Elsevier

Computer Networks

Volume 183, 24 December 2020, 107596
Computer Networks

Outage-aware Matching Game Approach for Cell Selection in LTE/WLAN Multi-RAT HetNets

https://doi.org/10.1016/j.comnet.2020.107596Get rights and content

Abstract

Heterogeneous networks (HetNets) which are consisting of multiple radio access technologies (Multi-RATs), offer a promising architecture for enhancing the mobile network capacity and improving the users’ quality of experience (QoE). However, the main challenge in such network architecture is the development of an efficient cell selection solution. In this paper, a distributed algorithm for the cell selection process is proposed to consider the impact of fast fading channel on the system performance. The cell selection process is formulated as an optimization problem, which is combinatorial, for minimizing the average outage probability and thus improving the system performance. Since there are many services that require a low outage probability, i.e. more sensitive to the system degradation, we are focusing in our paper on Long term evolution (LTE) wireless local area network (WLAN) aggregation (LWA) technology and how it may be used to enhance the outage performance of the system. Therefore, based on well-formulated utility functions for users and different stations in the multi-RAT HetNet, a matching game-based approach is proposed to provide a solution for the cell selection problem. Simulation results show the superiority of the proposed matching scheme compared to other cell selection schemes especially in terms of the achieved average outage probability with a very close performance to the optimal heuristic search solution. Also, the performance improvement as a result of deploying LWA stations in the network is shown. The reason for this improvement is that the aggregation enhances the achieved throughput significantly, which in turn improves the outage performance.

Introduction

The large spreading of mobile devices and applications has led to an explosive growth of data traffic, which in turn causes a continuous increase in network capacity demands [1]. Recently, Cisco researchers warned that the mobile data traffic will increase seven-fold between 2017 and 2022, with a monthly growth that will reach 77.5 exabytes by 2022, which is faster than static Internet Protocol (IP) traffic [2]. Therefore, during the peak times, users may face long latency for delivering their data, low throughput, and connection outage due to congestion on the cellular network. However, users have certain target rates for different services which should be provided with a low outage probability, regardless of network conditions, in order to satisfy their QoE. Therefore, mobile operators are interested in mitigating the load on cellular networks and they are seeking for alternative solutions. An efficient solution is to use the existing radio access technologies (RATs) like WiFi technology, to direct the data traffic in Heterogeneous networks (HetNets). The mobile operators are interested in data offloading to enhance the capacity of their networks while maintaining a good quality of service for customers. In general, offloading data to different small base stations (SBSs) is an efficient way to improve performance, as the user equipment (UE) is capable of accessing multiple RATs (multi-RATs). SBSs are stations that operate with low-power levels and few meters range, which is considered a short range compared to macrocell base stations (MBSs) range. These Small stations may be femtocells, picocells, microcells, or WiFi access points that are deployed by operators to enhance the network capacity and coverage. Therefore, HetNets consisting of MBSs overlaid with different SBSs, which can operate in licensed or unlicensed spectrum [3], provide a cost-efficient design, and support expansion of the present cellular networks to tackle the increase in network capacity demand. Also, deploying different SBSs can boost the user throughput and quality of service, thus enhancing the outage performance within the network. Therefore, data offloading in multi-RAT networks becomes a very interesting for the industry field and academia [4]. Currently, many interests are focused on WiFi technology to offload mobile data traffic as it uses unlicensed spectrum and can address the traffic demand [5,6]. Therefore, mobile operators can reduce the load on their networks using a cost effective solution, which is known as WiFi offloading [7,8].

In this context, an efficient scheme was introduced by 3GPP RAN in Release 13 named LTE WLAN Aggregation (LWA) [9], in which the user's data traffic is transmitted simultaneously over both LTE and WiFi networks. For LWA technology, two architectures are defined: collocated and non-collocated deployments. In the case of collocated scenario, the LTE eNB and WLAN access point (AP) are assumed to be integrated within one entity, while in non-collocated scenario both networks are separated and connected using a backhaul link with a new interface called xw. LWA provides an efficient scheme for LTE WLAN interworking that can enhance the achieved throughput of users by aggregating both licensed and unlicensed bands. However, deciding which RAT is preferred to select in Multi-RAT HetNets and the cell that supports this RAT can perform a problem known as, the user association or cell selection problem. Therefore, the user should be associated only with a cell that will satisfy his service requirements, while optimizing the network performance, which is a big challenge in Multi-RAT HetNets.

The network and cell selection in a HetNet scenario have been studied in previous works with different approaches. In [10], a centralized user association scheme in HetNet was proposed, based on game theory with a fairness and load balancing scheme. A load fairness index was defined to measure the degree of load imbalance among cells, and the proposed cell association scheme is only executed when this index is less than a certain threshold. In [11], the authors formulated an algorithm for the cell selection process based on matching game in order to maximize the network utility, which is defined as the sum of the logarithm of long-term rate of users. Their target was to enhance the network throughput, while considering fairness among the users. However, the instantaneous conditions of the link connecting a user with each cell, and the required service rate for the user were not considered. The authors in [12] focused on user QoE and its satisfaction, by proposing a scheme for associating the user in wireless small cell networks. They formulated a matching game for optimal user assignment to address the system QoE maximization problem, and used a qualitative mean opinion score with multiple levels to indicate the satisfaction related to different applications. However, this work did not consider HetNets which have different RATs, each with its own characteristics, and how to distribute the users among the network cells taking into account the demand for each user. In [13], the authors proposed a user association scheme in the downlink of small cell networks based on many-to-one matching game, considering the rate for cell-edge users. The users send their preference vector to each BS, for assigning the applicants priorities based on this vector. So, the BSs prioritize the users based on the information concealed in the preferences of each user, neglecting the demand required by the user's service to satisfy its QoS. They also did not address the different access networks that have different medium access control (MAC) protocols.

In [14], a Q-learning distributed algorithm was proposed in which each user learns from his local environment, and selects the best station (MBS or WiFi AP). They introduced a new parameter for offloading decision that considers the load of each WiFi AP, and the achieved signal-to-interference-plus-noise ratio (SINR). In [15], the authors presented a selection algorithm for WiFi and LTE HetNets that depend on continuously monitoring the state of a HetNet using received signal strength indicator (RSSI), throughput, and delay metrics. The authors in [16] formulated the network selection problem considering the user and base station information. The paper combined information such as Channel Quality Indicator, available bandwidth, and traffic load to make the decision for the network selection problem. However, they considered the service demand only for preventing BSs overloading during WiFi offloading process. In [17], a network selection scheme which is aware of UE preference was presented for mobile data offloading. The proposed scheme used the network information such as SINR and cell load in order to estimate the most preferable BS/AP for the user. The authors in paper [18], investigated the access network discovery and selection function (ANDSF) with multiple real time indicators of the network performance that can help in process of access networks ranking. The networks are evaluated against different criteria including congestion level, available bandwidth, delay and jitter. In [19], they considered the user's velocity to predict the staying time in WLAN coverage area to help in the initial network selection. Then, using the mobility prediction, network load, and service type, the algorithm can select the network which is most suitable for the user. In [20], the authors considered a multi-RAT system comprised of different access technologies which can be co-located or distributed in a C-RAN architecture or forming a HetNet structure. Therefore, a practical multi-connectivity scheduler for OFDMA-based multi-RAT systems was proposed to achieve a balance between system cost and utility satisfaction, while operating jointly across different RATs, accommodating real-system requirements, and guaranteeing system stability. The authors in [21] analyzed the performance of backhaul-aware cell selection by deriving the outage probability and using a larger association bias for small base stations (SBSs) with high-backhaul-capacity to offload users from low-backhaul-capacity SBSs. The concept of cell group selection is used instead of single cell selection, then the SBS within the cell group which provides the highest data rate to the user will be selected for transmission. The authors in [22] investigated the problem of RAN selection in a Heterogeneous Networks scenario with the objective of distributing the traffic among several RANs in a fair way. The RAN selection problem for HetNet is formulated as a sequential decision-making process. In addition, a mathematical analysis was carried out in order to estimate several performance metrics such as: the average blocking probability in the system, the average number of users served and the utilization of each RAN. In [23], the authors considered the RAT selection in a cellular/WLAN heterogeneous network which is based on Markov Decision Process and with the objective of maximizing the revenue. They highlighted the WLAN role in traffic offloading and how the QoS could be improved. Authors in paper [24] proposed a context-aware RAT selection mechanism which operates mainly on user equipment side to minimize signaling overhead and computation load on BS. The proposed mechanism collects and processes context information monitored by the UE, to get advantage of the latest 3GPP trends in the LTE-EPC architecture (e.g. ANDSF functionality).

There are many works that consider the architecture of LWA scheme and study its performance. The work in [25] presented the LWA architecture and proposed an algorithm for controlling the flow over LWA technology, which is based on the feedback from UE. The algorithm aims to get the most benefit from licensed and unlicensed spectrum by optimally splitting the traffic. In [26], the WiFi offloading with LWA is presented for balancing between user payment and QoS requirement. The authors revealed that LWA is chosen only when the aggregation gain is greater than the minimum value, which is determined by the average LTE and WiFi rate. In [27], a scenario where the user's traffic may be transmitted through both a macrocell and a small cell with LTE or WiFi technology was investigated and an optimal aggregation solution was developed. An optimal solution based on a “water-filling” interpretation was proposed, where the portion of user's traffic sent over LTE macrocell base station is proportional to ratio of peak capacity that a user can acquire on that macrocell and its throughput on the small cell. The authors in [28] proposed a new method for cell selection which is based on aggregation in order to maximize throughput of the system with LWA architecture. The authors used the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) for calculating the preference with four criteria: energy consumption, delay, load, and price. In [29], the authors proposed an algorithm for the adaptation of aggregation mode and controlling how the data is divided between LTE and WLAN networks under various load conditions. The work in [30] used TOPSIS technique in order to arrange the available modes, including LWA mode, and their own stations, according to several metrics and based on the user's service requirements, thus improving the network performance while satisfying the user requirements. In [31], the authors considered LWA architecture and proposed a scheduling technique with embedded mode selection criteria, which considers the user's service time deadline, aiming to save the delivering cost of the service as possible.

However, all of the aforementioned works did not consider that the cell selection may be affected by the instantaneous conditions of the connecting links, in addition to the required data rate for the user's service. Therefore, for the services and applications that cannot tolerate the system degradation, we have a very interesting parameter that cannot be neglected during the cell selection process, which is the outage probability metric. Additionally, in the literature surveyed, we observe that they rarely consider the MAC model for WiFi network and how the contention degree of its channel can affect the system performance. Also, to the best of our knowledge, there does not exist a research work for the cell selection process that applies the matching game while adopting LWA architecture within that game in order to improve the system outage probability, considering both cellular and WiFi RATs. In this sense, our proposal complements previous research works by applying the matching game as a strategy for the network selection process based on the outage probability, and aiming to satisfy the user's service requirements while optimizing the overall system performance.

The key contributions of this paper are summarized as follows:

  • In this paper, the problem of cell selection is formulated as a distributed process based on the matching game framework, where each player, i.e. user or BS, has its utility function that is used for ranking the other side players and construct its preference set.

  • We propose two scenarios considering different network architectures. In the first scenario, we apply the matching game in a HetNET model without deploying LWA technology and show the superiority of the proposed matching algorithm compared to other cell selection schemes. Then, another scenario will be adopted where some of nodes within that model, i.e. LTE and WiFi SBSs, are combined to form the LWA architecture so, we will show the performance improvement when applying our proposed matching algorithm with LWA scheme. Finally, having a stable matching that would make users better off and leads to an improved network performance.

  • We investigate a very important metric, which should be considered during the cell selection process and is represented in terms of the outage probability of user link over each network. This metric is based on the instantaneous signal to noise and interference ratio (SINR) over that links and will depend on the fading distribution of the channel and the required data rate for the user, i.e., service demand.

  • Focusing on LWA architecture, we split and distribute the users’ traffic according to the resources available over each network by determining the suitable splitting ratio, thus enhancing the outage performance by supporting the ability to achieve the user target rate.

  • Also, the MAC model for WiFi network and the contention degree of its channel is considered in this paper during the rate analysis and outage probability calculations.

  • The computation complexity analysis for the proposed algorithm is given. In addition, the convergence of the proposed algorithm to a stable matching is proved.

  • The performance of the proposed cell association algorithm is compared with another association scheme and with the optimal heuristic search in terms of average outage probability and the results are analyzed and concluded.

The rest of this paper is structured as follows. The system model for Multi-RAT HetNet is described in Section 2. The problem formulation is defined in Section 3. In Section 4, the cell selection algorithm based on matching game is proposed. The performance evaluation and numerical results are provided in Section 5, and finally Section 6 summarizes and concludes this paper.

Section snippets

System architecture

Without loss of generality, a two-tier heterogeneous network is considered in which the first tier consists of one LTE macro base station (MBS) while the second tier consists of several categories of small base stations (SBSs) overlaid under the area covered by the MBS as shown in Fig. 1. The SBS categories which are considered in this model include, LTE femto base stations (FBSs), stand-alone WiFi access points (WAPs), and LWA stations which will be referred as cooperative base stations

Problem formulation

Based on the above characterization, an optimization framework is developed for the base station selection problem in multi-RAT heterogeneous network. Let us first define a binary association matrix x, where xn, i is an element of matrix x nN,ik.Because all association indicators are binary variables, so thatxn,i={1,whenuseriisassociatedwithstationn0,otherwise

Here, the effect of the wireless channel that makes deterioration on the instantaneous achievable rate is considered. For this, the

Matching game-based cell selection

A matching game is defined as a pairing problem between two disjoint sets of players in which, according to preference relations, the players of each set are interested to be matched to the players of the other set. Individual preferences represent how a player would choose among different alternatives. The formulated network selection problem in (23) can be mapped into a matching game in which the first set of players consists of different nodes (MBS, FBSs, WAPs, and CBSs); meanwhile the

Simulation setup

For evaluating the performance of our proposed technique, we consider a MBS overlaid with two FBSs and two WAPs scenario. We evaluate our system performance in terms of the average outage probability metric. To show the importance of the proposed outage matching-game strategy, we evaluate its performance with other strategies, e.g. WLAN-first strategy and LTE-first strategy. For WLAN-first, the user always gives the preference to associate with a WiFi network, while for LTE-first, the user

Conclusion

In this paper, we have developed a distributed strategy for the cell selection problem based on matching game, in which the outage probability is used as a userlity for constructing the preference list. The preference list of each user is a ranking of the available networks according to the outage performance such that; the first ranked provides the lowest outage probability. We study the effect of the target data rates and number of users on the average outage probability and thus on the

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

References (44)

  • X. Kang et al.

    ‘’Mobile data offloading through a third-party wifi access point: an operator's perspective’

    IEEE Trans. Wireless Commun.

    (2014)
  • LTE-WLAN Radio Level Integration and Interworking Enhancement, document RP-152213, 3GPP RAN Plenary #70, Sitges, Spain,...
  • H. Du et al.

    A load fairness aware cell association for centralized heterogeneous networks

  • Q. Han et al.

    Matching-based cell selection for proportional fair throughput boosting via dual-connectivity

  • H. Shao et al.

    QoE-aware downlink user-cell association in small cell networks: a transfer-matching game theoretic solution with peer effects

    IEEE Access

    (2016)
  • O. Semiari et al.

    Matching theory for priority-based cell association the downlink of wireless small cell networks

  • E. Fakhfakh et al.

    ’ Optimised Q-learning for WiFi offloading in dense cellular networks’

    IET Commun.

    (2017)
  • S. Ahmed et al.

    Analysis of access network selection and switching metrics for LTE and WiFi HetNets

  • W.U. Jin et al.

    ’ Intelligent network selection for data offloading in 5G multi-radio heterogeneous networks’

    China Commun.

    (2015)
  • J. Orimolade et al.

    Intelligent access network selection for data offloading in heterogeneous networks

    AFRICON

    (2015)
  • K.M. Zaini et al.

    A combination of mobility, load and services for network selection in heterogenous wireless network

  • L. Diez et al.

    LaSR: a supple multi-connectivity scheduler for multi-RAT OFDMA systems

    IEEE Trans. Mobile Comput.

    (2020)
  • Cited by (6)

    View full text