˆ’π΄β„Ž2

2π‘ β„Ž1𝑠)2 βˆ’ 4β„Ž32𝑠(π‘₯1β„Ž2

1𝑠 βˆ’π΄β„Ž2π‘ β„Ž1𝑠)

2β„Ž32𝑠

.

(30)

Fig.3c refers to a new case where π‘₯1 < 𝛼 and π‘₯2 < 𝛽.No points in the power region validate (22) and consequentlythe MAC sum-rate is always lower than 𝑅𝐡𝐢

π‘šπ‘Žπ‘₯. For this casethe maximum achievable sum-rate occurs in the MAC phase.Adopting 𝑝1 = π‘₯1(1βˆ’ 𝑝2

π‘₯2), the sum-rate is:

𝑅12 +𝑅21 =1

2log(1 + π‘₯1β„Ž1𝑠 + 𝑝2(β„Ž2𝑠 βˆ’ π‘₯1β„Ž1𝑠

π‘₯2)). (31)

As long as π‘₯2

π‘₯1> β„Ž1𝑠

β„Ž2𝑠, the rate is increasing in terms of 𝑝2.

Therefore, the maximum sum-rate is as follows:

π‘…π‘€π΄πΆπ‘šπ‘Žπ‘₯ =

1

2log(1 + π‘₯2β„Ž2𝑠). (32)

This is obtained when 𝑝2 = π‘₯2. To have π‘…π‘€π΄πΆπ‘šπ‘Žπ‘₯ as the

optimum sum-rate, it is necessary that 𝐴 β‰₯ π‘₯2β„Ž2𝑠

β„Ž1𝑠. This leads

to a one-way relaying system where 𝑆1 transmits and 𝑆2

receives. When 𝐴 < π‘₯2β„Ž2𝑠

β„Ž1𝑠, the relay can not support 𝑅𝑀𝐴𝐢

π‘šπ‘Žπ‘₯ .Similar to our reasoning in the broadcast phase, the controllingsum-rate, it’s maximum and the power allocation strategy atthe nodes are according to (25), (27), and (28). For the casewhere π‘₯2

π‘₯1< β„Ž1𝑠

β„Ž2𝑠and 𝐴 β‰₯ π‘₯1β„Ž1𝑠

β„Ž2𝑠, the maximum achievable

sum-rate is:𝑅𝑀𝐴𝐢

π‘šπ‘Žπ‘₯ =1

2log(1 + π‘₯1β„Ž1𝑠). (33)

if 𝐴 < π‘₯1β„Ž1𝑠

β„Ž2𝑠, (29) and (30) determine the result.

In summery, to obtain the maximum sum-rate of the consid-ered TWRC, 𝑅𝐡𝐢

π‘šπ‘Žπ‘₯ and π‘…π‘€π΄πΆπ‘šπ‘Žπ‘₯ are determined. The one with

the lower value limits the sum-rate and is a possible answerfor (15). Based on the channel power gains and the powerconstraints of the nodes, it is then decided that the sum-rateis supported or a lower rate according to (28) or (30) is thetrue rate. The power level of the source nodes and the relayare assigned correspondingly. To have a general solution, some

modifications are necessary. In evaluating π‘…π΅πΆπ‘ π‘’π‘š, the following

extension is applicable:⎧⎨⎩

π‘βˆ—π‘Ÿ1 = 𝐴, π‘βˆ—π‘Ÿ2 = 0,

𝛼 = 𝐴, 𝛽 = π΄β„Ž1𝑠

β„Ž2𝑠,

if 1β„Ž2π‘ βˆ’ 1

β„Ž1𝑠β‰₯ 𝐴

π‘βˆ—π‘Ÿ1 = 0, π‘βˆ—π‘Ÿ2 = 𝐴,

𝛼 = π΄β„Ž2𝑠

β„Ž1𝑠, 𝛽 = 𝐴,

if 1β„Ž2π‘ βˆ’ 1

β„Ž1𝑠≀ βˆ’π΄

as stated before, otherwise.

(34)

We reconsider π‘…π΅πΆπ‘ π‘’π‘š = 1

2{log(1+π‘βˆ—π‘Ÿ1β„Ž1𝑠)+log(1+π‘βˆ—π‘Ÿ2β„Ž2𝑠)},π‘βˆ—1 =

π‘βˆ—π‘Ÿ2β„Ž2𝑠

β„Ž1𝑠and π‘βˆ—2 =

π‘βˆ—π‘Ÿ1β„Ž1𝑠

β„Ž2𝑠. 𝑄𝑠 is another important factor

that should be included in sum-rate optimization. It has a rolein shaping the power region of the source nodes, and thereforethe MAC sum-rate is directly dependent in. We assume thatπ‘₯2

π‘₯1> β„Ž1𝑠

β„Ž2𝑠. The maximum Mac sum-rate is:

π‘…π‘€π΄πΆπ‘šπ‘Žπ‘₯ =

1

2log(1 +π‘šπ‘–π‘›1β„Ž1𝑠 +π‘šπ‘–π‘›2β„Ž2𝑠) (35)

Where π‘šπ‘–π‘›2 = min(𝑄𝑠, π‘₯2), ��𝑠 = π‘₯1(1 βˆ’ π‘šπ‘–π‘›2

π‘₯2), and

π‘šπ‘–π‘›1 = min(𝑄𝑠, ��𝑠). Since 𝑆2 has a better link condition,β„Ž2𝑝

β„Ž2𝑠<

β„Ž1𝑝

β„Ž1𝑠, the maximum MAC sum-rate is derived by 𝑝2. If

π‘…π‘€π΄πΆπ‘šπ‘Žπ‘₯ < 𝑅𝐡𝐢

π‘šπ‘Žπ‘₯ then:

π‘…π‘ π‘’π‘šπ‘šπ‘Žπ‘₯ =

⎧⎨⎩

12 (log(1 +π‘„π‘ β„Ž2𝑠) + log(1 + (π΄βˆ’π‘„π‘ 

β„Ž2𝑠

β„Ž1𝑠)β„Ž2𝑠)),

if 𝑄𝑠 < π‘₯2, 𝑄𝑠 < π‘βˆ—2 andπ‘„π‘ β„Ž2𝑠

β„Ž1𝑠+ π‘šπ‘–π‘›1β„Ž1𝑠

β„Ž2𝑠(1+π‘„π‘ β„Ž2𝑠)> 𝐴

12 (log(1 +π‘„π‘ β„Ž1𝑠) + log(1 + (π΄βˆ’π‘„π‘ 

β„Ž1𝑠

β„Ž2𝑠)β„Ž1𝑠)),

if 𝑄𝑠 < π‘₯2, 𝑄𝑠 < ��𝑠, 𝑄𝑠 < π‘βˆ—1 andπ‘„π‘ β„Ž1𝑠

β„Ž2𝑠+ π‘„π‘ β„Ž2𝑠

β„Ž1𝑠(1+π‘„π‘ β„Ž1𝑠)> 𝐴

12 (log(1 + π‘π‘œβ„Ž1𝑠) + log(1 + (π΄βˆ’ π‘π‘œ

β„Ž1𝑠

β„Ž2𝑠)β„Ž1𝑠)),

if π‘₯2 < 𝑄𝑠 and π‘₯2 > π΄β„Ž1𝑠

β„Ž2𝑠

or,𝑄𝑠 < π‘₯2, ��𝑠 < 𝑄𝑠, ��𝑠 < π‘βˆ—1 andοΏ½οΏ½π‘ β„Ž1𝑠

β„Ž2𝑠+ π‘„π‘ β„Ž2𝑠

β„Ž1𝑠(1+οΏ½οΏ½π‘ β„Ž1𝑠)> 𝐴

π‘…π‘€π΄πΆπ‘šπ‘Žπ‘₯ ,otherwise

(36)where π‘π‘œ = min((27), 𝑄𝑠). As (36) shows, 𝑄𝑠 may limit the

104

Page 5: [IEEE 2011 8th International Symposium on Wireless Communication Systems (ISWCS 2011) - Aachen, Germany (2011.11.6-2011.11.9)] 2011 8th International Symposium on Wireless Communication

βˆ’5 0 5 10 150.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

Qp (dB)

Ave

rage

sum

βˆ’ra

te (

bits

/sec

/Hz)

Fig. 4. Maximum achievable sum-rate in terms of the primary user’sconstraint.

maximum sum-rate. For the case π‘…π‘€π΄πΆπ‘šπ‘Žπ‘₯ > 𝑅𝐡𝐢

π‘šπ‘Žπ‘₯ we have:

π‘…π‘ π‘’π‘šπ‘šπ‘Žπ‘₯ =

⎧⎨⎩

π‘…π΅πΆπ‘šπ‘Žπ‘₯,if π‘βˆ—1 < min(𝑄𝑠, π‘₯𝑝) and π‘βˆ—2 ≀ 𝑄𝑠

12 (log(1 +π‘„π‘ β„Ž2𝑠) + log(1 + (π΄βˆ’π‘„π‘ 

β„Ž2𝑠

β„Ž1𝑠)β„Ž2𝑠)),

if π‘βˆ—1 < min(𝑄𝑠, π‘₯𝑝) and π‘βˆ—2 > 𝑄𝑠12 (log(1 + π‘π‘œβ„Ž1𝑠) + log(1 + (π΄βˆ’ π‘π‘œ

β„Ž1𝑠

β„Ž2𝑠)β„Ž1𝑠)),

if π‘₯𝑝 < 𝑄𝑠 and π‘βˆ—1 > π‘₯𝑝12 (log(1 +π‘„π‘ β„Ž1𝑠) + log(1 + (π΄βˆ’π‘„π‘ 

β„Ž1𝑠

β„Ž2𝑠)β„Ž1𝑠)),

if 𝑄𝑠 < π‘₯𝑝 and π‘βˆ—1 > 𝑄𝑠

(37)where π‘₯𝑝 is the abscissa of the intersection of Line 1 andLine2. The maximum achievable sum-rate for π‘₯2

π‘₯1< β„Ž1𝑠

β„Ž2𝑠can

be obtained similarly.

IV. SIMULATION RESULT

In this section we provide numerical simulations to illustratethe maximum achievable sum-rate of the considered TWRCsystem. All the channel gains have exponential distributionwith unit mean. The average sum-rate is obtained over1000000 channel realization. Fig.4 shows how the achievablesum-rate varies in terms of the allowed interference at theprimary user. The transmit power constraints at the sourcenodes and the relay are set to 10 dB. As the power constraintat the primary user gets more relaxed, the system sum-rateapproaches the sum-rate of a TWRC system in a non-sharingenvironment.

Fig.5 depicts the achievable sum-rate of the system versus𝑄𝑠 for the optimal power allocation (OPA) and equal power al-location (EPA). In EPA scheme, the source nodes are assignedequal power based on the available power region. The relayemploys the same strategy. The power constraints of the relayand the primary user are 10 dB and 0 dB, respectively. As itwas expected OPA leads to a superior sum-rate performance.The simulations confirm that relaxing a constraint dose notresult in a considerable sum-rate increase. It can be seen inthe Fig.5 that the upward trend of the sum-rate slows down

0 5 10 15 200.32

0.34

0.36

0.38

0.4

0.42

0.44

0.46

0.48

0.5

0.52

Qs (dB)

Ave

rage

sum

βˆ’ra

te (

bits

/sec

/Hz)

OPAEPA

Fig. 5. Comparison of achievable sum-rate for OPA and EPA.

after a while. In this situation, 𝑄𝑝 is the dominant constraintand drives the sum-rate.

V. CONCLUSION

In this research, the sum-rate performance of the TWRC ina spectrum-sharing environment has been studied. It has beenshown that the proposed power allocation improves the sum-rate comparing to the equal power allocation. Furthermore,the impact of the model constraints in achievable sum-ratehas been investigated.

REFERENCES

[1] B. Rankov and A. Wittneben, β€œSpectral efficient protocols for half-duplex fading relay channels,” IEEE J. Sel. Areas Commun., vol. 25,no. 2, pp. 379–389, Feb. 2007.

[2] β€”β€”, β€œAchievable rate regions for the two-way relay channel,” in Proc.IEEE Int. Symp. on Information Theory (ISIT), Jul. 2006, pp. 1668 –1672.

[3] S. Haykin, β€œCognitive radio: brain-empowered wireless communica-tions,” IEEE J. Sel. Areas Commun., vol. 23, no. 2, pp. 201–220, Feb.2005.

[4] K. Jitvanichphaibool, Y.-C. Liang, and R. Zhang, β€œBeamforming andpower control for multi-antenna cognitive two-way relaying,” in Proc.IEEE Wireless Communications and Networking Conference (WCNC),Apr. 2009, pp. 515–520.

[5] I. Hammerstrom, M. Kuhn, C. Esli, J. Zhao, A. Wittneben, and G. Bauch,β€œMIMO two-way relaying with transmit CSI at the relay,” in Proc. IEEESignal Processing Advances in Wireless Communications, (SPAWC), Jun.2007, pp. 1–5.

[6] Y. Wu, P. A. Chou, and S. Y. Kung, β€œInformation exchange in wirelessnetworks with network coding and physical-layer broadcast,” in Proc.39th Annual Conf. on Information Sciences and Systems (CISS), Mar.2005.

[7] T. J. Oechtering, C. Schnurr, I. Bjelakovic, and H. Boche, β€œBroadcastcapacity region of two-phase bidirectional relaying,” IEEE Trans. Inf.Theory, vol. 54, no. 1, pp. 454–458, Jan. 2008.

[8] R. F. Wyrembelski, T. J. Oechtering, and H. Boche, β€œDecode-and-Forward Strategies for Bidirectional Relaying,” in Proc. 19th AnnualIEEE Int. Symp. on Personal, Indoor and Mobile Radio Communications(PIMRC ’08), Sep. 2008, pp. 1–6.

[9] S. J. Kim, P. Mitran, and V. Tarokh, β€œPerformance bounds for bidirec-tional coded cooperation protocols,” IEEE Trans. Inf. Theory, vol. 54,no. 11, pp. 5235–5241, Nov. 2008.

[10] T. Cover and J. Thomas, Elements of Information Theory. New York:Wiley, 1991.

105


Top Related