End-to-End Multiservice Delivery in Selfish Wireless Networks Under ...

1132

IEEE TRANSACTIONS ON COMMUNICATIONS, VOL. 64, NO. 3, MARCH 2016

End-to-End Multiservice Delivery in Selfish Wireless Networks Under Distributed Node-Selfishness Management Jinglei Li, Qinghai Yang, Peng Gong, and Kyung Sup Kwak, Member, IEEE Abstract—In this paper, we investigate the multiservice delivery between the source–destination pairs in distributed selfish wireless networks (SeWN), where selfish relay nodes (RN) expose their selfish behaviors, i.e., forwarding or dropping multiservices. Owing to the effect of the RNs’ node-selfishness on the multiservices, a distributed framework of the node-selfishness management is constructed to manage the RN’s node-selfishness information (NSI) in terms of its available resources, the employed incentive mechanism and the quality-of-service (QoS) requirements, and the other RNs’ NSI in terms of their historical behaviors. In this framework, the RNs’ NSI includes the degree of node-selfishness (DeNS), the degree of intrinsic selfishness (DeIS) and the degree of extrinsic selfishness (DeES). Under the distributed node-selfishness management, a path selection criterion is designed to select the most reliable and shortest path in terms of RNs’ DeISs affected by their available resources, and the optimal incentives are determined by the source to stimulate forwarding multiservices of the RNs in the selected path. Our simulation results demonstrate that this framework effectively manages the RNs’ NSI, and the optimal strategies of both the path selection and the incentives are determined. Index Terms—Node-selfishness management, selfish wireless networks, multi-services, path selection.

I. I NTRODUCTION

I

N RECENT years, we have witnessed a drastic growth in demand for multimedia services such as different styles of media streams (i.e., video, voice and data streams) [1] and different priority classes of one traffic streams [2], which are referred to as multiple services having different quality of service (QoS) requirements in wireless networks. Given the proliferation of smart devices in distributed intelligent networks, each node is expected to be endowed with smart autonomic functions. By instinct, the individual network nodes would prefer to act selfishly rather than altruistically in distributed network Manuscript received August 7, 2015; revised December 7, 2015; accepted January 21, 2016. Date of publication January 26, 2016; date of current version March 15, 2016. This research was supported in part by ITRC program, MSIP, Korea (IITP-2015-H8501-15-1019), NSF China (61471287). The associate editor coordinating the review of this paper and approving it for publication was W. Saad. J. Li and Q. Yang are with the State Key Laboratory of ISN, School of Telecommunications Engineering, Xidian University, Xi’an 710071, China (e-mail: [email protected]). P. Gong is with the National Key Laboratory of Mechatronic Engineering and Control, School of Mechatronical Engineering, Beijing Institute of Technology, Beijing 100081, China (e-mail: [email protected]). K. S. Kwak is Inha Hanlim Professror with the Department of Information and Communication Engineering, Inha University, Incheon 402-751, South Korea (e-mail: [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TCOMM.2016.2521741

scenarios. A distributed wireless network which consists of nodes exhibiting a selfish behavior is referred to as a distributed selfish wireless network (SeWN). In such network scenarios, the selfish behavior of network nodes, referred to as “nodeselfishness”, may degrade the network performance, e.g., the network connectivity [3], the reliability of the selected path and the probability of the successful End-to-End (E2E) multiservice delivery [4]. The node-selfishness of the network node is affected by some intrinsic and extrinsic factors, such as its own energy and bandwidth resources, the QoS requirements and the employed incentive mechanisms. For improving the network performance, the node individuals need to obtain the information on the node-selfishness of the other nodes and to determine the relationship between the aforementioned factors and the node-selfishness. In such distributed network scenarios, each network node may obtain the aforementioned information, directly collected by itself and/or indirectly received from its neighboring nodes. Accordingly, each network node should establish a distributed node-selfishness management for managing the aforementioned information on the node-selfishness, whilst improving the network performance of delivering multiservice, i.e., the reliability of the selected path and the successful probability of delivering multi-services. Many literatures have investigated the multi-service delivery in distributed wireless networks. A cross-layer resource allocation scheme was developed in [5] for guaranteeing the QoS requirements of the voice and data traffic. A spatial-correlationaware QoS routing algorithm was proposed in [6] for efficiently delivering visual service under QoS constraints. A geographic opportunistic routing was explored in [7] for delivering packets with both E2E reliability and delay constraints in wireless sensor networks. The feasibility of the aforementioned schemes disregard the impact of the node-selfishness on the multi-service delivery in distributed wireless networks. Due to the effect of the node-selfishness on the network performance, some management approaches of dealing with the nodes’ selfish behaviors have been studied for distributed wireless networks. A heterogeneous trust management was proposed in [8] for evaluating the trust of a network node in terms of its social and QoS behavior. A dynamic trust management was developed in [9] for minimizing the trust bias and maximizing the routing application performance in the presence of wellbehaved and selfish nodes. A group-based trust management scheme was proposed in [10] for evaluating the trust of a node group and also for excluding selfish nodes. Additionally, some incentive mechanisms have been investigated for improving

0090-6778 © 2016 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

LI et al.: END-TO-END MULTISERVICE DELIVERY IN SELFISH WIRELESS NETWORKS

the network performance. The incentive mechanisms based on credibility was employed in [11] to stimulate selfish nodes for forwarding traffic data. A penalty scheme for punishing selfish misbehavior was employed in [12] to share the wireless channel. The price-based collusion-resistant scheme was conceived in [13] for combating the selfish-user collusion, whilst optimizing overall spectral efficiency. In order to prevent the packet-dropping executed by selfish nodes, the reputation-based scheme was addressed in [14] for determining the trust routers. The d’Aspremont and Gerard-Varet approach was employed in [15] to enforce truth telling upon selfish relay nodes in relay networks. However, the aforementioned literatures neglect the deep analysis of the node-selfishness from the perspectives of all impact factors, i.e., the nodes’ available resources, the QoS requirements of the multi-services and the factor of the employed incentive mechanism. In this paper, we investigate the E2E multi-service delivery in distributed SeWNs, where selfish relay nodes (RN) expose their selfish behaviors, i.e., forwarding or dropping multi-services. In such distributed networks, we refer to the RNs’ available resources as their intrinsic factors, and both the service impact factors, i.e., the QoS requirements, and the factor of an employed incentive mechanism. Accordingly, the degree of intrinsic selfishness (DeIS) and the degree of extrinsic selfishness (DeES) are defined as the influences of the intrinsic factors and the extrinsic factors on the node-selfishness, respectively, while the degree of node-selfishness (DeNS) exhibits the selfish behavior of forwarding multi-services. These DeIS, DeES and DeNS are all the RN’s node-selfishness information (NSI). For the E2E multi-service delivery, the source and every RN should determine its selfish behavior in terms of its intrinsic and extrinsic factors, and extract other RNs’ NSI. According to those NSI, the source selects a reliable and short path and maintains the reliability of this selected path by adjusting the incentives provided for stimulating selfish RNs under an employed incentive mechanism. The main contributions of this paper are outlined as follows: • A distributed framework of the node-selfishness management at every RN is conceived to quantify the effects of its intrinsic and extrinsic factors on its node-selfishness, and also to obtain the other RNs’ NSI in terms of their behaviors of forwarding multi-services and/or the recommended NSI of the RNs around it. • A path selection criterion is designed to select the most reliable and shortest path in terms of the RNs’ DeISs for delivering multi-services. • The optimal incentives are determined for maintaining the reliability of the multi-service delivery under the influence of the node-selfishness of the RNs within the selected path. The remainder of the paper is organized as follows. Section II depicts the system model in SeWNs. The distributed framework of the node-selfishness management is developed to manage the RNs’ NSI in Section III. In Section IV, the sources effectively and reliably deliver their multi-services under the distributed framework of the node-selfishness management. Simulation results are provided in Section V, and Section VI concludes this paper.

1133

II. S YSTEM M ODEL In this section, we introduce the network model including source-destination pairs and some selfish RNs, and as well the profit and cost of each RN while forwarding multi-services. Finally, the RN’s DeIS, DeES and DeNS are defined in terms of its intrinsic and extrinsic factors. A. Network Model The multi-services under our consideration are categorized into H priority classes in terms of their QoS requirements, and refer to priority class h (1 ≤ h ≤ H ) as service h (cf. [2]). Each service h has the unique service impact factor, denoted by λh (λh ≥ 0), and the service impact factor λh 1 ≥ λh 2 , ∀h 1 , h 2 ∈ {1, · · · , H } as h 1 < h 2 . The multi-services are delivered between the source-destination pairs in the distributed SeWNs including some selfish RNs, denoted by V, and the behavior of each selfish RN i (∀i ∈ V) is either to forward multi-services or to drop multi-services. The performance of the E2E multi-service delivery is degraded by the selfish behaviors of RNs, owing to their limit available resources, i.e., their residual energy resources E (E ≥ 0) and available bandwidth W (W ≥ 0). Meanwhile, since the multi-services have different impact factors, the RNs have different resource consumption amounts for satisfying their QoS requirements, which may lead to different selfish behaviors of these RNs. Nevertheless, for successfully delivering multi-services, the source employs an incentive mechanism to depress the node-selfishness of these RNs, and different incentives have different stimulating levels for the selfish behaviors of these RNs. Accordingly, the selfish behavior of each RN i (∀i ∈ V) are affected by its intrinsic factors, i.e., its residual energy E i and available bandwidth Wi , and its extrinsic factors, i.e., the incentive ρi received by it and the impact factor λh of the received service. Due to the effect of the RNs’ selfishness on the E2E multiservice delivery, every source in the SeWNs should obtain the NSI of all RNs for determining the reliability of all path existing between this source-destination pair. Nevertheless, considering the distributed framework of SeWNs where no central network entity collects the NSI of all selfish RNs, every distributed source/RN should have its own node-selfishness management for its selfish behavior and other RNs’ NSI. The node-selfishness management at every RN obtains the NSI of the RNs around it in terms of their historical behaviors and the remote RNs’ NSI recommended by the RNs around it, referred to as the recommended NSI. Additionally, the nodeselfishness management of each RN also determines the effects of the intrinsic and extrinsic factors on its node-selfishness of forwarding multi-services. For the E2E multi-service delivery, every source in such SeWN may obtain some paths of delivering multi-services, denoted by , by using a traditional routing protocol, e.g., the dynamic source routing (DSR) in [16]. Nevertheless, since the node-selfishness of the RNs within these paths degrades the reliability of these paths, each source selects the most reliable and shortest one of these obtained paths, denoted by θ ∗ (θ ∗ ∈ ), in terms of the RNs’ NSI under this distributed framework of the node-selfishness management. Additionally,

1134


Fig. 1. The E2E multi-service delivery in the distributed SeWNs.

since each time of the path selection introduces a large signaling overhead in distributed SeWNs, the source should reduce the selection frequency and maintain the reliability of the selected path. Nevertheless, the node-selfishness of the RNs within the selected path varies with the aforementioned factors, thus leading to the variation of the path reliability. Accordingly, every source may adjust the incentives to depress the node-selfishness of the RNs within the selected path and to maintain the reliability of this path for delivering multi-services. Here, the network model of the distributed SeWNs including a distributed framework of the node-selfishness management and the multi-service delivery of the sources is shown in Fig. 1. B. RN’s Profit and Cost of Forwarding Multiservices When the multi-services are successfully forwarded, RN i (i ∈ V) obtains the revenue of forwarding multi-services, denoted by Pi . Since the multi-services have different bit-errorrate (BER) requirements, via the adaptive modulation [19], the transmission rate of RN i to RN j for service h is expressed as P |g |2

i,h i, j 1.5 , where Pi,h is the transRi,h = Wi log2 1 − ln(5η N0 Wi h) mit power of RN i for service h, N0 is the thermal noise power, gi, j is the channel gain from RN i to RN j, Wi is the available bandwidth possessed by RN i, ηh is the target BER of forwarding service h. The transmission time of RN i for service h is expressed as RLi,hh , where L h is the service length. The resource consumption of RN i for service h with target BER ηh is given by

E i,h = Pi,h

(1)

Accordingly, the resource cost of RN i for forwarding service h depends on the corresponding resource consumption E i,h , expressed as Ci,h = π E i,h .

S = f (S I , S E ),

(3)

which is a non-decreasing function with respect to (w.r.t.) its DeIS and DeES. When the RN’s DeIS S I = 0 or its DeES S E = 0, its DeNS S = 0, meaning that the RN having either the infinite available resources or the infinite incentive is altruistic. While the RN’s DeIS S I = 1 and its DeES S E = 1, its DeNS S = 1, meaning that this RN of no available resources and no incentive for the service of the high impact factor is completely selfish. Naturally, it is more common that 0 < S I < 1 and 0 < S E < 1, yielding the RN’s DeNS 0 < S < 1. III. D ISTRIBUTED F RAMEWORK OF N ODE -S ELFISHNESS M ANAGEMENT

Ri,h Wi

Lh (1 − 2 ) ln(5ηh )L h N0 Wi = . Ri,h 1.5|gi, j |2 Ri,h

are the employed incentives and the impact factors of the multi-service. The effects of the RN’s intrinsic factors on its selfishness are its own characteristics owing to the fact that these resources are possessed by itself [22], and the effects of the RN’s extrinsic factors on its selfishness are its own characteristics owing to the fact that the extrinsic factors are provided by the source (cf. [13]). Although the RN’s extrinsic factors affect its node-selfishness, these factors are unable to influence the relationship between its intrinsic factors and its selfishness. Therefore, there are uncorrelated among the effects of the intrinsic and extrinsic factors. Accordingly, it is rational that we characteristics the effects of the intrinsic and extrinsic factors on the RN’s selfishness by using the DeIS and DeES, which are defined as follows. Definition 1 (DeNS, DeIS and DeES): The RN’s DeIS S I is defined as the degree reflecting the effects of intrinsic factors on its selfish behavior, while the RN’s DeES S E is defined as the degree reflecting the effects of extrinsic factors on its selfish behavior. The RN’s DeNS is defined as the degree reflecting the effects of all impact factors on its selfish behavior, denoted by S. The DeIS, DeES and DeNS, which are referred to as the RN’s NSI, vary from 0 (altruistic) to 1 (completely selfish) via a normalization. From Definition 1, the contributions of DeIS and DeES to DeNS are decoupled, since DeIS and DeES are uncorrelated. Furthermore, the RN’s DeNS is a comprehensive formulation of both DeIS and DeES, expressed as

(2)

where π is the energy price, which is a constant and is the same for all RNs. C. RN’s Node-Selfishness Information Note that the RN’s intrinsic and extrinsic factors affect its node-selfishness. The intrinsic factors of the RN are its residual energy and available bandwidth, while its extrinsic factors

Under the distributed framework of the node-selfishness management, the node-selfishness management of each RN consists of two parts: the management of its NSI and the management of the other RNs’ NSI. For the management of the RN’s node-selfishness, the models of the intrinsic and extrinsic selfishness are designed for computing its DeIS and DeES in terms of its intrinsic and extrinsic factors, respectively. Additionally, for the management of the other RNs’ NSI, their DeISs, DeESs and DeNSs are obtained in terms of their historical behaviors and the NSI of these RNs recommended by the RNs around it. Accordingly, the framework of the RN’s node-selfishness management is illustrated in Fig. 2. Nevertheless, since the sources have no selfishness, the node-selfishness management of every source has no management of its node-selfishness, and only manages the NSI


1135

available bandwidth to its DeIS is plausibly expressed as (c.f. the epidemic model in [21]) S FI = e− F W , I

FI

Fig. 2. The framework of the node-selfishness management of each RN to manage its own NSI in terms of its intrinsic and extrinsic factors and to obtain the other RNs’ NSI in terms of their historical behaviors and the recommended NSI from the RNs around it.

of the RNs in terms of their historical behaviors and the recommended NSI, which is similar to the second part of the node-selfishness management of every RN. Accordingly, in this section, we just analyze the node-selfishness management of one RN. A. Management of RN’s Node-Selfishness Owing to the effects of the RN’s intrinsic and extrinsic factors on its node-selfishness of forwarding multi-services, it should quantify these effects, thus the models of the intrinsic and extrinsic selfishness are designed to obtain its DeIS and DeES, respectively. 1) Model of Intrinsic Selfishness: Owing to the effects of both the RN’s residual energy and its available bandwidth on its DeIS, the model of intrinsic selfishness is divided into two parts: the effect of its residual energy on its DeIS, denoted as S EI (0 ≤ S EI ≤ 1), and the effect of its available bandwidth on its DeIS, denoted as S FI (0 ≤ S FI ≤ 1). From the residual-energy point of view, the RN’s willingness of forwarding multi-services decreases as its residual energy is depleted. Correspondingly, its DeIS increases as its residual energy decreases. From the available bandwidth point of view, the RN has the low willingness of forwarding multi-service if the available bandwidth is insufficient and vice versa. Hence, in order to balance the effects of both its residual energy and available bandwidth on its DeIS, the RN’s DeIS is developed as S I = ω I S FI + (1 − ω I )S EI ,

(4)

where ω I (0 ≤ ω I ≤ 1) is the weight for balancing these two effects. Here, we analyze the influence of the RN’s available bandwidth on its DeIS. In the presence of large available bandwidth, the RN has high willingness of forwarding multi-services, and thus exposing low DeIS. In the presence of mediocre available bandwidth, its willingness of forwarding multi-services slowly decreases. Extremely, in the presence of the null available bandwidth, this RN may drop multi-services, thus exposing high DeIS. The RN’s DeIS may exponentially increase as its available bandwidth decreases. Hence, mapping the RN’s

(5)

( FI

where ≥ 0) represents the spectral efficiency of its available bandwidth. Additionally, we formulate the relationship between the DeIS and the RN’s residual energy. The DeIS increases as the residual energy retained at the RN is depleted. Typically, when an RN has sufficiently high residual energy, it is likely to be more willing to forward multi-services, while in the presence of mediocre residual energy level, the willingness of forwarding multi-services is reduced. Finally, in the presence of a low residual energy level, the RN may drop multi-services. By mapping the residual energy to the DeIS, we arrive at the plausible formulation as (c.f. the epidemic model in [21] and the hyperbolic selfishness behavior in [22]) κ I 1 − e− E E I , (6) SE = 1 − I ¯ 1 − e− E E where E¯ ( E¯ ≥ 0) is the total energy initially possessed by the RN, κ (κ ≥ 0) and EI ( EI ≥ 0) are the parameters scaling the impact of its residual energy on the DeIS. The parameter κ represents the prejudice characteristics of the RN’s energy consumption. When κ = 1, the RN has no prejudice against its energy consumption; when κ < 1, the RN has an altruistic prejudice against its energy consumption; while κ > 1, the RN has a selfish prejudice against its energy consumption. The parameter EI represents the useful value of the energy resource. The parameter EI of large useful value causes the RN’s altruistic behavior. While considering two extreme cases: for E = 0, S EI = 1, which means that the RN is completely selfish for for¯ S I = 0, which means that warding multi-services; for E = E, E the RN is altruistic to forward multi-services. Naturally, it is ¯ yielding 0 < S I < more common that the RN has 0 < E < E, E 1, which is affected by both the parameters κ and EI . 2) Model of Extrinsic Selfishness: Owing to the effects of both the service impact factor and the incentive on the RN’s DeES, the model of extrinsic factor is formulated including two parts: the effect of the service impact factor on its DeES, denoted as SSE (0 ≤ SSE ≤ 1), and the effect of the incentive on its DeES, denoted as S IE (0 ≤ S IE ≤ 1). From the service impact factor point of view, the RN has the high willingness of forwarding the service of low impact factor, because forwarding the service of high impact factor needs more resource consumption than that of low service impact factor. Its DeES increases as the impact factor of the service increases. From the incentive point of view, the received incentive is employed to depress the RN’s node-selfishness. The RN’s DeES increases as the received incentive decreases. Here, in order to balance these two effects SSE and S IE in terms of its sensitivities to the service impact factor and the received incentive, the RN’s DeES is expressed as S E = ω E SSE + (1 − ω E )S IE , ωE

(7)

where ∈ [0, 1] is the weight value to balance these two effects SSE and S IE .

1136


Now, we further analyze the RN’s DeES from the perspective of the impact factors of the multi-services. In the presence of high service impact factor, the RN has low willingness of forwarding this service owing to large resource consumption for satisfying its QoS requirement, thus exposing its large DeES. By contrast, in the presence of low service impact factor, the RN has high willingness of forwarding this service owing to little resource consumption, thus exposing its small DeES. Here, we model that the RN’s DeES exponentially increases as the service impact factor increases, and mapping the service impact factor to its DeES is plausibly expressed as (c.f. the epidemic model in [21]) SSE = 1 − e− S λ , E h

(8)

where SE ( SE ≥ 0) is the difficulty coefficient of processing the multi-services. Additionally, we further portray the effect of the incentive on the RN’s DeES from the perspective of an incentive. In the presence of high incentive, the RN has the high willingness of forwarding multi-services. In contrast, in the presence of the low IPF, this RN has the low willingness of forwarding multi-services. Plausibly, we model that the RN’s DeES may exponentially decrease as the incentive increases. Here, the relationship between the incentive and the RN’s DeES is plausibly formulated as (c.f. the epidemic model in [21]) S IE = e− I ρ , E

(9)

where IE ( IE ≥ 0) is the impact coefficient of the incentive under the employed incentive mechanism. B. Management of Other RNs’ NSI In distributed SeWNs, besides managing the RN’s own NSI, its node-selfishness management also acquires the other RNs’ NSI, including their DeISs, DeESs and DeNSs, for the multiservice delivery. The RN obtains other RNs’ NSI by two ways, i.e., obtaining the recommended NSI from the RNs around it and extracting other RNs’ NSI in terms of their historical behaviors. For the RN’s node-selfishness management, the other RNs’ NSI directly extracted is referred to as their direct ˆ while the other RNs’ NSI recommended NSI, denoted by S, ˇ Meanwhile, is referred to as their indirect NSI, denoted by S. the RN weights the direct and indirect NSI of other RNs in terms of their locations. For the neighbor nodes of the RN, their direct NSI extracted by it is more reliable than their indirect NSI. Nevertheless, for the RNs far away from the RN, this RN may only have its indirect NSI recommended by its neighbor RNs. Accordingly, the node-selfishness management of each RN finally obtains the NSI of RN i, expressed as φ j Sˇ i, j , (10) S˜ i = μSˆ i + (1 − μ) j∈i

where Sˆ i is the direct NSI of RN i, Sˇ i, j is the indirect NSI of RN i recommended by RN j, μ is the weight value to balance their direct and indirect NSI, i is the set of the RNs recommending the NSI of RN i, φ j (0 ≤ φ j ≤ 1) represents the trust

level of RN j, i.e., how much trust level of the information recommended by the j-th RN. In this management of the other RNs’ NSI, we mainly care about the method of extracting the direct NSI Sˆ in terms of their historical behaviors. Since the other RNs’ available resources gradually decreases with the resource consumption of forwarding multi-services, the available resource states of each one of the other RNs at the adjacent time instances are time-correlated. Hence, during a period of time, the available resource states of each of the other RNs are highly timecorrelated and have low changing rate, thus the other RNs’ DeNSs and DeISs may be directly extracted by such timecorrelated characteristics. Nevertheless, since both the service impact factors and the received incentives also affect the RN’s behaviors of forwarding multi-services during this period of time, it causes the fluctuation of these behaviors and correspondingly degrades their time correlation. Hence, the other RNs’ DeESs may be indirectly obtained by the time-correlated characteristics together with the extracted DeISs and DeNSs. Accordingly, the direct NSI S˜ extracted by the RN’s nodeselfishness management is portrayed into the DeNS extraction, the DeIS extraction and the DeNS extraction. 1) DeNS Extraction: The RN’s node-selfishness management extracts the other RNs’ DeNSs in terms of their historical behaviors, i.e., forwarding and dropping multi-services, which are referred to as the binary events. The other RNs’ DeNSs are extracted by the beta probability density function, which is used to represent the probability distribution of these binary events and is illustrated in Appendix A. A time window and a forgetting factor are adopted to deal with the others RNs’ historical behaviors for extracting their DeNSs. Within a time window, the RN’s available resource states are highly time-correlated. Nevertheless, since the earlier behaviors are less correlated with the current behaviors than the later ones, a forgetting factor is used to provide a small weight on the earlier behavior and a large weight on the later behavior. Within the time window with length T , the historical behaviors of RN i are expressed as Bi = {Bi,t |0 ≤ t ≤ T − 1}, where Bi,t ∈ {0, 1} is the behavior information of RN i at time t, and T is properly set in terms of the time-correlation of its behaviors. The behavior information Bi,t = 1 represents that RN i forwards multi-services at time t, while Bi,t = 0 represents that it drops multi-services at time t. The behavior information Bi,0 is the latest action information of RN i, while Bi,T −1 is its earliest behavior information. Additionally, the forgetting factor τi (0 ≤ τi ≤ 1) is used to weight the historical behaviors of RN i within the time window, and (τi )t is used to weight on the behavior information Bit . Accordingly, the extracted DeNS of RN i is expressed as Sˆi =

1+

T −1

(1 − Bi,t )(τi )t , T −1 2 + t=0 (τi )t t=0

(11)

where τi is related to the variance measure of the historical behaviors Bi during the time window of length T . The above equation is obtained in Appendix A. 2) DeIS Extraction and DeES Extraction: In the RN’s node-selfishness management, the other RNs’ DeISs and


DeESs are extracted in terms of their historical behaviors together with their DeNSs. The other nodes’ DeISs extraction is related to their intrinsic information, i.e., the other RNs’ residual energy resources and their available bandwidth resources. Nevertheless, since the other RNs’ available resources are private to themselves, their resource information is unable to be directly obtained, so that the observable behaviors are employed to extract their DeISs. Within the time window, since the other RNs’ available resource states are highly timecorrelated, their DeNSs may be unchanged. Furthermore, the RN’s node-selfishness management extracts the other RNs’ DeISs in terms of their behaviors of forwarding the service of low impact factor without the source’s incentive or under the sufficiently large incentive, because in these cases these behaviors are hardly affected by their extrinsic factors. Hence, in this special case, the DeIS of RN i is approximately equal to its extracted DeNS, expressed as S˜iI = S˜ i ,

(12)

where S˜ i is the extracted DeNS in the aforementioned special case. In addition, the other RNs’ DeES extraction is executed by the RN’s node-selfishness management in terms of their extracted DeNSs and DeISs by Eq. (3), thus the extracted DeES of RN i is expressed as S˜iE = g( S˜i , S˜iI ),

(13)

where g(·) is the inverse function of Eq. (3) w.r.t. the DeES of RN i. 3) Parameters Extraction: For further analyzing the effect of each impact factor on the other RNs’ node-selfishness, the RN’s node-selfishness management also needs to evaluate the parameters of the other RNs ω I , FI , EI , ω E , SE and IE for effectively optimizing the network performance. The kernel methods in regression [17] and the neural network [18] have been employed to evaluate the parameters of a nonlinear function and a video transmission. Hence, these methods may also be employed for the RN’s parameter extraction, which will not be studied in this paper due to the limited space of this paper. IV. E2E M ULTISERVICE D ELIVERY U NDER D ISTRIBUTED F RAMEWORK OF N ODE -S ELFISHNESS M ANAGEMENT In this section, in terms of the obtained NSI under the distributed framework of the node-selfishness management, every source selects the most reliable and shortest path and provides the optimal incentives to the RNs within the selected path for the multi-service delivery, as illustrated in Fig. 3. A. Path Selection For delivering multi-services, the sources find some paths by virtue of the traditional routing protocol. Nevertheless, these paths may not be all reliable for successfully forwarding multiservices due to the node-selfishness of the RNs within these paths. Although the RN’s DeNS represents its behavior of forwarding multi-services, the DeNS information is not the best

1137

Fig. 3. The source node selects a short and reliable path and determines the optimal incentives for maintaining the reliability of the multi-service delivery under the distributed framework of the node-selfishness management.

option to select the most reliable path. When the RNs of a few available resources have high DeIS, their historical behaviors may be the ones of forwarding the multi-services owing to the large incentives, thus leading to their low DeNSs. If these RNs of high DeISs are selected within the path of delivering multi-services in terms of the RNs’ DeNSs, the sources should provide large incentives for stimulating the multi-service forwarding of these RNs and maintaining the reliability of the selected path. In contrast, if the RNs of low DeISs are selected within the path of delivering multi-services in terms of the DeIS information, the sources only provide a few incentives for maintaining the path reliability. Accordingly, the DeIS is regarded as the better parameter of selecting the most reliable path than the DeNS. Therefore, the sources employ the RNs’ DeISs, rather than their DeNSs, to select the most reliable and shortest one of the potential paths. By using a traditional routing protocol, each source may obtain M paths, each one of which is possible to be used for delivering multi-services, denoted by . For obtaining the most reliable and shortest one from M paths, the reliability of a path of delivering multi-services is measured by the RNs’ DeISs, which are extracted in the source’s node-selfishness management. For the sake of analysis, let us denote the set of all RNs within path θ (θ ∈ ) by θ = {θ1 , . . . , θ|θ| }, where |θ | is the number of RNs within path θ , unless this might lead to ambiguity. Here, we define the path reliability of path θ (θ ∈ ) as the product of the reliable degrees of all RNs within this path, expressed as θ =

|θ|

(1 − S˜iI ),

(14)

i=1

where S˜iI is the extracted DeIS of RN i within path θ . Based on the path reliability of each path, the path selection criterion is proposed for selecting the most reliable and shortest one of all paths obtained by each source. Definition 2 (Path Selection Criterion): In order to select the most reliable and shortest one from the obtained paths , every source selects one path, which has the maximum path reliability among all path , which is formulated as θ ∗ = arg max{θ |θ ∈ }. θ

(15)

Note that, under the path selection criterion, the path reliability of path θ is related to the number of the RNs within

1138


path θ and their extracted DeISs S˜iI (∀θi ∈ Rθ ). The path reliability of path θ increases as the DeIS of each RN within this path decreases. Accordingly, the path with the maximum path reliability is the most reliable one. Meanwhile, the path reliability of path θ increases as the RN number within this path decreases. The path with the maximum path reliability may be the shortest one. Therefore, by using the path selection criterion, every source obtains the most reliable and shortest one from the obtained paths . B. Multi-Service Delivery With RNs’ NSI After selecting the shortest and most reliable path θ ∗ = for the multi-service delivery by using the path selection criterion, the source should maintain the reliability of path θ ∗ for delivering multi-services by adjusting the incentives for depressing the RNs’ node-selfishness. Nevertheless, the incentives provided by the source are the cost of delivering multi-services, thus the source should minimize such incentive costs. Therefore, for maintaining the path reliability, the source should minimize the incentive costs subject to the successful multi-service forwarding of each RN within the selected path θ ∗. Before analyzing the problem of the multi-service delivery, we should analyze the effect of the node-selfishness on the multi-service forwarding of each RNs within path θ ∗ . Since the RN’s behavior of forwarding multi-service is affected by its available resources, it has the expected value of the resource consumption cost affected by its DeIS. The RN of high DeIS cares its lack available resource and conserve its available resource, thus the resource cost of forwarding multi-services is regarded as a large value. In contrast, the RN of low DeIS does not care the resource consumption of forwarding multiservices and has high willingness of forwarding multi-services, thus the resource consumption cost is regarded as a small value. Here, the expected resource cost of each RN within path θ ∗ for forwarding multi-services is expressed as {θ1∗ , . . . , θ|θ∗ ∗ | }

C˜ i,h = S˜iI Ci,h , ∀θi∗ ∈ θ ∗ , ∀h ∈ {1, . . . , H }.

(16)

Additionally, we study the relationship between the RN’s DeES and its profit of forwarding multi-services. From Eq. (7), the RN’s DeES decreases as the provided incentive increases for forwarding service h. The RN of low DeES, for which the larger incentive is provided by the source, appreciates service h from this source and has high willingness of forwarding service h. In contrast, the RN of high DeES has the smaller incentive from the source and may drop service h. Here, the expected service forwarding profit of the RN within path θ ∗ for forwarding multi-services in terms of its DeES, expressed as Pi P˜ i,h = E , ∀θi∗ ∈ θ ∗ , ∀h ∈ {1, . . . , H }. ˜S i,h

(17)

In this distributed SeWN, the source and every RN obtain the other RNs’ NSI, i.e., their DeISs and DeESs, under the distributed framework of the node-selfishness management. According to Eqs. (16) and (17) together with the NSI of the

RNs within path θ ∗ , the problem of the multi-service delivery of the source is expressed as min∗ ρi,h (18) ρi,h ,∀θi ∈θ ∗

s.t.

θi∗ ∈θ ∗

Pi ≥ S˜iI π E i,h , ∀θi∗ ∈ θ ∗ . ˜S E i,h

(19)

Meanwhile, Eq. (18) is a linear function of the incentives provided by the source for each RNs within the selected path θ ∗ and IEq. (19) is convex w.r.t. the incentive. For satisfying IEq. (19), the source provides the high incentive to the RNs within the selected path θ ∗ . In contrast, in order to minimize the objective function, the source tends to provide the incentive to each RN as few as possible. Therefore, there exists an optimal incentives for stimulating the RNs’ multi-service forwarding, at which the objective function is minimized whilst satisfying the constraints. Here, the optimal incentive of RN θi∗ (∀θi∗ ∈ θ ∗ ) for forwarding service h is expressed as

E ωiE (1−e− S,i,h λh ) Pi 1 ∗ ρi,h = − E ln − . I,i (1−ωiE ) (1−ωiE )π S˜iI E i,h (20) ∗ increases as the DeIS From Eq. (20), the optimal incentive ρi,h of each RN within paht θ ∗ decreases. V. S IMULATION R ESULTS In this section, our simulation results are provided for characterizing the RN’s node-selfishness management and effectively demonstrating both the path selection and the path reliability of delivering multi-services. A. RN’s Node-Selfishness Management Figs. 4 and 5 show the effects of the RN’s intrinsic and extrinsic factors on its DeIS and DeES under the framework of the node-selfishness management, respectively. We set the RN’s total available bandwidth and its total residual energy as 10 Mbps and 50 Joules. Fig. 4 depicts the RN’s DeIS affected by its available bandwidth and its residual energy. When the RN is lack of the available bandwidth and the residual energy, its DeIS S I = 1. In contrast, its DeIS S I = 0 if it possesses the sufficient available bandwidth and residual energy. Meanwhile, the RN’s DeIS increases as its available bandwidth or residual energy decreases. Fig. 5 shows the RN’s DeES affected by the service impact factor and the incentive. While the incentive ρ = 0 and with the sufficient service impact factor, the RN’s DeES S E = 1. In contrast, the RN’s DeES S E = 0 if the service impact factor λ = 0 and the sufficient incentive. Meanwhile, the RN’s DeES increases as the service impact factor increases or the incentive decreases. Based on the time correlation and fluctuation of the RN’s behaviors, we develop two kinds of the RN’s behaviors of forwarding multi-services, such as Bi,h,1 = {Bi,h,t = 1|0 ≤ t ≤ T − 1} and Bi,h,0 = {Bi,h,t = 0|0 ≤ t ≤ T − 1}. Fig. 6 shows the variation of the extracted DeNS under three forgetting factors: τ in Eq. (24), τ = 1 and τ = 0.8 together


Fig. 4. The RN’s DeIS is affected by its available bandwidth and its residual energy under the framework of the node-selfishness management.

Fig. 5. The RN’s DeES is affected by the service impact factor and the incentive.

1139

Fig. 7. There is an SeWN topology consisting of one source-destination pair as well as five selfish RNs. Four selfish RNs have confirm NSI, and RN 1 has dynamic NSI.

Fig. 8. The index and reliability of the selected path versus the dynamic DeIS of RN 1.

during the intervals from 100 to 120 and from 200 to 220, the DeNS of τ in Eq. (24) is higher than that of τ = 1 during the interval from 100 to 120, whereas it reverses during the interval from 200 to 220, owing to the fact that the extraction method with τ in Eq. (24) has the function of maintaining the RN’s original willingness.

B. Path Selection and Path Reliability Fig. 6. The variation of the extracted DeNS with time under a certain behavior information of forwarding multi-services.

with the time-window of length T = 20. The curve of legend “Behavior” illustrates the behaviors of dropping a certain service during the period from 1 to 100, forwarding it from 101 to 200 and dropping it from 201 to 300. The forgetting factor τ = 1 represents that all RN’s behaviors are equally weighted for the DeNS extraction, while τ = 0.8 represents that the latest behavior is weighted by a larger value than the previous ones. Meanwhile, τ in Eq. (24) is employed to adaptively weight the RN’s behaviors. As the RN’s behaviors are the same within the time window of length T , the DeNS curves of τ in Eq. (24) and τ = 1 are overlapped, and the DeNS curve of τ in Eq. (24) is closer to the “Behavior” curve than that of τ = 0.8, because the extraction methods with τ in Eq. (24) and τ = 1 exploit more behavior information than that with τ = 0.8. However, in the case that the RN’s behaviors are varying in the time window

An SeWN topology consisting of one source-destination pair and eight selfish RNs includes Path 1 consisting of RN 1, RN 2 and RN 3, Path 2 consisting of RN 4 and RN 5, and Path 3 consisting of RN 6, RN 7 and RN 8, which is shown in Fig. 7. In this topology, RN 1 has dynamic NSI with its DeNS being 0.05 and its DeIS varying, and the other seven selfish RNs have fixed NSI. Here, the NSI of all RNs has been obtained by the source in terms of their historical behaviors and the recommended NSI under the distributed framework of the node-selfishness management. For delivering multi-services, the source obtains the existing path to destination by a traditional routing protocol, and then selects the best path by three methods, i.e., the DSR, the proposed path selection criterion of the RNs’ DeISs, and as well the criterion of the RNs’ DeESs. Meanwhile, the RNs’ available resources are the main principal elements for delivering multi-services, thus it is reasonable to measure the path reliability in terms of the RNs’ DeISs. Fig. 8 shows the path index and path reliability versus the dynamic DeIS of RN 1. The above subfigure provides the index

1140

Fig. 9. The sum of optimal incentives versus the DeIS of RN 1.

of the selected path. In the traditional routing protocol, i.e., DSR, the source selects path 2 for delivering multi-services owing to that this path only includes two RNs and is the shortest path. In terms of the RNs’ DeNSs, the source always selects path 1, which has the smallest DeNSs in all paths. By using the proposed path selection criterion, the source selects path 3 from path 1 when S I = 0.15, as the DeIS of RN 1 increases. The below subfigure shows the reliability of the selected path. By using the DSR, the source selects path 2, which has fixed NSI, thus the reliability of path 2 is fixed. In contrast, the reliability of path 1, selected by using the criterion of the RNs’ DeNSs, decreases as the DeIS of RN 1 increases. By using the proposed criterion, the path reliability decreases as the DeIS of RN 1 increases from 0 to 0.1 owing to path 1 being selected and then, when the DeIS of RN 1 increases from 0.15 to 1, the path reliability is fixed owing to path 3 being selected. Meanwhile, by comparing these three methods, the path reliability in the proposed criterion is larger than that in the other criterions. When one path is selected in a criterion for effectively delivering multi-services, the source provides optimal incentives to stimulate the RNs’ behaviors. Fig. 9 shows the sum of optimal incentives provided for the RNs of the selected path versus the DeIS of RN 1 for delivering the service of λ = 0.0064. The sum of optimal incentives is unchanged owing to path 2 being selected in the criterion of DSR, while the sum of optimal incentives increases owing to path 1 being selected in the criterion of the RNs’ DeNS. Additionally, in our proposed criterion, the sum of optimal incentives increases when the DeIS of RN 1 increases from 0.05 to 0.1 and the sum of optimal incentives is unchanged when the DeIS of RN 1 increases from 0.15 to 1. Meanwhile, the sum of optimal incentives in our proposed criterion is smaller than that in the other two criterions. Fig. 10 plots the RN’s DeIS versus the optimal incentive provided by the source for delivering multi-services of the impact factors λ = {0.0031, 0.0064, 0.0170}. The optimal incentive provided by the source increases as the DeIS of the RN increases. Meanwhile, for a certain RN, when the service of higher impact factor is delivered by the source, the more optimal incentive is provided. Fig. 11 shows the DeNS of the RN versus its DeIS for forwarding three services of the impact factors λ = {0.0031, 0.0064, 0.0170}, respectively. The DeNSs of the RN is unchanged regardless of its DeIS, and its DeNSs are also the same while forwarding different services.


Fig. 10. The optimal incentive versus the DeIS of the RN for forwarding multiservices.

Fig. 11. The DeNS of the RN versus its DeIS for forwarding multi-services.

Fig. 12. The related errors of the optimal incentives versus the DeIS for forwarding multi-services.

Under the distributed framework of the node-selfishness management, the DeNS and DeIS extractions may generate some errors for the RN’s true DeNS and DeIS. Since the source adjusts the incentives in terms of the extracted RNs’ DeISs, the optimal incentives may generate some errors owing to that these extracted RNs’ DeISs may have some errors. Fig. 12 shows the relative errors of the optimal incentives versus the extracted DeIS when the RN’s true DeISs are 0.3, 0.5 and 0.8. The relative errors of the optimal incentives are small owing to that the error of the extracted NSI is small observed in Fig. 6.


VI. C ONCLUSIONS In this paper, we have constructed a distributed framework of the node-selfishness management, where every RN manages its NSI and other nodes’ NSI and every source manages the RNs’ NSI in distributed SeWNs. In this framework, the RN’s models of intrinsic and extrinsic selfishness have been developed to manage its DeIS and DeES, and the other RNs’ NSI has been obtained in terms of the RNs’ historical behaviors and their recommended NSI. Under this distributed framework of the node-selfishness management, the path selection criterion has been designed to select the most reliable and shortest path for the multi-service delivery. Additionally, the optimal incentives have been adjusted by the source for maintaining the path reliability of the E2E multi-service delivery. APPENDIX A D E NS E XTRACTION The RN’s NSM employs the beta distribution to extract the other RNs’ DeNSs. The beta distribution has been used to build the reputation system in [20], and the mathematical analysis has been developed on probability theory [12]. The results of the beta distribution and the RN’s DeNS extraction show as follows. The beta-family of probability density function is a continuous family of functions with the two parameters α and β. The beta distribution f ( p| α, β) is expressed using the Gamma function as f ( p| α, β) =

(α + β) α−1 p (1 − p)β−1 . (α)(β)

(21)

where 0 ≤ p ≤ 1, α >0 and β > 0, and the Gamma function +∞ is expressed as (z) = 0 e−t t z−1 dt . The expectation of the beta distribution is given by 1 α . (22) p = E( p) = p × f ( p| α, β)d p = α+β 0 The beta distribution is introduced to determine the behaviors whether the RNs forward multi-services. The behavior information is expressed as 1 meaning forwarding multi-services and 0 meaning dropping multi-services. For RN i, the behavior numbers of forwarding and dropping service h are denoted by u i,h and vi,h , respectively. By α = u i,h + 1, β = vi,h + 1 together with Eq. (21), the posteriori probability density function of the i-th RN for forwarding service h is expressed as

(u i,h+vi,h+2) (pi,h)u i,h (1− pi,h )u i,h , f (pi,h u i,h ,vi,h) = (u i,h+1)(vi,h+1) (23) where 0 ≤ pi,h ≤ 1 is the probability of RN i for forwarding service h. For the DeNS extraction, a time window and a forgetting factor are introduced to deal with the time correlation and the fluctuation of the RN’s historical behaviors. Within the time window of length T , the node-selfishness management directly observes the behaviors of RN i for forwarding service h, denoted by Bi,h = {Bi,h,t |0 ≤ t ≤ T − 1}. At time t,

1141

the behavior Bi,h,t = 1 means that RN i forwards service h, and the behavior Bi,h,t = 0 means that RN i drops service h. The behavior Bi,h,0 represents the latest behavior of RN i for service h, while the behavior Bi,h,T −1 is its earliest behavior in this time window. Additionally, we use the variance of the observed behaviors of RN i for forwarding service h to describe this forgetting factor, expressed as τi,h = 1 − Var(Bi,h ),

(24)

where Var(Bi,h ) is the variance value of the historical behaviors of RN i for forwarding service h. The node-selfishness management weights the historical behaviors of RN i for forwarding service h by (τi,h )t at time t of the time window, and the behavior weight (τi,h )t increases as the time instance of this time window decreases. Within this time window of length T , the time-weight numbers of forwarding service h and dropping service h as T −1 B (τ )t u i,h = t=0 T −1 i,h,t i,h (25) vi,h = t=0 (1 − Bi,h,t )(τi,h )t , where Bi,h,t is the behavior of forwarding service h of RN i at time t, τi,h is the forgetting factor to weight the behavior of forwarding service h of RN i. By Eq. (22), the extracted DeNS of RN i for forwarding service h is defined as

E = 1 − E pi,h u i,h , vi,h = 1 − Sˆi,h

u i,h + 1 , u i,h + vi,h + 2

(26)

where E( pi,h u i,h , vi,h ) is an expected value with the posteriori probability density function of Eq. (23). By plugging Eqs. (25) into Eq. (26), we obtain the extracted DeNS of RN i. R EFERENCES [1] Y. Xiao and H. Li, “Voice and video transmissions with global data parameter control for the IEEE 802.11e enhance distributed channel access,” IEEE Trans. Parallel Distrib. Syst., vol. 15, no. 11, pp. 1041– 1053, Nov. 2004. [2] M. v. d. Schaar, Y. Andreopoulos, and Z. Hu, “Optimized scalable video steaming over IEEE 802.11a/e HCCA wireless networks under delay constraints,” IEEE Trans. Mobile Comput., vol. 5, no. 6, pp. 755–768, Jun. 2006. [3] J. Li, Q. Yang, K. S. Kwak, and L. Hanzo, “The connectivity of selfish wireless networks,” IEEE Access, vol. 3, pp. 2814–2827, Nov. 2015. [4] J. Li, Q. Yang, and K. S. Kwak, “Neural-network based optimal dynamic control of delivering packets in selfish wireless networks,” IEEE Commun. Lett., vol. 19, no. 12, pp. 2246–2249, Dec. 2015. [5] H. Jiang and W. Zhuang, “Cross-layer resource allocation for integrated voice/data traffic in wireless cellular networks,” IEEE Trans. Wireless Commun., vol. 5, no. 2, pp. 457–468, Feb. 2006. [6] R. Dai, P. Wang, and I. F. Akyildiz, “Correlation-aware QoS routing with differential coding for wireless video sensor networks,” IEEE J. Multimedia, vol. 14, no. 5, pp. 1469–1479, Oct. 2012. [7] L. Cheng, J. Niu, J. Cao, S. K. Das, and Y. Gu, “QoS aware geographic opportunistic routing in wireless sensor networks,” IEEE Trans. Parallel Distrib. Syst., vol. 25, no. 7, pp. 1864–1875, Jul. 2014. [8] F. Bao, I. Chen, M. Chang, and J. Cho, “Hierarchical trust management for wireless sensor networks and its applications to trust-based routing and intrusion detection,” IEEE Trans. Netw. Serv. Manage., vol. 9, no. 2, pp. 169–183, Jun. 2012. [9] I. Chen, F. Bao, M. Chang, and J. Cho, “Dynamic trust management for delay tolerant networks and its application to secure routing,” IEEE Trans. Parallel Distrib. Syst., vol. 25, no. 5, pp. 1200–1210, May 2014.

1142


[10] R. A. Shaikh, H. Jameel, B. J. d’Auriol, H. Lee, S. Lee, and Y. Song, “Group-based trust management scheme for clustered wireless sensor networks,” IEEE Trans. Parallel Distrib. Syst., vol. 20, no. 11, pp. 1698– 1712, Nov. 2009. [11] H. Zhu, X. Lin, and R. Lu, “SMART: A secure multilayer creditbased incentive scheme for delay-tolerant networks,” IEEE Trans. Veh. Technol., vol. 58, no. 8, pp. 4628–4639, Oct. 2009. [12] P. Kyasanur and N. F. Vaidya, “Selfish MAC layer misbehavior in wireless networks,” IEEE Trans. Mobile Comput., vol. 4, no. 5, pp. 502–516, Sep. 2005. [13] Z. Ji and K. J. R. Liu, “Multi-stage pricing game for collusion-resistant dynamic spectrum allocation,” IEEE J. Sel. Areas Commun., vol. 26, no. 1, pp. 182–191, Jan. 2008. [14] Y. Rebahi, V. E. Mujica-V, and D. Sisalem, “A reputation-based trust mechanism for ad hoc networks,” in Proc. IEEE Symp. Comput. Commun., Jun. 2005, pp. 37–42. [15] J. Li, Q. Yang, and F. Fu, “Game theoretic approach for enforcing truthtelling upon relay nodes,” IEICE Trans. Commun., vol. E94-B, no. 5, pp. 1483–1486, May 2011. [16] C. E. Perkins, E. M. Royer, S. R. Das, and M. K. Marina, “Performance comparison of two on-demand routing protocols for ad hoc networks,” IEEE Pers. Commun., vol. 8, no. 1, pp. 16–28, Feb. 2001. [17] D. You, C. F. Benitez-Quiroz, and A. M. Martinez, “Multiobjective optimization for model selection in kernel methods in regression,” IEEE Trans. Neural Netw. Learn. Syst., vol. 25, no. 10, pp. 1879–1893, Oct. 2014. [18] S. Mohamed and G. Rubino, “A study of real-time packet video quality using random neural networks,” IEEE Trans. Circuits Syst. Video Technol., vol. 12, no. 12, pp. 1071–1083, Dec. 2002. [19] X. Qiu and K. Chawla, “On the performance of adaptive modulation in cellular systems,” IEEE Trans. Commun., vol. 47, no. 6, pp. 884–895, Jun. 1999. [20] A. Josang, “The beta reputation system,” in Proc. 15th Bled Electron. Commerce Conf. e-Reality: Construct. e-Economy, Jun. 2002. [21] R. Meester and P. Trapman, “Bounding basic characteristics of spatial epidemics with a new percolation model,” Adv. Appl. Probab., vol. 43, no. 2, pp. 335–347, 2008. [22] E. Ataie and A. Movaghar, “Performance evaluation of mobile ad hoc networks in the presence of energy-based selfishness,” in Proc. Int. Conf. Broadband Commun. Netw. Syst., Oct. 2006, pp. 1–6.

Jinglei Li received the B.S. degree in electronic information engineering from the PLA Information Engineering University, Zhengzhou, China, in 2008, and the M.S. degree in communication and information systems from Xidian University, Xi’an, China, in 2011. He is currently pursuing the Ph.D. degree in communication and information systems at Xidian University. His research interests include wireless network connectivity and node selfishness analysis.

Qinghai Yang received the B.S. degree in communication engineering from Shandong University of Technology, Zibo, China, in 1998, the M.S. degree in information and communication systems from Xidian University, Xi’an, China, in 2001, and the Ph.D. degree in communication engineering from Inha University, Incheon, South Korea, in 2007. From 2007 to 2008, he was a Research Fellow with the UWB-ITRC, Korea. Since 2008, he has been with Xidian University. His research interests include autonomic communication, content delivery networks, and LTE-A techniques. He was the recipient of the UniversityPresident Award from Inha University.

Peng Gong received the B.S. degree in mechatronical engineering from Beijing Institute of Technology, Beijing, China, in 2004, and the M.S. and Ph.D. degrees from the Inha University, Incheon, South Korea, in 2006 and 2010, respectively. In July 2010, he joined the School of Mechatronical Engineering, Beijing Institute of Technology, China. His research interests include network emulation, link/system level performance evaluation, and radio resource management in wireless systems such as 3GPP LTE, UWB, MIMO, and cognitive radio.

Kyung Sup Kwak (M’81) received the B.S. degree from the Inha University, Incheon, South Korea, in 1977, and the M.S. degree from the University of Southern California, Los Angeles, CA, USA, in 1981, and the Ph.D. degree from the University of California at San Diego, San Diego, CA, USA, in 1988. From 1988 to 1989, he was a member of technical staff with Hughes Network Systems, San Diego, CA, USA. From 1989 to 1990, he was with the IBM Network Analysis Center at Research Triangle Park, NC, USA. Since then he has been with the School of Information and Communication, Inha University, Incheon, South Korea, as a Professor. He was the Chairman of the School of Electrical and Computer Engineering from 1999 to 2000, and the Dean of the Graduate School of Information Technology and Telecommunications with Inha University from 2001 to 2002. He is the current Director of Advanced IT Research Center, Inha University and UWB Wireless Communications Research Center, a key government IT research center in South Korea. His research interests include multiple access communication systems, mobile communication systems, UWB radio systems and ad-hoc networks, and high-performance wireless Internet. He is a Member of IEICE, KICS, and KIEE. He was the recipient of Inha University Fellowship and the Korea Electric Association Abroad Scholarship.