Outage Analysis in SWIPT Enabled Cooperative AF/DF Relay Assisted Two-Way Spectrum Sharing Communication

This paper reports relative performance of decode-and-forward (DF) and amplify-and-forward (AF) relaying in a multi-antenna cooperative cognitive radio network (CCRN) that supports device-to-device (D2D) communications using spectrum sharing technique in cellular network. In this work, cellular system is considered as primary and internet of things devices (IoDs), engaged in D2D communications, are considered to be secondary system. The devices access the licensed spectrum by means of the cooperation in two-way primary communications. Furthermore, IoDs are energized by harvesting the energy from radio frequency (RF) signals, using simultaneous wireless information and power transfer (SWIPT) protocol. Closed form expressions of outage probability for both cellular and D2D communications are derived and the impact of various design parameters for both AF and DF relaying techniques are studied. Based on the simulation results, it is found that the proposed spectrum sharing protocol, for both DF relaying and AF relaying schemes, outperform another similar network architecture in terms of spectrum efficiency. It is also observed that the performance of the proposed system using DF relaying is better than AF relaying scheme in terms of energy efficiency at same transmit power.

Signal transmission between IoD 1 (SU 1 ) and IoD 2 (SU 2 ) Pp 1 , Pp 2 Transmission power of PU 1 , PU 2 h i Channel gain of link PU i → SU 1 (i ∈ 1, 2) g i Channel gain of link PU i → SU 2 (i ∈ 1, 2) D i Distance between PU i and SU 1 (i ∈ 1, 2) D j Distance between PU i and SU 2 (for i=1, j=4 and i=2, j=5 ) nsu, npu Received noise at SU, PU, respectively h 3 Channel gain of link SU 1 → SU 2 ρ Power splitting factor of SU 1 α Power sharing factor at SU 1 L Distance between PU 1 and PU 2 D 3 Distance between SU 1 and SU 2 m Path loss exponent η Energy conversion efficiency at SU 1 R P U , R SU Target rate of PU and SU communication, respectively considered to be an energy inefficient approach as noise gets amplified by the relay node. The performance improvement of SU communication over [13] is studied in [15] using bidirectional SU communication. Needless to mention that the presence of multiple antennas at the source and or the relay in any SWIPT enabled relay assisted communication over fading channel would not only improve the reliability of information transfer, but also enhances in energy harvesting at the relay node. Motivated by this, a preliminary study on the performance of the same system model [13] with multiple antenna PUs is presented in [16]. In this paper, a comparative study between DF and AF relaying scheme is presented using PS protocol in CCRN.

Scope and Contributions:
The work in [13] follows three-phase communication using DF relaying with a single-antenna in a CCRN framework. However, it does not explore the impact of multi-antenna on SE and EE aspects of the proposed D2D operation in 5G HetNets. Two-phase communication is more preferable over three-phase to enhance the system throughput. This leads us to explore the problem of spectrum sharing for bidirectional cellular communication and unidirectional D2D communication simultaneously. To offer improved SE, the present cellular system use multiple antennas, the trend seen on 5G [17]. To make this study more general, non-identical number of antennas are considered at two ends of the PU system. The nodes engaged in D2D communication are equipped with necessary hardware to harvest energy from the RF signals of the cellular users. EH meets the energy requirement of the transmitting node for D2D communication to relay (AF/DF) the signals of cellular users and also its own message transmission over the cellular spectrum simultaneously. The main objective of this work is to highlight the relative improvements in SE and EE performances of AF and DF relaying as the PU system moves from single antenna to multi-antenna system. Our contributions can be summarized as follows.
• A novel CCRN architecture with multi-antenna PU system is proposed where two-phase protocol supports two-way SWIPT enabled communications between the pair of cellular users (PUs) and also one-way D2D (SUto-SU) communication. Two fold benefits, the first one is the improved SE due to multi-antenna and the other one is the throughput improvement due to two-phase protocol are achieved.
• Closed form expressions of both PU and SU outage probability are derived for both DF and AF relay assisted communication using PSR protocol for multi-antenna CCRN framework. Simulation results closely match the analytical expressions.
• The exact dependence of the system performance on various parameters like power sharing factor, power splitting factor, transmission power is also shown through the simulation results. DF relaying mostly performs better than AF relaying. As we have used two-phase spectrum sharing protocol, therefore both of the relaying mechanisms show almost equally spectrum efficiency compared to the similar network architecture [13]. However, AF relaying is found to be better in terms of incremental improvement in SE with increase in the number of antennas.
The various symbols used are introduced in Table I. The remaining part of the paper is arranged as follows. The system model is described in Section II and communication protocol description is given in Section III. The outage performance of both PU and SU are analysed in Section IV. The necessary simulation and the numerical results in terms of the various system parameters is presented in Section V. Finally, the paper is concluded in Section VI.

II. SYSTEM MODEL A. Assumptions and Notations
The system model consists of a pair of eNB and UE in long term evolution (LTE) network architecture as depicted in Fig. 1. The eNB and UE are equipped with multiple antennas N a and N b , respectively. In absence of direct communication link, the cellular nodes (eNB and UE) intend to exchange their information via a single antenna IoT device (IoD) (denoted as IoD 1 in Fig. 1.), aiming to achieve the target rate of R P U at each side.
Simultaneously, IoD 1 sends its own message signal to IoD 2 to meet a target rate of R SU . Both of them use single antenna for communication. Here eNB and UE are considered as PU 1 and PU 2 , respectively and IoD 1 and IoD 2 are modelled as SU 1 and SU 2 , respectively. We assume that PU i , (i ∈ 1, 2) uses fixed power supply, i.e., P pi , SU 1 is powered through harvested energy from PUs' RF signals ,using SWIPT, for relaying PUs' messages and transmitting to SU 2 . We consider both DF and AF relaying mechanisms to support the two-way PU communications. 1 , g i,2 , ..., g i,Np ], (i ∈ 1,2) are the vector channel coefficients from the multiple-antenna PU i to SU 1 and PU i to SU 2 , respectively, where h i,n ∼ CN (0, 1) and g i,n ∼ CN (0, 1). Channel between SU 1 → SU 2 is represented by h 3 , where h 3 ∼ CN (0, 1). All the channel distribution between PU and SU links follow independent and identically distributed (i.i.d) Rayleigh fading. Due to short distance the link between SU 1 and SU 2 is considered as Nakagami-m fading. The instantaneous channel state information (CSI) is assumed to be unavailable at PU i (i ∈ 1, 2). It is also considered that the full-diversity space-time codes (like GABBA codes [18]) are used at the PU nodes. Normally, in space-time code, power is uniformly distributed among the transmitting antennas. The distances between the users PU 1 −SU 1 , PU 2 −SU 1 , PU 1 −SU 2 , PU 2 −SU 2 and SU 1 −SU 2 are given by D 1 , D 2 , D 4 , D 5 and D 3 , respectively with 'm'as the path-loss exponent. Here n su and n pu indicate the received noise at SU and PU, respectively. These noises are additive white Gaussian noise (AWGN) with zero (0) mean and the variance σ 2 .

B. Protocol description
The two-way communications between PU 1 and PU 2 via SU 1 take place in two time phases viz., multiple access channel (MAC) and broadcast (BC) phase, as shown in Fig. 2. In the MAC phase, both PU 1 and PU 2 simultaneously transmit their information signals to SU 1 . SU 1 is able to harvest energy from part of the received signal using PSR protocol. In the case of AF relaying, SU 1 broadcasts an amplified PU signal superimposed with the SU signal X s in the BC phase, using the total harvested energy. The PU receivers are able to receive the signals using the maximal-ratio-combining (MRC) and separate the desired signal using the self-interference cancellation (SIC). Now, decoding of SU 1 's message at SU 2 is performed in the two phases; first the stronger PU signal is decoded considering SU 1 's desired signal as interference. Once the strong PU signal is separated from the received signal in the the first phase, SU 2 decodes the desired SU 1 's signal in presence of channel noise.
In the case of DF relaying, SU 1 first decodes the PUs' message signals received in the MAC phase, and then broadcasts a network coded primary signal superimposed on the secondary signal X s in the BC phase. Both the PU nodes decode the network coded (XOR operation) PU signal in presence of SU interfering signal and noise. Thereafter, SU 2 decodes SU 1 signal by cancelling PU signal's based on the PUs' messages received previously in the MAC Phase.

A. Rate Analysis
Now the received power in MAC phase at SU 1 from PU 1 and PU 2 can be expressed as follows 1 [19].
where P p1 and P p2 are the transmit powers of PU 1 and PU 2 , respectively. Similarly, the received power at SU 2 from PU 1 and PU 2 can be expressed as A part ρ (referred to as power splitting factor in the subsequent discussion) of the received signal power at SU 1 is used for energy harvesting, and the rest (1-ρ) portion is used for information processing. The harvested energy in MAC phase is given by where 0 < η < 1 represents the energy conversion efficiency. After power splitting, the received power in MAC phase at the receiver of SU 1 for information processing is given by Now following the linear EH model, the available power at SU 1 for transmission in the BC phase is given by In the MAC phase, the successful decoding of the received signals from PU 1 and PU 2 is possible at Similarly, in the MAC phase, the successful decoding of the received signals from PU 1 and PU 2 is possible In DF relaying, SU 1 uses α (0 < α < 1) fraction of its total transmit power P s to relay the network-coded primary information and rest (1-α) fraction of P s is used to send its own independent message X s to SU 2 . After MRC and cancellation of self-interference terms by applying SIC technique, the received signal-tointerference-noise ratio (SINR) at PU i (i ∈ 1,2; N p ∈ N a , N b ) receivers in BC phase can be expressed as It is assumed that SU 2 , like SU 1 , succeeds in decoding the PUs' signals received in MAC phase. Based on these prior knowledges, SU 2 is able to separate the desired SU signal from the PUs' interference received in BC phase [13]. Therefore, the received signal at SU 2 can be rewritten as Now, the achievable rate at P U i is given by (based on the SINR at P U i ) (w ∈ n, p) where a 1 = ηρ Based on (9), the achievable rate at SU 2 is expressed as

B. Outage Probability Analysis
An outage occurs when the achievable rate of data transmission on any transmission link falls below the target rate of data transmission.
1) Outage Probability Analysis of Primary System: Based on the definition, PU outage probability 2 using DF relaying mechanism can be determined as follows [13]: Applying (6), the success probability of data transmission between both of the PU nodes (PU 1 and PU 2 ) and SU 1 is expressed by following three conditions satisfied together. where Proof : See Appendix A.
The probability of successful data transmission from SU 1 to PU 1 can be written as . Now, the solution to (16) is obtained as (17) Similarly, the probability of successful data transmission from SU1 to PU2 can be written as The closed form solution to PU outage probability can be determined using (14), (17)- (18). Closed form expression of PU outage probability using DF relaying is given in Appendix F.
2) Outage Probability Analysis of Secondary System: SU outage probability 3 using DF relaying is shown as follows: Similar to (13), the probability of success due to data transmission from PUi (i ∈ 1, 2) to SU2 is defined by the following three conditions together.
p=1 | g2,p | 2 follow the same nature of gamma distribution like X1 and Y1, respectively.
Based on (20), P{Q2} can be expressed as follows Following (11), the probability of successful data transmission between the link SU1 to SU2 is expressed as where u4 = 2 (2R SU ) − 1. Z =| h3 | 2 and it follows the gamma distribution with Nakagami shaping parameter m k . The PDF of Z can be described as fZ (z) = 3 Since SU 2 needs to decode the message of SU 1 by removing PUs message in BC phase, therefore successful information transmission from both PUs to SU 1 is necessary in MAC phase. Now (22) can be evaluated as follows [21,Sec. 3.471.9,8.352.2]: Proof : See Appendix C. Finally, the closed form SU outage expression using DF relaying can be determined using (14), (21) and (23). Closed form expression of SU outage probability using DF relaying is given in Appendix G.

IV. RATE & OUTAGE ANALYSIS OF AF RELAYING A. Rate Analysis
In the BC phase of AF relaying, SU1 broadcasts an amplified version of the signal generated after combining the PU signals, received in the MAC phase. To include this in the analysis, harvested power Ps in (5) is normalized using the factor ξ 4 , expressed as Based on the superposition coding and AF relaying principle SU1 uses α (0 < α < 1) the fraction of its total transmit power Ps to relay the combined signal of primary information and the rest (1-α) portion is used to send its own independent message Xs to SU2. As both PUs know their individual transmitted signals, consequently, they are able to cancel their self-interference terms. Therefore, the instantaneous end-to-end SINR at PUi can be expressed using (25).
Similarly, SU2 is able to detect PU signal by considering SU signal as an interference [22]. The instantaneous SINR can be expressed applying (26).
The PU signals being a strong one, SU2 first decodes PU signals considering SU1 signal as interference and removes it from the received signal. Then it decodes its own signal in the presence of noise. The instantaneous SNR at SU2 can be expressed in (27).
Achievable rate at P Ui is given by (based on the SINR) The achievable rate for PU information decoding at SU2 is as The achievable rate at SU2 is given by (based on the SNR at SU2) 4 As noise power is negligible with respect to the power received for information processing, therefore noise power is neglected.
B. Outage Probability Analysis 1) Outage Probability of Primary System: PU outage probability using AF relaying is determined as follows: The symbol Kv(.) indicates the modified Bessel function. Similarly, it is also possible to determine the probability of successful data transmission between the link SU1 to PU2.
Closed form expression of PU outage probability using AF relaying is given in Appendix H.
2) Outage Probability Analysis of Secondary System: SU outage probability, using AF relaying mechanism, can be determined as follows: Proof : See Appendix E. Similar to (34), the probability of success for the decoding of the PU information in presence of SU interference and noise can be written as . Finally, the closed form SU outage expression using AF relaying can be determined using (34) and (35). Closed form expression of SU outage probability using AF relaying is given in Appendix M.

V. NUMERICAL RESULTS AND DISCUSSIONS
Numerical values for the system parameters used are enlisted in Table II. The constraint of the maximum PU outage limit (say 0.1) is found to be met at α=0.81 by both AF and DF relaying for Na=2. This value of α is used for the results shown in Fig. 4, Fig. 5 and Fig. 6. Considering the practicability of the number of transmitting antennas on mobile devices we set N b =1 in our simulation results.
The outage performances of both PU and SU systems are shown as variation in power-sharing factor (α) for both DF and AF relaying in Fig. 3.a and Fig. 3.b. It is clearly seen that the analytical results match perfectly with the simulation results. For the small value of α, the higher fraction of the harvesting power is allocated to SU information transmission and small fraction of power is used for DF relaying of PU information. Thus SU outage is found to be low while PU outage to be high. The reverse situation is to observed at high value of α. The small value of α (close to 0) causes PU outage of AF relaying same to that of DF relaying and SU performance is found to be the worst by following (33)-(35). Since 50% or more power is assigned to PU information transmission to detect PU signals perfectly in presence of SU interference, therefore the minimum value of α is shown as 0.5 for AF relaying in Fig. 3.b. The threshold limit of 10% PU outage is achieved at α=0.29 and Na=1 for DF relaying and the same outage limit is achieved at α=0.89 and Na=1 for AF relaying. If the number of antennas is increased then PU performance is improved accordingly. If Na is increased from 1 to 2, about 46% improvement (reduction) in PU outage is observed for DF relaying and about 39% improvement of PU outage is observed for AF relaying. If Na is increased from 1 to 2, about 58% improvement of SU outage is observed at α=0.27 and RSU =1 bps for DF relaying and about 42% improvement of SU outage is observed at α=0.81 and RSU =1 bps for AF relaying. Fig. 4. and 5. show the outage performances of PU and SU, respectively with respect to power splitting factor (ρ) for both DF and AF relaying mechanisms. As depicted in the figures, the outage performances of both PU and SU are very poor for ρ → 0 and ρ → 1. Initially, the performances of PU and SU outage are improved with the increase in the value of ρ, and they attain their minimum values of outage at the optimal value of ρ * . Thereafter, when the value of ρ is increased further, it leads to an increase in both PU and SU outage. The reason behind the characteristics of this graphical plot can be explained as follows. Initially, when ρ value is very low, the energy harvested at SU1 is insufficient to broadcast the information at the BC phase and effectively the outage performance of both PU and SU are very poor. When the value of ρ increases, the harvesting energy at SU1 is also increased and consequently both PU and SU outage performances improve accordingly. Further degradation on the outage performances are found due to major power allocation for EH and less power allocation for decoding the information received from PU to SU signal transmission. Performance of PU outage is significantly improved for DF relaying as compared to AF relaying mechanism whereas the SU outage performances are almost same for both the relaying mechanisms with respect to ρ. The overall outage performances for both PU and SU are improved with the increase in the number of antennas used at PU nodes. Now SE and EE of the system can be defined as [13] EE can be defined as the ratio of SE to the total energy consumed for PUs transmission [23]. Value PU target rate of information transmission (R P U ) 0.2 bps/Hz SU target rate of information transmission (R SU ) 1 bps/Hz Distance between PU 1 and PU 2 (L) 20 m Distance between SU 1 and SU 2 (D 3 ) 10 m PU transmit power (Pp 1 =Pp 2 =Pp) -23 dBm (Fig. 3 to Fig. 5), -50 dBm to -15 dBm (Fig. 6) Average noise power (npu=nsu=np) -100 dBm [11] Path loss exponent (m) 2.7 Energy conversion efficiency (η) 0.9 Power splitting factor (ρ) 0.9 Power sharing factor (α) 0.81 The SE and EE performances of the DF and AF relaying protocols are shown with respect to the total PU transmit power 5 in Fig. 6.a and Fig. 6.b, respectively. The performance of the proposed system using AF and DF relaying schemes are compared with similar system supporting two-way PU and one-way SU communication simultaneously [13]. As shown in figure, SE is improved and EE is deteriorated with the increase in PU transmit power. The characteristics of this graphical plot may be explained as follows. As transmission power of PUs are increased, the PU and SU outage probabilities are decreased accordingly following (14)- (18), (21), (23), (32), (34) and (35) for DF and AF relaying mechanisms, respectively. As inverse relationship is maintained between SE and the outage probability of both PU and SU system, the SE is improved with the increase in the transmission power of PU. Less increment of SE as compared to more energy consumption leads to a degradation of EE with the increasing value of PUs transmit power. Our proposed system using DF relaying mechanism is more efficient as compared to AF relaying mechanism and similar model of two-way PU and one-way SU communication.
In terms of SE, our proposed system using DF relaying mechanism is performing 224%, 196% and 142% better as compared to [13] for Na=3, Na=2 and Na=1, respectively at 52 dB transmit power. Our system using AF relaying is performing 34%, 31% and 22% better in terms of SE as compared to [13] for given value of Na=3, Na=2 and Na=1, respectively at 70 dB transmit power. In terms of EE, our proposed system using DF relaying mechanism is performing 62%, 48% and 21% better as compared to [13] and on the other hand, the performance of [13] is 56%, 60% and 69% better as compared to the proposed system using AF relaying for given value of Na=3, Na=2 and Na=1, respectively at 52 dB transmit power. However, the proposed model is less energy efficient at high transmit power as compared to [13].

VI. CONCLUSIONS
This paper has investigated the scope of SWIPT enabled IoT communication, on the licensed spectrum using overlay spectrum sharing mode of cognitive radio networks. Transmitting IoD node harvests energy from the information bearing RF signals of both eNB and UE transmission and this energy is used in relaying between eNB and UE. Apart from the close match between the analytical and simulation results on outage experiences in IoD and cellular systems, it is also to be noted that DF relaying mechanism is more efficient over AF relaying in terms of EE and AF relaying is more sensitive to the impact of the number of antennas used by the cellular system. The proposed network architecture and the spectrum sharing model can be extended further (i) to analyse the outage for more realistic non-linear RF-EH model, (ii) to improve the secrecy of the relay assisted cellular communication using the PUs' signals for friendly jamming in addition to EH and (iii) to study game theoretic modelling of the possible negotiations between cellular users and multiple IoT device pairs for efficient trading of the available radio resources of the former. x
pa pa! dx1 dz + I (Using Binomial Series Expansion, we can write as follows) (1/Na) Na Na−