I Introduction
Wireless powered communication networks (WPCNs) are one of the promising technologies to achieve sustainable wireless networks, where the communicating nodes are powered by radio frequency (RF) signals. The simultaneous wireless information and power transfer (SWIPT) concept has been investigated extensively to realize WPCNs, e.g., authors in [1, 2] survey recent works in energy harvesting communication domain. In SWIPT, energy and information are transferred from the same RF signal by using either time sharing or power splitting protocol [3, 4, 5]. The time sharing protocol allocates dedicated time for energy harvesting and information transfer, while power splitting extracts energy and information from the same RF signal.
Relays are used in wireless networks to extend the coverage and increase the information reliability. Relay selection problem using amplifyandforward (AF) and decodeandforward (DF) techniques in WPCNs has been addressed in literature quite extensively, e.g., [6, 7]. Similar to the traditional relay selection schemes with fixed power supply, battery status information (BSI) enabled relay selection schemes are proposed to improve the system performance of multirelay wireless powered cooperative networks by exploiting both channel state information (CSI) and BSI to make the selection decision [8, 9]. In such systems, the relaying nodes, with BSI indicating the amount of stored energy above a predetermined power threshold required for successful information transmission, will first create a subset and then will send their CSI back to the source node. By the next step, the ’best’ relay from the subset will be chosen to forward the source information, while the remaining relays will harvest energy from the source information signal.
In literature, there are three main relaying architectures for SWIPT systems, namely, ideal relaying receiver (IRR), powersplitting relaying (PSR) and timeswitching relaying (TSR) protocols [4, 6, 10, 7, 11, 12, 13, 14]. A recent work in [7] discusses an RF based cooperative network, where the relays are used for transmitting information to a designated receiver and for transmitting energy to an associated ambient RF energy harvester. For the case where the number of relays is more than two, two relay selection methods are developed and the tradeoff between outage probability and average energy transfer is discussed. The authors in [11] analyze the performance of a network consisting of a single source, single relay and single destination. The throughput of the TSR protocol for both AF and DF relaying schemes for the same model is also investigated in [14]. A similar system model is considered in [15], where the relay is equipped with multiple antennas and the outage probability and ergodic capacity of the system are studied. In [12, 16], the authors investigate the ergodic capacity and outage probability over fading channels of an AFbased network consisting of the source, relay and destination nodes for the IRR, PSR and TSR protocols. The outage probability analysis in twohop DF and AF relaying systems over lognormal fading channels is provided in [17, 18], respectively. Another work on the outage is presented in [19], where the authors study multirelay wireless powered cooperative systems over Nakagami fading channels.
In [13], the authors analyze the outage performance of the wireless powered fullduplex (FD) AF and DF relaying networks in environment. Furthermore, the outage in twoway (TW) FD relaying networks with multiple pairs of users is investigated in [20]. A similar model with multiple sourcedestination pairs communicating through a single energy harvesting relay is considered in [21] and the system effect of the harvested energy distribution among the users is investigated. In [22], the authors analyze outage performance obtainable under a decentralized relay selection strategy in multiuser multihop DF based relaying networks over Nakagami fading channels. Multihop relaying for a cognitive radio network is investigated in [23], where the authors aim at minimizing endtoend outage probability for a secondary user under the energy causality and primary user cooperation rate constraints. The system outage probability is investigated in [24], where the authors propose optimal and suboptimal joint relayantenna selection schemes for TW AF relaying networks. Moreover, the authors in [25] introduce a general relay selection strategy for the PSRenabled TW FD relaying network, where average sum capacity and outage probability are studied. Other performance metrics such as energy efficiency and security issues in wireless power transfer enabled relaying networks are studied in [26, 27, 28, 29]. For instance, the authors in [27, 28] propose relay selection schemes to improve secrecy outage probability in cooperative DF and AF relaying networks under the presence of eavesdroppers. In [29], joint relay selection and power allocation scheme is proposed for largescale multipleinput multipleoutput AFbased relaying systems with passive eavesdroppers. The authors provide closedform expressions for ergodic secrecy rate and secrecy outage probability over Rayleigh fading channels.
Ia Motivation and Contributions
In this work, we aim to minimize the outage probability for a cooperative system, comprising of a single source, multiple energy harvesting (EH) relays and a single destination. A single relay is selected to forward the source signal to the destination. We assume half duplex relay communication such that the relays receive the source signal in a time slot and the selected relay forwards it to the destination in time slot . The channels on both the sourcerelay and the relaydestination links are mutually independent, and are independently and identically distributed (i.i.d.). Due to mutual independence and i.i.d. channel assumption on both links, relay selection poses new challenges as the relay selected to receive data from the source in time slot will have a completely independent (and unknown) channel realization in time slot for transmission on the relaydestination link.
For a similar setting, the work in [30, 31] assumes that the CSI is not available on the sourcerelay link at the relay node. In contrast, we assume availability of the CSI on the sourcerelay link at the relay node throughout this work. When the CSI is available on the sourcerelay link, the relay selection exploits the CSI to decide which relays are dedicated for data/energy transfer [32]. Then, conditioned on the availability of transmit CSI (CSIT) at relay on the relaydestination link, we propose novel relay selection schemes. Furthermore, we derive closedform analytical expressions of the outage probability which are unified in terms of the amount of the harvested power. The consideration of the availability of both transmit and receive CSI at the relay node requires a different relay selection approach as compared to relay selection in [31]. We formulate the outage minimization problem and evaluate the performance of the proposed relay selection schemes numerically. Then, we compare our schemes with the existing schemes available in literature and show their superiority in terms of outage performance.
Ii System Model
Consider a twohop DF wireless communication system where a source node transmits information to a destination node by means of one relaying node selected from available energylimited relay nodes (a relay node is denoted by ). Moreover, due to the apartness between and , we assume that there is no direct link between them. We consider a broadcast channel between the source and the relays and assume that all nodes are equipped with a single antenna. The sourcetorelay () and relaytodestination () links, indicated by and , are subject to quasistatic i.i.d. Rayleigh fading. The and distances are denoted by and , respectively; the corresponding pathloss exponent is chosen to be identical for all the links.
The CSIT is not available at the source and, therefore, the source transmits with a fixed power . The received signal at the relay is expressed as,
(1) 
where , and
denote the source transmit power, the normalized information signal and the Gaussian noise with zero mean and variance
, respectively.We make the following assumptions regarding the cooperative system.

We assume that the fading coefficients remain constant for the duration of a time slot, but vary independently from one slot to another. The schematic diagram for the system model is shown in Fig. 1 with a centralized controller, which collects stored energy and channel state information from all the relays and makes relay selection in every time slot.

We use time sharing protocol for SWIPT at relay nodes.

The transmissions are interference free and interference cannot be exploited for energy harvesting at relays.

We consider a half duplex communication system. The selected relay node decodes information from the source in a time slot of duration and forwards it to the destination in the next time slot. Hence, node is not available for information (or energy) reception in time slot . However, all other nodes are available for information/energy reception from the source signals, thereby mimicking a FD communication system [33].

A single relay is selected for forwarding information to the destination. It is well known that transmission from multiple relays provides spatial diversity and improves data reliability at the cost of increased complexity. We focus on single relay selection schemes to reduce the complexity of the system, but all the proposed schemes can be extended to multiple relay transmission schemes in a straight forward manner.

The circuit energy consumed in energy harvesting and information decoding at relay is negligible. However, if circuit energy consumption is not negligible but same for all relay nodes, it will not have any effect on the design of proposed relaying schemes as all the relay nodes behave symmetrically. Therefore, circuit power consumption can be neglected in this problem without affecting problem and solution design.

We assume that the relay has no external power supply (i.e., powered by energyharvesting from sourcetransmitted signal) while being deployed with a battery of infinite capacity. It is also assumed that there is negligible leakage within the time period of interest [34].
The rate provided on the link in a time slot is given by
(2) 
where the factor shows that the to transmission requires two time slots. For a relay transmit power , the rate provided by the link is given by,
(3) 
where denotes the channel coefficient between the relay node and the destination D.
All the nodes selected for energy transfer harvest energy from the source signal. Assuming and linear energy harvesting model, without loss of generality, the energy harvested by a relay node is given by
(4) 
where is the energy harvesting efficiency. In recent literature, nonlinear energy harvesting models have been used for various studies, e.g., [35, 36], which account for nonlinear effects in RF energy harvesting in practical systems. For simplicity, we assume a linear model here to focus more on the impact of CSI assumptions on system outage performance.
Iia Problem Formulation
For a traditional grid powered DF relaying strategy, the outage probability that a rate is not supported by the system is given by
(5)  
(6) 
where is the selected node for information relaying. However, when the relays are powered by EH, unavailability of the harvested energy is an additional source of outage.
For a wireless powered relay network, the outage probability is given by,
(7) 
where and denote the success probability on the and links, respectively while is the probability that the selected relay can support a transmit power to forward the information to the destination. depends on the relay selection scheme and energy harvesting efficiency . If there is always sufficient energy available for decoding, and the EH system behaves as a grid powered system.
Our objective is to find a relay selection policy that minimizes the outage probability for the EH cooperative system with a fixed number of relays. The optimization problem is formulated as:
(8)  
(9) 
where and are constants representing a fixed number of relay nodes in the system and the fixed source power, respectively. is the rate condition. is neutrality constraint which implies that stored energy , for the node selected for forwarding , must be greater than the energy required for transmission. We intend to find a relay selection policy which follows the constraints in (9) and minimizes network failure (outage probability).
A closedform solution for the problem is difficult to achieve due to involvement of multiple relays and multiple energy queues at the relay nodes, which depend on stochastic fading channels. The energy queue states are mutually coupled for the outage analysis, making computation of in (7
) difficult, and the analysis is not tractable for a large number of relays. Therefore, we propose heuristic relay selection schemes and evaluate the outage performance numerically. To give insight, we provide outage performance analysis for a certain relay
depending on the number of energy queues.Iii Relay Selection Schemes
To avoid the outage event, there must be at least a single relay available with sufficient energy to transfer data to the destination in time slot . To minimize the outage probability, the relay selection scheme should aim at maximizing energy harvesting of the relays and minimizing the energy expenditure on the link, thereby maximizing the network lifetime with minimum failure.
A simplified approach to model such a system is to assume the length of fading blocks long enough such that the channels on the and links remain constant for both reception and forwarding phases at the relay [37, 38]. This has the advantage that both receive and transmit channels are known at the time of relay selection. In contrast, we assume that the relay reception and transmission occurs in two consecutive time slots. Due to mutually independent channels on the and links, and the fact that the CSI for the channel is not available when a relay is selected for forwarding at time , relay selection becomes challenging.
The relay selection scheme is dictated by the availability of the CSI at relay on the both and links. Throughout this work, we assume that the CSI is available at the relay on the link^{1}^{1}1
The CSI estimation at the relay can be performed by pilot/data aided techniques.
.Regarding the availability of the CSI at relay node on the link, we consider the following two scenarios:

The CSIT is not available on the link.

The CSIT is available on the link.
Iiia No CSIT on the Link
When the CSIT is not available on the link, no power allocation can be performed. Therefore, the selected relay transmits with a fixed power at time .
At time , the relay selection is performed. As the CSIT on link is not available, relay selection is solely based on the available information at time . As a single relay is selected at time , this scheme is called single relay selection with No CSIT (SRSNCSI).
A relay is selected for decoding information such that,
(10) 
where (10) is evaluated for a relay only if,
(11) 
such that,
(12) 
and
(13) 
denotes the stored energy for relay . The indicator functions and in (11) ensure that a selected node can decode the signal from the source and has energy to transmit with a fixed power in time slot . Equation (10) selects the node with the minimum for information decoding out of the nodes which satisfy (11). The rationale behind the selection of the node with minimum is to provide relatively ’average’ channel for information decoding at relay as information decoding is already ensured by the condition . This implies that good channels (which satisfy (11)) can be better utilized for energy harvesting as decoding on the best channel does not improve the outage performance as long as (12) is satisfied. If for the selected relay or no relay satisfies (11), an outage event occurs.
All the relays with harvest energy from the source signal such that,
(14) 
The selected relay node is not a candidate for selection for both decoding and harvesting from the source signal in time slot , which implies that and the stored energy for node after making a transmission at time is given by,
(15) 
It is clear from (6) that is determined by the rate provided by the ’bottleneck’ link. However, the outage probability in WPCN is also characterized by the amount of energy harvested by the relay nodes. The harvested energy is a function of the source power, channel distribution and the energy harvesting efficiency . When is large, very small number of relay nodes provide enough stored energy such that there is always a node available with enough energy to transmit on the link and the outage probability in (7) converges to (6). However, when or is small, (6) is only a lower bound on .
It is worthwhile pointing out that the results on outage performance associated with (10)  (13) will take into account the number of relays and randomness of each channel realization of all possible and links to calculate the amount of harvested energy. However, it is intractable to consider these features in the outage performance analysis since it is not feasible to capture a dynamic behavior of the EH and communication activities in one generalized analytical framework. Hence, our analysis will be performed for a relay at certain time slot (which can be regarded as a static process with initial inputs on the stored power available for successful endtoend communication) with various levels of the harvested power available for information transmission.
Proposition 1.
With respect to the defined relay selection strategy, the outage probability given in (7) can be rewritten in its closedform as in (16), shown at the top of the current page. The generalized expression of the outage probability provided here considers not only the number of time slots allocated for EH purposes () but also how many times a relay was chosen for data transmission ().
Proof.
See Appendix A.
(16) 
IiiB CSIT Available on the Link
In the case when the CSIT is available at the relay on the link, the relay can benefit from this information through power allocation. CSI can be made available at relay node either by using data aided estimation or making use of explicit training sequences. As we show later in this section, this helps improve system outage performance and overhead in channel estimation is justified. This feature is commonly used in 4G and 5G systems, especially when there are no latency constraints.
The required transmit power to make a successful transmission for a relay is computed from (3), and given by,
(17) 
This scenario provides more flexibility for relay selection. However, the CSIT on the link is available only at time due to i.i.d. channel assumption and the relay is selected at time .
The main challenges in relay selection are:

If the relay selection is made based on the channel quality on the link, the selected relay node may not have enough energy to transmit on the link at time . At the same time, the use of channel quality on the channel will not be optimal as the selected relay may not necessarily have the best channel at time .

If the relay selection is based on the stored energy maximization, the availability of the CSI on and links at the relay node is not exploited.
To take the advantage of CSI availability at different times, we propose a two step relay selection algorithm.
Phase I: In the first phase, a subset of (maximum) relays^{2}^{2}2To avoid confusing it with multiple relay forwarding, please note that only one relay will be selected for forwarding after phase 2 of the scheme. is selected out of relays for decoding information. As the CSIT on the link is not known at time , more than one relay decode information to provide multiuser diversity for the transmission on the link. It is worth noting that selecting a single relay in phase 1 makes available CSIT on the link at time useless as all other relays cannot be used for forwarding in phase 2 of the scheme. Due to multiple relay selection in first phase, this scheme is termed as Multiple Relay Selection with available CSI (MRSACSI).
The selection for the forwarding set is made such that,
(18) 
where (18) is evaluated for a relay only if,
(19) 
denotes the fading channel with smallest amplitude, selected out of nodes satisfying . Cardinality of is limited by , where is a system parameter for the scheme. Equation (18) states that relays with the smallest fading channels are dedicated for decoding information.
This metric chooses the relay nodes with the weakest channels, but capable of decoding the information. This implies that the rest of the relays harvest energy from the source signals on good channels and the stored energy for the nodes increases at a faster rate.
Phase II: In the second phase of the relay selection algorithm, the forwarding relay from the set at time is selected such that,
(20) 
where (20) is evaluated for a relay only if,
(21) 
such that,
(22) 
The scheme selects the relay with the best transmit channel out of the relays, which have enough stored energy for transmission as in constraint (21). This ensures transmission with minimum expenditure and is the optimal decision for the relays in . Note that is calculated individually for every relay via (17). If the cardinality of set is zero or no relay in the set satisfies (21), outage occurs. The pseudocode for the second phase of the relay selection algorithm is presented in Algorithm 1.
After transmission, the stored energy for the node is updated as,
(23) 
The rest of the nodes harvest and store energy depending on the received signal strength from the source such that
(24) 
with the nodes not able to harvest energy as they were reserved for decoding.
(25) 
Proposition 2.
According to the chosen relay selection strategy, the outage probability given in (7), when CSIT is available at the relay, can be expressed in its closedform as in (25), shown at the bottom of the next page. The generalized solution for the outage probability provides an insight into how the number of time slots allocated for EH purposes () as well as the number of time slots devoted for data transmission () will affect the outage performance of the communication system under consideration.
Proof.
See Appendix B.
There is a tradeoff involved with the selection of parameter for a fixed . If is large, there is greater chance of finding a good channel for transmission on the link, but fewer relays are available for EH and the relay system becomes power limited. On the contrary, if is too small, less multiuser diversity is exploited on the link, but more relays harvest energy. Thus, for the proposed scheme, it is important to optimize for a given and .
Iv Numerical Results
We numerically evaluate the performance of the proposed schemes in this section. A Rayleigh fading channel with mean one is considered on the and links. Time slot is assumed to be one while noise variance . is fixed to dBW throughout.
In Fig. 2, we compute the outage probability for the SRSNCSI scheme for different number of available relay nodes. The CSIT is not available at the relay node on the link and the selected relay node transmits with a fixed power 10 dBW. The outage probability decreases as the number of relays increases, as expected. However, when is sufficiently large, any further increase in does not benefit. For a small , there is high probability that the selected relay stored energy is not enough to transmit successfully on the channel and ’power limitation’ of the relay node contributes to the outage significantly. As increases, the outage performance improves. At , the effect of power limitation vanishes completely and does not help to decrease outage. The system behaves like a grid powered system and the outage performance is given by (6).
In Fig. 3, we demonstrate the probability that a certain relay will be selected for information transmission for the SRSNCSI scheme (i.e., the CSIT is not available at the relay node on the link and the selected relay node transmits with a fixed power dBW) for different amounts of the harvested energy stored in the battery. The amount of energy is related to the number of time slots dedicated to EH purposes and to the number of time slots when this relay is acting as a communication node . To investigate the number of time slots dedicated for EH, we assume , i.e., the relay has not been selected for information transmission yet. This assumption is reasonable for the case when the number of available relays is relatively large. The case means that the communication is just initiated and no power is available at the relays, and they operate in the EH mode only. It can be also observed that the probability of relay selection increases when the amount of the stored energy increases, as expected. However, further increase of the stored energy does not contribute to the probability of relay selection significantly, i.e., it starts saturating after . Next, for the case when (a relay has been selected once), the selection probability severely deteriorates when the battery charge is low which, in turn, corresponds to the outage probability degradation shown in Fig. 4.
Next, we compare SRSNCSI scheme with other similar available schemes. As a benchmark, we consider two commonly used schemes. In the first scheme, the relay is selected such that [31, 37, 6],
(27) 
with the notation . We denote it by SRSNCSIbestenergy scheme, where the relay with the largest residual energy is selected for transmission.
The second scheme selects the relay which has the best chance of decoding on the channel. The concept is similar to relay antenna selection scheme in [39], where the antennas with large channel gains are selected for decoding. Thus,
(28) 
We denote this scheme by SRSNCSIbestdecoding. Please note that the forwarding in all schemes is made only if , which saves transmit energy on unsuccessful transmission.
From Fig. 5, we see that SRSNCSI outperforms the other schemes. The SRSNCSIbestenergy performs better than SRSNCSIbestdecoding because the major cause of outage is insufficient energy to forward data for the selected relay. The best channel selection on the link is not optimal as decoding is already ensured by the condition .
In Fig. 6, we compute the outage probability of the selected relay for the SRS scheme and various when CSIT on the link is available at the relay node which transmits with . The outage performance is shown versus the data rate threshold for . In the case when CSIT is available, the outage probability outperforms that of the case when a relay is not aware of the channel properties of the link. It is clear that the performance deteriorates when is increased, as expected. It is worthwhile pointing out that the result of the scenario with available CSIT always outperforms one when no CSIT on the link is available at the relay (see Fig. 4 for comparison).
In Fig. 7, we plot the outage probability for the MRSACSI scheme and compute the optimal value of parameter for a given and . If is too small, multiuser diversity is not exploited effectively on the link. On the contrary, if is too large, EH is not enough for the relays to store enough energy to avoid outage events. We observe that is optimal at small , while is optimal at large . This is attributed to the fact that large rate requirements require more relay nodes to harvest energy to have sufficient energy for successful transmissions to the destination.
In Fig. 8, we compare the performance of the MRS scheme with the two schemes mentioned above. We assume that the CSIT is available at relay on the link and power is allocated for the selected relay by (17). However, due to unavailability of at time , we eliminate transmit power term from (27) and evaluate the metric,
(29) 
It is worth noting that the relay is not available for harvesting at time slot even if because can only be calculated at instant due to delayed CSIT on the link. Fig. 8 shows that power allocation due to available CSIT on the channel at time improves the outage performance for the SRSNCSIbestenergy and SRSNCSIbestdecoding schemes as compared to their respective performance in Fig. 5, but MRSACSI outperforms both schemes comfortably due to inherent multiuser diversity exploitation.
To demonstrate the effect of relay selection metrics in (18) and (20), we compare MRSACSI scheme with a similar 2phase relay selection scheme proposed in [31]. Like MRSACSI, relays with the largest stored energies are selected in first phase. In the second phase, a relay out of relays is selected such that,
(30) 
This scheme is denoted by MRSACSIbestenergy. For the MRSACSIbestenergy scheme, for the parameters [31]. The results in Fig. 9 reveal that our scheme performs better that the MRSACSIbestenergy scheme in spite of the fact that both schemes are exploiting multiuser diversity. From the numerical evaluation in Fig. 8 and Fig. 9, we conclude that based on exploitation of multiuser diversity and careful design metrics for both phases as explained in Section IIIB, our proposed MRSACSI scheme performs better that the other schemes available in literature.
V Conclusions
We propose novel relay selection schemes for the WPCNs and discuss the scenarios where the CSI is available at relay on the link. At the time of selection of relay at the link, the unavailability of any information on CSI at the link is a very practical scenario as compared to ideal scenarios where CSI availability for the fading channels is assumed throughout at the time of relay selection at the source. Conditioned on the availability of the CSIT on the link at time slot , we propose two heuristic relay selection schemes to minimize system outage as the problem is not tractable due to unavailability relay power at the time of relay selection. To provide insight, we provide closedform analytical expressions of the outage probability for both scenarios. Next, we evaluate the performance of the proposed schemes numerically and compare it with the commonly used relay selection schemes. When the CSI is available at the relay on both and links, consideration of mutually independent i.i.d. channels on two hops makes the half duplex relay selection problem challenging. A two–phase relay selection scheme in conjunction with our proposed relay selection metrics is proposed to exploit the multiuser diversity effectively. The numerical evaluation shows that our proposed scheme outperforms the other schemes from the literature comfortably when power allocation is applied on the link.
Acknowledgement
The authors are thankful to Ioannis Krikidis for useful discussions that helped immensely to improve the quality of the paper.
Appendix A Outage for the SRS scheme without CSIT
Since the relay node selected for information transmission does not simultaneously harvest energy, the probability that a certain relay
will be selected for information delivery at time moment
can be expressed as(31) 
where denotes the overall amount of stored harvested power at a relay after previous time slots and can be expressed using (4), (14) and (15) as
(32) 
where and , where is the mean of the exponential RV . denotes the number of time slots when this relay was chosen for the relaytodestination transmission, and its maximum equals when the relay is chosen for information transmission every two time slots^{3}^{3}3This scenario is applicable when a number of relays is low and will result in the high system outage since the energy harvested within one or two time slots, due to the channel randomness, is less likely to be sufficient to support information transmission with a fixed transmit power , i.e., the higher is the required the more severe outage occur..
Considering i.i.d. Rayleigh fading channels, the PDF of can be given for different as
(33) 
Therefore, due to the independence of the involved RVs, the relay selection probability can be further written as
(34) 
where denotes the SNR value associated with the rate threshold .
is the cumulative distribution function (CDF) of RV
given by(35) 
where denotes the lower incomplete Gamma function [40].
These derivations prove the results on outage performance presented in Proposition 1.
Appendix B Outage for the SRS scheme with CSIT
In the case when CSIT on the link is available at the relay nodes, the selected relay can efficiently allocate the power as . Therefore, the outage probability of a relay can evaluated as
(37) 
where and is the complementary CDF defined as in (B), shown at the top of the next page, where , and denote the modified Bessel function of the second kind of order 1, the Meijer Gfunction [41, (8.4.3.1)] and the Fox’s Hfunction [42, (1.2)], [43], respectively.
This proves the result in Proposition 2.
(38) 
References
 [1] Y. He, X. Cheng, W. Peng, and G. L. Stuber, “A survey of energy harvesting communications: models and offline optimal policies,” IEEE Communications Magazine, vol. 53, no. 6, pp. 79–85, June 2015.
 [2] I. Ahmed, M. M. Butt, C. Psomas, A. Mohamed, I. Krikidis, and M. Guizani, “Survey on energy harvesting wireless communications: Challenges and opportunities for radio resource allocation,” Computer Networks, vol. 88, pp. 234–248, 2015.
 [3] L. Liu, R. Zhang, and K.C. Chua, “Wireless information and power transfer: A dynamic power splitting approach,” IEEE trans. Communications, vol. 61, no. 9, pp. 3990–4001, Sep. 2013.
 [4] N. Ali, A., Z. Xiangyun, D. Salman, and K. Rodney, A., “Relaying protocols for wireless energy harvesting and information processings,” IEEE Transactions on Wireless Communications, vol. 12, no. 7, pp. 3622–3636, 2013.
 [5] X. Gao, P. Wang, D. Niyato, K. Yang, and J. An, “Auctionbased time scheduling for backscatteraided rfpowered cognitive radio networks,” IEEE Transactions on Wireless Communications, vol. 18, no. 3, pp. 1684–1697, March 2019.
 [6] I. Krikidis, S. Timotheou, and S. Sasaki, “RF energy transfer for cooperative networks: Data relaying or energy harvesting?” Communications Letters, IEEE, vol. 16, no. 11, pp. 1772–1775, 2012.
 [7] D. Michalopoulos, H. Suraweera, and R. Schober, “Relay selection for simultaneous information transmission and wireless energy transfer: A tradeoff perspective,” Selected Areas in Communications, IEEE Journal on, vol. 33, no. 8, pp. 1578–1594, Aug 2015.
 [8] Y. Gu, H. Chen, Y. Li, Y. Liang, and B. Vucetic, “Distributed multirelay selection in accumulatethenforward energy harvesting relay networks,” IEEE Transactions on Green Communications and Networking, vol. 2, no. 1, pp. 74–86, Mar. 2018.
 [9] K. Liu, “Performance analysis of relay selection for cooperative relays based on wireless power transfer with finite energy storage,” IEEE Transactions on Vehicular Technology, vol. 65, no. 7, pp. 5110–5121, Jul. 2016.
 [10] S. Arzykulov, G. Nauryzbayev, T. A. Tsiftsis, and M. Abdallah, “Error performance of wireless powered cognitive relay networks with interference alignment,” in IEEE 28th Annual International Symposium on Personal, Indoor, and Mobile Radio Communications (PIMRC), Oct. 2017, pp. 1–5.
 [11] C. Zhong, H. Suraweera, G. Zheng, I. Krikidis, and Z. Zhang, “Wireless information and power transfer with full duplex relaying,” Communications, IEEE Transactions on, vol. 62, no. 10, pp. 3447–3461, Oct 2014.
 [12] G. Nauryzbayev, K. M. Rabie, M. Abdallah, and B. Adebisi, “Ergodic capacity analysis of wireless powered AF relaying systems over fading channels,” in IEEE Global Communications Conference (GLOBECOM), Dec. 2017, pp. 1–6.
 [13] G. Nauryzbayev, M. Abdallah, and K. M. Rabie, “Ergodic capacity analysis of wireless powered AF relaying systems over fading channels,” in Proc. of IEEE Conference on Vehicular Technology (VTCFall), Aug. 2018, pp. 1–6.
 [14] A. Nasir, X. Zhou, S. Durrani, and R. Kennedy, “Wirelesspowered relays in cooperative communications: Timeswitching relaying protocols and throughput analysis,” Communications, IEEE Transactions on, vol. 63, no. 5, pp. 1607–1622, May 2015.
 [15] G. Zhu, C. Zhong, H. Suraweera, G. Karagiannidis, Z. Zhang, and T. Tsiftsis, “Wireless information and power transfer in relay systems with multiple antennas and interference,” Communications, IEEE Transactions on, vol. 63, no. 4, pp. 1400–1418, April 2015.
 [16] G. Nauryzbayev, K. M. Rabie, M. Abdallah, and B. Adebisi, “On the performance analysis of WPTbased dualhop AF relaying networks in  fading,” IEEE Access, vol. 6, pp. 37 138–37 149, 2018.
 [17] K. M. Rabie, B. Adebisi, and M. Alouini, “Wireless power transfer in cooperative DF relaying networks with lognormal fading,” in IEEE Global Communications Conference (GLOBECOM), Dec. 2016, pp. 1–6.
 [18] K. M. Rabie, A. Salem, E. Alsusa, and M. Alouini, “Energyharvesting in cooperative AF relaying networks over lognormal fading channels,” in IEEE International Conference on Communications (ICC), May 2016, pp. 1–7.
 [19] H. Yu, D. Wang, G. Pan, R. Shi, J. Zhang, and Y. Chen, “On outage of WPC system with relay selection over Nakagami fading channels,” IEEE Transactions on Vehicular Technology, vol. 66, no. 9, pp. 8590–8594, Sep. 2017.
 [20] B. Xia, C. Li, and Q. Jiang, “Outage performance analysis of multiuser selection for twoway fullduplex relay systems,” IEEE Communications Letters, vol. 21, no. 4, pp. 933–936, Apr. 2017.
 [21] Z. Ding, S. M. Perlaza, I. Esnaola, and H. V. Poor, “Power allocation strategies in energy harvesting wireless cooperative networks,” Wireless Communications, IEEE Transactions on, vol. 13, no. 2, pp. 846–860, Feb. 2014.
 [22] R. Senanayake, S. Atapattu, J. S. Evans, and P. J. Smith, “Decentralized relay selection in multiuser multihop decodeandforward relay networks,” IEEE Transactions on Wireless Communications, vol. 17, no. 5, pp. 3313–3326, May 2018.
 [23] A. Banerjee, A. Paul, and S. P. Maity, “Joint power allocation and route selection for outage minimization in multihop cognitive radio networks with energy harvesting,” IEEE Transactions on Cognitive Communications and Networking, vol. 4, no. 1, pp. 82–92, March 2018.
 [24] J. Li, L. J. Cimini, J. Ge, C. Zhang, and H. Feng, “Optimal and suboptimal joint relay and antenna selection for twoway amplifyandforward relaying,” IEEE Transactions on Wireless Communications, vol. 15, no. 2, pp. 980–993, Feb. 2016.
 [25] D. Wang, R. Zhang, X. Cheng, L. Yang, and C. Chen, “Relay selection in fullduplex energyharvesting twoway relay networks,” IEEE Transactions on Green Communications and Networking, vol. 1, no. 2, pp. 182–191, Jun. 2017.
 [26] Z. Chang, X. Hou, X. Guo, T. Ristaniemi, and Z. Han, “Secure and energyefficient resource allocation for wireless power enabled full/halfduplex multipleantenna relay systems,” IEEE Transactions on Vehicular Technology, vol. 66, no. 12, pp. 11 208–11 219, Dec. 2017.
 [27] W. Wang, K. C. Teh, and K. H. Li, “Generalized relay selection for improved security in cooperative DF relay networks,” IEEE Wireless Communications Letters, vol. 5, no. 1, pp. 28–31, Feb. 2016.
 [28] ——, “Relay selection for secure successive AF relaying networks with untrusted nodes,” IEEE Transactions on Information Forensics and Security, vol. 11, no. 11, pp. 2466–2476, Nov. 2016.
 [29] A. Kuhestani, A. Mohammadi, and M. Mohammadi, “Joint relay selection and power allocation in largescale mimo systems with untrusted relays and passive eavesdroppers,” IEEE Transactions on Information Forensics and Security, vol. 13, no. 2, pp. 341–355, Feb. 2018.
 [30] M. M. Butt, A. Nasir, A. Mohamed, and M. Guizani, “Trading wireless information and power transfer: Relay selection to minimize the outage probability,” in IEEE Global Conference on Signal and Information Processing (GlobalSIP), Dec 2014, pp. 253–257.
 [31] M. M. Butt, A. Salama, A. Mohamed, and M. Guizani, “Relay selection schemes to minimize outage in wireless powered communication networks,” IET Signal Processing, vol. 10, no. 3, pp. 203–209, May 2016.
 [32] M. M. Butt, I. Krikidis, and N. Marchetti, “Minimizing outage probability by exploiting csi in wireless powered cooperative networks,” in IEEE Global Communications Conference (GLOBECOM), Dec 2016, pp. 1–7.
 [33] A. Ikhlef, J. Kim, and R. Schober, “Mimicking fullduplex relaying using halfduplex relays with buffers,” IEEE Transactions on Vehicular Tech., vol. 61, no. 7, pp. 3025–3037, September 2012.
 [34] F. Yuan, K. Q. T. Zhang, S. Jin, and H. Zhu, “A harvestusestore mode for energy harvesting communication systems with optimal power policy,” in Proc. IEEE International Conference on Communications (ICC), Sydney, Australia, June 2014, pp. 5366–5371.
 [35] E. Boshkovska, D. W. K. Ng, N. Zlatanov, and R. Schober, “Practical nonlinear energy harvesting model and resource allocation for SWIPT systems,” IEEE Communications Letters, vol. 19, no. 12, pp. 2082–2085, Dec 2015.
 [36] P. N. Alevizos and A. Bletsas, “Sensitive and nonlinear farfield RF energy harvesting in wireless communications,” IEEE Transactions on Wireless Communications, vol. 17, no. 6, pp. 3670–3685, June 2018.
 [37] Y. Luo, J. Zhang, and K. B. Letaief, “Relay selection for energy harvesting cooperative communication systems,” in IEEE Global communications conference (Globecom), Atlanta, GA, USA, Dec. 2013.
 [38] H. Ju and R. Zhang, “Throughput maximization in wireless powered communication networks,” Wireless Communications, IEEE Transactions on, vol. 13, no. 1, pp. 418–428, January 2014.
 [39] I. Krikidis, S. Sasaki, S. Timotheou, and Z. Ding, “A low complexity antenna switching for joint wireless information and energy transfer in MIMO relay channels,” Communications, IEEE Transactions on, vol. 62, no. 5, pp. 1577–1587, May 2014.
 [40] I. S. Gradshteyn and I. M. Ryzhik, Table of Integrals, Series, and Products. San Diego, USA: Academic, 2007.
 [41] A. P. Prudnikov, Y. A. Brychkov, and O. I. Marichev, Integrals and Series, Vol. 3: More Special Functions. New York: Gordon and Breach, 1990.
 [42] A. Mathai, R. Saxena, and H. Haubold, The HFunction: Theory and Applications. Springer New York, 2009.
 [43] G. Nauryzbayev, K. M. Rabie, M. Abdallah, and B. Adebisi, “On the performance analysis of WPTbased dualhop AF relaying networks in fading,” IEEE Access, vol. 6, pp. 37 138–37 149, 2018.
Comments
There are no comments yet.