挖掘异构无线网络中的用户需求多样性提升性能

98 浏览量更新于2024-08-29 收藏 1.43MB PDF 举报

“利用异构无线网络中的用户需求多样性——IEEE无线通信 Transactions on Wireless Communications, Vol.14, No.8, August 2015” 本文主要探讨了在异构无线网络中如何利用用户需求多样性来提升网络资源管理的效果。在当前的无线通信环境中，无线电资源管理（Radio Resource Management, RRM）对于提高资源利用率至关重要。传统的做法是通过挖掘网络的多样性来增加用户的吞吐量，但这种方法往往忽视了用户需求对RRM的影响。作者提出了一种新的观点，即通信的最终目标是满足用户个性化的服务需求，因此存在一种潜在的性能提升空间，即用户需求多样性带来的增益。这种增益源于用户需求与无线资源的精细匹配，而现有的以吞吐量为中心的优化策略由于用户对最大化吞吐量的盲目追求，无法直接实现这种匹配。为了获取这种需求驱动的优化效果，文章提出了用户需求中心化的优化方法。在这个框架下，用户不再单纯追求最大吞吐量，而是寻求最大化服务质量（Quality of Service, QoS），这可能包括数据传输速率、延迟、连接稳定性等多个维度。通过考虑用户的不同需求，网络可以更智能地分配资源，使得整体性能得到提升。为了实现这一目标，文章可能涉及以下关键技术和概念： 1. **异构无线网络**：由不同类型的无线接入技术（如宏蜂窝、微蜂窝、微微蜂窝、Wi-Fi等）构成的复杂网络环境，这些技术协同工作以提供广泛的覆盖和容量。 2. **用户需求分析**：理解用户对网络服务的具体需求，如视频流、语音通话、实时游戏等，以及这些需求在时间、空间和带宽上的变化。 3. **资源分配算法**：设计有效的算法，根据用户需求动态调整频谱、功率和其他网络资源，以优化QoS。 4. **匹配理论**：借鉴数学中的匹配理论，建立用户需求与网络资源之间的映射关系，以达到最优的资源分配。 5. **需求感知优化**：将用户的需求状态纳入优化模型，以期在满足用户满意度的同时提高系统效率。 6. **多目标优化**：因为用户需求多样，优化不仅要考虑吞吐量，还要平衡延迟、能量效率、公平性等多个因素。 7. **机器学习和人工智能**：可能利用AI和机器学习技术预测用户行为和需求，以实现更加智能化的资源分配策略。 8. **性能评估**：通过模拟或实验证明用户需求中心化优化方法相对于传统方法的性能优势，包括吞吐量、延迟、公平性和能源效率等方面的提升。通过上述技术的应用和创新，该研究论文旨在为异构无线网络提供更高效、更个性化的服务，从而提升整体网络性能。这样的研究对于未来5G及6G网络的发展具有重要的理论指导意义和实践价值。

4144 IEEE TRANSACTIONS ON WIRELESS COMMUNICATIONS, VOL. 14, NO. 8, AUGUST 2015

A distributed multi-service resource allocation algorithm for

multi-homing terminals is proposed in [8]. To improve the low

system efﬁciency due to users’ selﬁsh behavior, Haddad, et al.

in [9] propose a concise network load broadcasting scheme

to coordinate and assist users’ network selections. Practical

learning algorithms for network selection in dynamic and un-

certain environments are discussed in [10]. Authors in [11]

propose a probability model to analyse users’ network selection

behaviors. Game theory based network selection is discussed

in [12].

In centralized schemes, a centralized controller is responsi-

ble for inter-network resource allocation, network selection of

users and bandwidth allocation among users, etc. To exploit

the network diversity gain, the system sum-rate of orthogonal

frequency division multiple access (OFDMA) based heteroge-

neous networks under the proportional user rate constraint is

studied in [13]. On the network side, Gomes, et al. in [14]

propose an architecture and optimization algorithms to improve

network capacity and minimize the number of used base sta-

tions by splitting cells in congested areas and merging cells

in low load areas. Aiming at harvesting the integration gain of

WiFi/WiMAX networks, effective user network association al-

gorithms are proposed in [15] to achieve the max-min through-

put fairness and the proportional fairness. A hierarchical RRM

game is studied in [16] considering the admission control,

capacity reservation and bandwidth allocation of networks.

Our work differs from the above work in that the user

demand-centric perspective in RRM is proposed, where users

seek to maximize QoE in terms of classical mean opinion score

(MOS). QoE based network selection in a single user case was

ﬁrst studied in [17], while our work focus on the QoE game

in a multiple user case. Although satisfaction equilibriums in

[18]–[20] have similar ﬂavor in bounded demand, our work

is inspired by the subjective QoE levels, resulting in different

equilibrium deﬁnitions and properties and learning algorithms.

We mainly focus on heterogeneous demand scenarios and

characterize the resulting user demand diversity gain. The

simulation results in Section VI also show the superiority of

QoE equilibrium over the satisfaction equilibrium.

III. S

YSTEM MODEL

We consider a heterogeneous wireless network with N

networks coexisting in a given area. For simplicity, we use

“network” to represent a cell in cellular networks or an access

point (AP) in WLAN. The involved networks can be generally

classiﬁed as macrocell or small cell, where macrocells refer to

common 3G, 4G and 5G cellular network cells and small cells

could be picocells and femtocells in cellular networks, APs in

WLANs. The network set is denoted by N = {1,...,N}.A

set of users M = {1,...,M} locate in the networks. Each

user has a set of available networks A

⊆N. In particular,

users located in overlapping areas of networks can access any

one of the networks, but access only one network at any given

time.

Speciﬁcally, a user’s throughput is determined by the phys-

ical layer data rate, the load on the associated network and

the network side resource allocation policy. For network side

resource allocation policy, we assume that the proportional-

fairness and soft-QoS based service differentiation are adopted

in networks. These schemes are widely used in cellular net-

works including LTE-A. In this context, the average throughput

of user m ∈Mgiven the accessed network n is derived by

m,n

. (1)

where R

m,n

is the physical layer data rate between user m

and network n, w

is user m’s weight, W



i∈M

is the total user weight of network n indicating the load of

the network, M

is the set of users associated with network

n. The above throughput model incorporates several practical

considerations. First, the data rate R

m,n

reﬂects the physical

layer characteristics such as radio channel condition, modu-

lation and coding scheme. Second, the discount factor

on R

m,n

shows the nature of multiple user resource sharing.

Third, the user weight here differentiates users in the same

network with diverse trafﬁc or applications, which can be seen

as an indication of application layer information. The model is

similar to the one in [3] and much more general than [2], [5],

since we generalize the user weight from 1 to w

IV. QOEGAME IN HETEROGENEOUS WIRELESS

NETWORK RESOURCE MANAGEMENT

To exploit the user demand diversity in the RRM of heteroge-

neous wireless networks, we convert conventional throughput-

centric optimization to user demand-centric optimization. In

this section, we ﬁrst present the idea of user demand-centric

optimization. Then, we propose a new game formulation, QoE

game, to model users’ behavior in the system, with an emphasis

on features different from the throughput-centric optimization.

Finally, we discuss the properties of the QoE game and present

the deﬁnition of user demand diversity gain.

A. User Demand-Centric Optimization

Most commonly, the throughput-centric optimization is used

in distributed RRM of heterogeneous wireless network, or net-

work selection. Speciﬁcally, each user attempts to maximize the

achievable throughput θ [5] or a strict increasing utility function

f(θ) of θ (e.g., a logarithmic function in [3]). The underlying

assumption is that a larger throughput is always better.In

response to the weakness of throughput-centric optimization

in exploiting user demand diversity, we propose user demand-

centric optimization, where users directly seek for maximizing

QoE. Our intuition is that from user demand perspective, users

can tolerate throughput variation to some extent. Inspired by

the widely used QoE metric, mean opinion score (MOS), we

use MOS to represent users’ goal. In MOS, users’ QoE is

classiﬁed into ﬁve levels, “Excellent”, “Good”, “Fair”, “Poor”

and “Bad”, as shown in Table I. Denote the selected network of

user m by a

∈A

, the network selection strategy proﬁle of

all users by a =(a

,...,a

) with the joint strategy proﬁle set

A = ×

m∈M

. We deﬁne a QoE function q(·) to represent the

剩余13页未读，继续阅读

weixin_38621365

粉丝: 7
资源: 906

挖掘异构无线网络中的用户需求多样性提升性能

异构无线网络资源管理技术研究.pdf

利用异构无线网络中的多无线电协作以绝对安全地防止窃听

优化异构无线网络的协同多样性：基于效用的容量框架

异构无线网络中的模糊多属性决策访问方案

异构无线网络中基于人工神经网络的自适应垂直切换算法.pdf

异构无线网络负载均衡的网络架构

异构无线传感器网络中具有成本效益的屏障覆盖形成

网络游戏-基于合作的异构无线网络分散资源分配方法.zip

智能配电网络中基于Q学习的异构无线网络选择算法

异构无线网络中的效用驱动呼叫接纳与流量均衡策略优化

最新资源