Cooperative Game-Based Resource Allocation Scheme for Heterogeneous Networks with eICIC Technology

,

Motivated by the abovementioned advantages and characteristics of HetNets, many studies on HetNets have begun to be investigated. One of the main challenges in HetNets is the macrocell to small-cell interference. In HetNets, SBSs are characterized by low transmission power. Generally, most IoT devices refer to the maximal signal to interference and noise ratio (SINR) criterion to select their corresponding cells. Terefore, IoT devices may connect to a distant MBS due to its high transmission power rather than to any nearby SBS. It leads to an inefcient load distribution. Due to this reason, it is clear that the benefts of HetNets strongly depend on an efcient resource management between high-power MBS and low-power SBS [3][4][5].
For the resource management of HetNets, the 3rd Generation Partnership Project (3GPP) has proposed the notion of enhanced intercell interference coordination (eICIC). As a time-domain solution, the eICIC uses two mechanisms, i.e., Cell Range Extension (CRE) and Almost Blank Subframes (ABSs), to protect the small-cell transmissions by mitigating the interference from the MBS. Te CRE is a technique to expand a small-cell range virtually by adding a bias value to the received power; it can be realized by performing user association. Terefore, small-cell coverage, cell-edge throughput, and overall network throughput can be improved. Te ABS is a technique to mitigate the interference experienced by small-cell devices in cell range expansion region. During certain time intervals, the MBS is muted except for overhead signal transmissions to improve SINRs to the small-cell devices. Tese intervals are called ABS. Tus, multiple IoT devices can transmit to their corresponding SBSs over ABS with very little interference from the MBS [3][4][5].
To implement the eICIC technology, joint resource allocation among MBSs and SBSs is vitally important to reduce the mutual interference and achieve spectrum sharing. Since the proportion of ABSs is closely related to the association of device and the resource allocation scheme adopted by the SBSs, it is of critical importance that the proportion of ABSs, user association, and the resource allocation decisions should be jointly optimized so as to achieve the HetNet performance gain. In this paper, we aim to design a novel control scheme for the joint resource allocation problem involving the following three fundamental issues: (i) how to determine user association for load balancing, (ii) how to decide the optimal proportion of ABSs and which devices are allocated on ABSs or non-ABSs in time domain, and (iii) how to allocate spectrum resource in frequency domain. Tese three control issues are tightly coupled with each other. To optimize the system performance and realize the potential of HetNets, it is necessary to study these control problems jointly [6,7].
Cooperative optimization is inspired by the principle of coordination work, which considers the interactions among agents to optimize their payofs through their objective functions. It has a solid theoretical foundation and outstanding practical performance, which has aroused widespread attention and research interest. As a branch of cooperative optimization, cooperative game theory is concerned with the game players forming alliances and working together to achieve their common goals. In the cooperative games, how to decompose the reward generated by the grand alliance to each participant has always been a hot research topic. At present, cooperative game models are present in almost every feld of research such as economics, biology, sociology, politics, computer science, and telecommunications. In this study, we adopt the basic idea of cooperative game theory to develop a novel resource control scheme for the eICIC-based HetNet platform. From the perspective of coordination, the MBS, SBSs, and IoT devices cooperate and difuse information so as to obtain a globally optimal solution to maximize the HetNet system performance [8,9].
Te rest of this article is organized as follows. In Section 2, we describe the basic ideas of cooperative game solutions and main contributions of this study. Section 3 summarizes the state-of-the-art research papers related to the control issues for eICIC-based HetNets. Section 4 presents the scenario of HetNet system platform and formulates the control problem. And then, our proposed scheme is elaborated in detail by using diferent solution concepts. To help readers understand better, we also provide the primary steps of the proposed algorithm. Section 5 presents the testbed implementation setup and simulation results are shown along with the achieved experimental results. Finally, a conclusion is drawn and future research plans are explained in Section 6.

Technical Concepts and Main Contributions
In a cooperative game, multiple players have to agree on a feasible utility allocation. Te key idea is that players can achieve superior outcomes by working together rather than working against each other. In 1950, it was originated in a fundamental paper by J. Nash. He introduced an idealized representation of the cooperative problem and developed a methodology that gave the hope that the uncertainty of cooperation could be resolved. Nash solution for cooperative games is that under certain axioms, there is a unique solution. However, one of his axioms, i.e., independence of irrelevant alternatives (IIAs), came under criticism. Since then, diferent solution concepts have been provided with alternative axiom; they also lead to another unique solutions. Te most famous of them is the Kalai and Smorodinsky solution (KSS). It is suggested as an alternative to Nash's solution, which was suggested 25 years earlier. Te main diference between the two solutions is that the Nash solution satisfes the IIA axiom, while the KSS solution satisfes the axiom of monotonicity [9,10].
Recently, multicriteria cooperative games are concerned with situations in which a number of players must take into account several criteria, each of which depends on the decision of all players. A multicriteria cooperative game is a generalization of the classic cooperative game, where each player has a set of criteria for the evaluation of any decision. In these situations, there are two decision issues to be jointly considered: one related to the preferences of individual players with respect to their own criteria and the other 2 Journal of Computer Networks and Communications related to the problem of selecting a solution that could be accepted by all the rational game players. However, until now, the literature on multicriteria cooperation is scarce. In 2009, the idea of multicriteria Kalai and Smorodinsky solution (MCKSS) was introduced while maintaining the multidimensional nature of each player's payof. An important feature of MCKSS is that the set of feasible outcomes is a polyhedron, and as a consequence, if the solution can be rationalized by means of linear functions, then the outcomes induced by the solution can be computed by solving multicriteria linear problems [11]. As another cooperative game solution, H. Raifa introduced the concept of sequential Raifa solution (SRS) for a two-player cooperative problem. Based on the arbitration protocol, the SRS can be characterized by the standard cooperative game axioms. In addition, an additional specifc axiom expresses the key concept of repeated negotiation of the same procedure to the decreasing sequence of remaining games. Given a disagreement point, the most preferred outcome for a player is the one that gives his maximal utility while keeping the other player at his disagreement payof. Te interim agreement is the average of these two, most preferred points. By using each interim agreement as a new disagreement point, the SRS sequentially repeats this cooperative game process and converges to a Pareto optimal point of the cooperative game set. By doing so, the SRS bridges the gap between cooperative and noncooperative game processes via relative gains and concessions [12,13].
In this study, we aim to optimize the performance of eICIC-based HetNets. Based on the KSS, MCKSS, and SRS, our proposed scheme is designed as a joint control paradigm in a step-by-step repeated manner. During the interactions between the MBS and SBSs, the situation of multicriteria cooperation arises. Terefore, we formulate the MBS-SBSs control problem as a multicriteria cooperative game. In the viewpoint of each individual SBS, control issues can be modeled as a two-player cooperative game. For individual IoTdevices, the assigned resource can be shared according to the single-criterion cooperative game. In the proposed scheme, these three cooperative game models are sophisticatedly combined in a dynamic interactive manner and work together during HetNet operations. In the eICIC-based HetNet system infrastructure, our joint control approach can harness the synergies of cooperation among MBS, SBSs, and IoT devices to achieve a socially optimal solution. Te signifcant major contributions of our study are summarized as follows: (1) To maximize the resource efciency in the HetNet platform, we design a joint resource allocation scheme based on cooperative games. To the best of our knowledge, our cooperative game-based resource control paradigm is the frst in the literature for the eICIC-based network infrastructure. (2) To decide the fraction ratio for the MBS-SBSs control problem, we formulate a multicriteria cooperative game model and select best strategies based on the idea of MCKSS. To choose the fraction rate for each individual SBS service, we formulate a two-person cooperative game model and adopt the idea of SRS. (3) For multiple IoT devices in the coverage areas of MBS or SBS, the resource distribution problem is solved based on the KSS. In a dynamically changing eICIC-based HetNet environment, three cooperative game processes are operated in a step-by-step repeated manner. (4) Our cooperative game models are jointly combined and work together to get reciprocal advantages. With the eICIC technology, our hybrid approach explores the sequential interactions between MBS, SBSs, and IoT devices and reaches a consensus for the excellent HetNet performance. (5) Extensive simulation results demonstrate the performance superiority of our proposed scheme by comparing to the existing state-of-the-art eICIC control protocols. Especially, the excellency of our cooperative game-based control paradigm is confrmed in terms of system throughput, device payof, and fairness.

Related Work
After 3GPP has proposed the eICIC standard, academic research papers are largely focused on advancing the practical implication of eICIC-based HetNets. In this section, we touch on currently published papers relevant to the research topic of our study. In [14], Deb et al. proposed a novel radio resource sharing (RRS) scheme for eICIC in HetNets. Two important challenges in the RRS scheme are (i) to determine the amount of radio resources for the SBSs and (ii) to determine the device association rules for the SBSs. Te RRS scheme addresses these two tightly coupled control issues in a joint manner. By using the notion of ABS and CRE, the RRS scheme can account for network topology, trafc load, and MBS-SBS interference map to guarantee the HetNet network performance. Terefore, the SBS transmissions are ensured without badly hit by interference from the MBS based on the enhanced intercell interference coordination implementation. Extensive simulations are conducted to study the performance evaluation, and numerical results show that RRS scheme can achieve better reliability and sustainability [14]. Te paper [15] investigates the concept of dualconnectivity, joint device association, and resource allocation for HetNets. Te benefts of dual-connectivity paradigm necessitate the careful device association and resource allocation techniques in the dense HetNet scenario. To control the problem of intertier interference, the dual-connectivity resource allocation (DRA) scheme is developed. First, in dual-connectivity enabled HetNets, the joint problem of device association and resource allocation is formulated as a mixed integer nonlinear programming problem. Second, by linearizing the products of the variables, the original problem is reformulated as a tractable linear problem, which is optimized to maximize the overall network throughput. Tird, a new low-complexity heuristic method is designed to Journal of Computer Networks and Communications achieve the near-optimal system solution. Last, numerical results confrm that the DRA scheme outperforms in the average system throughput with less computation-time and low complexity [15].
Te joint association and fairness (JAF) scheme is designed for the eICIC-based HetNet. [16]. Tis scheme concentrates on the joint device association and interference coordination problem, and it is formulated as a network-wide max-min fairness problem. However, it is a mixed integer nonlinear programming problem, which is NP-hard. To overcome this computation difculty, this NP-hard problem is decomposed into two phase subproblems. Te frst subproblem is defned as the ABS ratio problem for given device association; it fgured out the explicit expression of the optimal ABS ratio. Based on this derived result, the second subproblem is defned to maximize network utility through the device association. Based on the marginal utility, a new low complexity greedy algorithm with polynomial complexity is proposed to solve the second subproblem. Finally, simulation results show that the JAF scheme has good performance in terms of system throughput and fairness among mobile devices [16].
As discussed above, the earlier schemes in [14][15][16] have been studied on the resource allocation problem for the eICIC-based HetNet. Although these studies tackled the same control issues, which we concern in this study, they did not consider a joint control paradigm based on the cooperative game theory. Unlike the aforementioned RRS, DRA, and JAF schemes, our proposed scheme concerns the combination of diferent cooperative game solutions for controlling the activities of MBS, SBSs, and mobile IoT devices and guides them toward a fair-efcient outcome in the HetNet platform.

Proposed Resource Control Scheme for eICIC-Based HetNets
Tis section introduces the HetNet platform that we focus on. Ten, we explain the basic ideas of KSS, MCKSS, and SRS, which are adopted to design our proposed joint control scheme. After that, our proposed resource allocation algorithm is described strategically in the nine-step procedures. Individual IoT devices are randomly placed in the cellular network area following a uniform distribution. Base stations carry out the spectrum resource management, scheduling, admit control, and more functionalities for their corresponding devices. As depicted in Figure 1, IoT devices can access diferent base stations, such as MBS or SBS, and we summarize the notations used in this paper in Table 1 [14][15][16].

HetNet System Infrastructure and Cooperative Game
To alleviate the trafc load of MBSs in HetNets, we employ time-domain eICIC mechanism. Assume a same ABS pattern that each MBS has the same ABS ratio α, where α fraction is assigned to ABS and (1 − α) fraction is assigned to non-ABS. MBS maintains silence in the ABS position and transmits at normal power in a non-ABS position. Terefore, each MBS can only use the non-ABSs mode to serve its corresponding devices. However, SBSs can serve their corresponding devices by using the ABS or non-ABS mode. In the HetNet platform, we defne the IoT devices within the nominal coverage of the SBS as SBS-centric devices. By assigning a positive bias to the reference signal received power, the SBS coverage area can be extended. Te devices in the SBS-extended range are defned as SBS-extended devices. Over the non-ABS mode, the SBS-extended devices sufer excessive interference from MBS, while the SBS-centric devices receive relatively less interference [7,[14][15][16].
Without loss of generality, it is assumed that the SBScentric and SBS-extended devices are served by non-ABS and ABS, respectively. However, some SBS-centric devices are seriously interfered by MBS. In order to solve this problem, we confgure a certain proportion of ABS fraction by using ratio factor β. Terefore, the resource scheduling policy of SBS is to serve the SBS-extended devices by using the β proportion of ABS fraction, while (1 − β) proportion of ABS fraction is assigned for the SBS-centric devices. Based on the eICIC technology, the values of α and β are dynamically adjusted to share the limited spectrum resource to maximize the HetNet system. By combining the time domain and the frequency domain, the diagram of spectrum allocation for the MBS and SBS is shown in Figure 2.
During the eICIC operations, each SBS has two type devices. Devices are served on ABS mode (the set denoted by D A S ), and the other devices are served on non-ABS mode (the set denoted by D nA S ). Each MBS only uses the non-ABS mode to serve their corresponding devices, which are denoted by the set D M . Terefore, we have |M| + 2 × |S| virtual base stations in total. Te SINR of D j ∈ D from virtual base station V is given by the following equation [16]: where P V is the transmission power of V, |h D j ,V | 2 is the channel gain from V to D j , and σ 2 is the noise power of each base station. T D j denotes the interfering device set of D j .
When D j is served on ABS mode, Based on the quasistatic fading channel, SINR D j ,V is assumed to be constant during an allocation period. According to the Shannon capacity, the instantaneous communication rate from V to D j , i.e., r D j ,V , can be obtained, and we fnally get the normalized link rate from V to D j , i.e., R D j ,V as follows [3,16]: where W is the channel bandwidth from V to D j . In this paper, we develop three cooperative games, which are hierarchically organized based on the joint control paradigm. Te frst cooperative game (G M S ) decides the ABS ratio α, and the second cooperative game (G S ) decides the proportion ratio factor β for each SBS. Finally, the third cooperative game (G D ) distributes the assigned resource for corresponding IoT devices. Trough the G M S , G S , and G D games, the MBS, SBSs, and devices are sequentially interacted with each other to reach a mutually acceptable solution, which is called social optimal consensus. It is noteworthy that we formulate the M-S-D association in an iterative coordinated manner. Formally, we defne the tuple entities in our proposed G M S , G S , and G D game models, such as (1) M, S, and D represent the sets of MBSs, SBSs, and IoT devices, respectively (2) In the frst game for M i , i.e., G

Technical Concepts and Ideas of KSS, MCKSS, and SRS.
In this subsection, we quickly review the fundamental concepts of KSS, MCKSS, and the sequential Raifa solution for cooperative games.

Multicriteria Kalai and Smorodinsky Solution for
Cooperative Games. To characterize the basic concept of MCKSS, we preliminarily defne some mathematical expressions. Let N � 1, 2, . . . , n { } be the set of players, and each player considers the same m criteria, M � 1, 2, . . . , m { }, to value the possible agreements. An n-person multicriteria cooperative game is formally described by the pair (S, d), where S ⊆ R n×m is the set of all feasible outcomes and d � × n i�1 d i is the disagreement point or status quo. Te outcomes in S are obtained as the result of a joint decision of all the players. Terefore, there exists an agreement X � (x 1 , x 2 , . . . , x n ) ∈ S that gives the player i ∈ N an outcome vector is the payof of player i in criterion j. d is the result obtainable if the players fail to reach an agreement. In a multicriteria cooperative game, individual rationality establishes that each player will only negotiate at or above those outcomes that improve upon the disagreement point d. Tus, the set of outcomes where players will negotiate is d . Usually, a multicriteria cooperative game solution is a procedure to incorporate partial information on the importance of the criteria in multicriteria linear problems. Based on this idea, MCKSS takes into account each player's maximum payofs with respect to the criteria, which are represented by the ideal outcomes; they are denoted as I � (I 1≤j≤m i∈N ). MCKSS is obtained by replacing the payof gains of players by the proportion with respect to their ideal outcomes. Terefore, for each feasible outcome is a major decision factor, where i ∈ N and 1 ≤ j ≤ m. Originally, E. Kalai and M. Smorodinsky introduced the idea of KSS as a single criterion cooperative game solution. It provides a payof that is proportional to the achievable maximum payof while ensuring efciency. Terefore, each player gets the same fraction of his maximum possible payof in KSS. Geometrically, KSS is the intersection point between the cooperative game set S and the straight line between the disagreement point and the utopian point. By considering single criterion X and I, the KSS, i.e., X KSS � (x * 1 , x * 2 , . . . , x * n ), must satisfy the following equation [10,11]: To defne MCKSS, let P(X) denote the minimum proportional payof vector whose components are P j (X) � min 1≤i≤n For each feasible outcome X ∈ S, the minimum utility gains vector is z(X) � (z 1 (X), z 2 (X), . . . , z m (X)), where z 1≤j≤m (X) � min 1≤i≤n x j i − d j i . Each component of z(X) represents the guaranteed minimum utility gains of the set of players for the corresponding criterion. MCKSS is based on the idea that the players jointly agree on those outcomes whose minimum utility gain levels cannot be simultaneously improved with respect to all the criteria. Terefore, the players will choose an outcome such that there is no other outcome whose minimum utility gain vector is better in terms of components. For the multicriteria cooperative game (S, d), a feasible outcome X ∈ S ∩ d ≥ is MCKSS, if there does not exist Y ∈ S ∩ d ≥ such that P(Y) ≥ P(X). Mathematically, MCKSS, i.e., X MCKSS � (x 1 , x 2 , . . . , x n ), is given by the following equation [11]:

Journal of Computer Networks and Communications
s.t., ⋮, Terefore, MCKSS is the Pareto optimal by criteria in S ∩ d ≥ and the set of efcient solutions for multicriteria cooperative game (S, d).

Sequential Raifa Solution for Cooperative Game.
To characterize the fundamental idea of SRS, we assume a two-player cooperative game problem. R is a real number set, and R 2 denotes the two-dimensional Euclidean space. Let S ⊂ R 2 be a nonempty and fnite set, and N is a set of natural numbers. For any x, y ∈ R 2 , we write x ≥ y (x ≫ y, resp.) if for any i ∈ S, Te set of all two-person cooperative games is denoted as B. For any nonempty subset C of B, a mapping F: For any game (S, d) ∈ B, we consider the mapping f (S,d) ∶R 2 ⟶ R 2 ; it defned as follows [13]: Now consider the sequence (m (S,d) k ) k∈N� 1,2,3,... { } defned by the following equation [13]: According to (6) and (7), SRS(S, d) is defned by the following equation [13]: For the game player M i , we defne his utility function U M i (·). However, U M i (·) consists of two subfunctions for the criteria of sensitivity (U S M i (·)) and productivity

Te Proposed Joint Cooperative Game
. Tey are given as follows: To decide the β S j value in S j , we construct the G S j game.
In this game, D where κ is a control parameter for the U where d e and d c are the disagreement points for D is within a predefned minimum bound (φ), this change can be negligible. At this time, we can think that we converge a fair-efcient solution, and our negotiation process is terminated. Finally, we can get the β S j value, and the (α For the devices in D, the G D game is designed to distribute the assigned resource. In  , its utility function is defned as follows: where R D k and %Γ D k are the requested and allocated resource amounts for D k , respectively. Based on the corresponding devices, the G , is obtained as follows: where I D k is the ideal outcome for the D k . For the G S j e and G S j c games, the utility function for each game player is defned as the same by using (15). Finally, devices in D

Main Steps of Our Hierarchical Joint Cooperative Game
Algorithm. By deploying diferent network infrastructures, such as MBSs and dense deployment of SBSs, the HetNet platform for IoT paradigm is widely seen as a solution for the 5G mobile networks. However, the performance of SBSs could be impacted by the interference from the MBS. To protect the SBS transmissions by mitigating the MBS interference, the eICIC technology is introduced. In this study, we design a hierarchical game model for the eICIC-based HetNet system. Our proposed game consists of diferent cooperative games such as G M S , G S , and G D . In these games, game players can obtain, through cooperation, better outcomes than those obtained when they do not cooperate. Te G M S , G S , and G D games are operated in a distributed parallel fashion according to MCKSS, SRS, and KSS. In dynamically changing HetNet system environments, our joint control paradigm is especially appropriate to negotiate conficting requirements. Te primary steps of the proposed scheme are described as follows: Step 1: To carry out the simulation analysis, the parameter and control factor settings are shown in Table 2. Our simulation scenario is described in Section 5.
Step 2: At a sequence of time steps, the G M S , G S , and G D games are operated sequentially and interactively to reach a mutually consensus.
Step 3: First, the G M S game is designed as a two-player game with two decision criteria. In this game, M and S M are game players, and their utility functions are defned using (9) and (10).
Step 4: For the G M S game, MCKSS is adopted as a solution; it is obtained based on the equations (5) and (11). Finally, the α M value is calculated by using (12).
Step 5: Second, the G S game is designed as a single criterion two-player game. In this game, D S e and D S c are the game players, and their utility functions are defned using (13). For the G S game, the SRS is adopted as a solution. Finally, the β S value is obtained based on the equations (5)- (7) and (14).
Step 6: Tird, the G D game consists of three subgames such as G  (4) and (16).
Step 8: In our joint control approach, the G M S , G S , and G D games are sophisticatedly combined and work together to achieve reciprocal advantages.
Step 9: Constantly, individual MBS, SBSs, and IoT devices are self-monitoring the current eICIC-based HetNet system condition, and it proceeds to Step 2 for the next game iteration.

Performance Evaluation
Tis section presents the numerical results derived to validate the proposed method. By comparing with the existing RRS, DRA, and JAF schemes [13,14,15], we confrm that our approach outperforms other schemes via Matlab simulations. System parameters and their values are listed in Table 2, and the simulation environment and system scenario are given as follows [17,18]: (1) Simulated the eICIC-based HetNet system platform consists of fve MBSs, twenty-fve SBSs, and twohundred ffty IoT devices, i.e., |M| � 5, |S| � 25, and |D| � 250 (2) Each MBS has fve SBSs, and IoT devices are randomly distributed over in the MBS covering area (3) Each IoT device D 1≤i≤l generates six diferent type service requests (%Γ D ) (4) Te arrival process of %Γ D is the rate of the Poisson process (ρ). Te ofered range is varied from 0 to 3.0. (5) Te total spectrum resource of each MBS (M M ) is 500 Gbps (6) We assume the absence of physical obstacles in the network area (7) Te resource allocation process through the cooperative games is specifed in terms of basic To evaluate the proposed scheme, we compare its performance in terms of system throughput, IoT device payof, and fairness over ofered service request generation ratios. Table 2 shows the control parameters and system factors used in the simulation. Figure 3 shows the system throughput in the eICIC platform as a function of IoT device service request ratios. It is observed that our joint game approach consistently outperforms all other existing protocols. Tis is expected because when our proposed scheme is operated, the MBS SBSs, and devices work together to improve the resource efciency. Especially, MCKSS is applied between the MBS and SBSs, SRS is adopted for each SBS's intraresource sharing process, and KSS is used to distribute the assigned resources for each device. Terefore, there will be more room to efectively exploit an available resource in order to maximize the overall system capacity. Based on the desirable features, our cooperative game model can guide selfsh network agents to efectively share their limited resources in a coordinated manner. So, we can attain the highest system throughput among the four diferent schemes. Figure 4 reports the normalized device payof achieved by all considered schemes under diferent workload ratio. Compared with the other existing schemes, IoT devices in our proposed scheme adaptively share the resource in a step-by-step dynamic coordinated fashion. Finally, the assigned resource for each diferent device set is adaptively distributed based on the KSS; it ensures a fair-efcient solution for the resource allocation problem. Terefore, our approach is quite fexible to maximize the device payof under widely diferent and diversifed eICIC-based HetNet situations. Due to this reason, it can be seen that the proposed scheme achieves a higher device payof over the existing RRS, DRA, and JAF protocols.
For diferent protocols, the Jain's fairness index of IoT devices is shown in Figure 5. It can be used to evaluate the degree of fairness, and a larger Jain's index corresponds to a fairer allocation. As shown in Figure 5, our proposed scheme outperforms the RRS, DRA, and JAF protocols by  a large margin. In the proposed scheme, individual IoT devices can dynamically adjust their strategy through the cooperative game process, which is implemented as a hierarchical interactive manner. Terefore, IoT devices can collectively capture how to adapt their strategies to achieve the better beneft while getting reciprocal advantages. Terefore, all devices fairly share the limited resource in the eICIC-based HetNet system. Simulation results confrm the excellency of our proposed method for the fairness index; we maintain the highest fairness compared to other methods.

Summary and Conclusion
Traditionally, network trafc load in the cellular network varies continuously as a result of user mobility and trafc dynamics. Te HetNet architecture is one of the most promising solutions to adaptively handle trafc fuctuations while enhancing the network coverage, spatial spectrum reuse, and system throughput. In this paper, we investigate a joint resource allocation scheme for the eICIC-based HetNet system, which ofers several benefts, including a more equitable distribution of network trafc across the MBS and SBSs. For the eICIC technology, previous work did not consider dynamically changing network conditions in HetNets. Terefore, they studied the benefts of the eICICbased HetNet with static optimization of eICIC parameters. Our proposed scheme is dynamic and can improve the system performance by using a hierarchical and repeated cooperative game model. Especially, we take into account both the ratio of non-ABS for MBS devices and the ratio of ABS fraction for SBS-extended devices. Tese ratios are adaptively adjusted based on the concept of MCKSS and SRS through the G M S and G S game models. And then, the assigned resources for MBS, SBS-centric, and SBSextended devices are fair-efciently distributed by using the idea of KSS through the G D game. By tightly coupling the G M S , G S , and G D games, we can lead to consensus outcomes that may be accepted by all IoT devices of MBS and SBSs. Te main novelty of our joint game approach is to explore the reciprocal advantages through sequential interactions among the MBS, SBSs, and IoT devices. Under dynamically changing eICIC-based HetNet environments, our proposed scheme achieves a socially optimal solution in a step-by-step repeated manner. Te extensive simulation results demonstrate the efciency of the proposed algorithm and verify the superiority of our hierarchical control paradigm by comparing the existing RRS, DRA, and JAF schemes.
As a future work, we will study the concept of dual connectivity and joint association in the eICIC-based HetNet platform. Furthermore, it is interesting to formulate the resource allocation process as a multiobjective optimization problem. Moreover, we will take the ABS ratio issue to design a network-wide max-min fairness process and develop a new low complexity greedy algorithm based on the marginal utility. In addition, considering the privacy protected downlink and uplink decoupling is also an interesting research topic.

Data Availability
Te data used to support the study are available from the corresponding author (swkim01@sogang.ac.kr).

Conflicts of Interest
Te author, Sungwook Kim, declares that there are no conficts of interest regarding the publication of this paper. no. IITP-2023-2018-0-01799) supervised by the IITP