SDN驱动的物联网网络：一种基于SMDP的无线电资源分配策略
56 浏览量
更新于2024-08-26
收藏 1.55MB PDF 举报
"这篇研究论文探讨了在软件定义的物联网（IoT）网络中，基于有限状态马尔可夫决策过程（SMDP）的无线电资源分配方案。随着物联网技术的迅速发展，机器对机器（M2M）通信技术涌现，为大量物联网设备提供了普遍的无线连接。然而，这也对大规模IoT网络的管理和控制提出了重大挑战。软件定义网络（SDN）因其动态可重构的网络元素特性，被视为优化网络管理的有效技术。SDN与IoT的结合为增强物联网网络的管理和控制能力提供了可行的解决方案。利用SDN技术，可以进一步提高IoT网络的资源利用率。论文首先提出了一种适用于IoT的SDN架构，然后重点介绍了基于SMDP的资源分配策略，旨在优化网络性能并提高效率。"
本文深入研究了在物联网环境中如何利用SDN进行有效的无线电资源管理。物联网的快速增长导致了对网络控制和管理的需求激增，而SDN的引入解决了这一问题，通过动态配置网络元素，使得网络管理变得更加灵活和高效。SMDP是一种强化学习方法，用于在不确定环境中做出最优决策，它在此处被用于无线电资源分配，以最大限度地提高网络性能。
首先，论文提出了一种新的SDN架构，该架构专门为物联网设计，强调了SDN控制器在整体资源管理中的核心作用。控制器负责收集网络状态信息，分析并做出决策，以优化资源分配。此外，这种架构考虑了物联网设备的多样性和动态性，使得资源分配更具适应性和智能性。
接着，SMDP模型在无线电资源分配中的应用是本文的重点。通过建模网络状态、动作、奖励函数以及转移概率，SMDP能够动态地调整资源分配策略，以适应不断变化的网络条件。这种方法考虑了物联网设备的能量效率、服务质量（QoS）要求以及网络拥堵等因素，以实现长期性能的最大化。
在实际应用中，SMDP-based方案能够有效地平衡物联网网络的负载，减少冲突，提高频谱效率，并确保关键服务的可靠性。同时，由于SMDP的优化特性，它还能帮助降低设备的能耗，延长电池寿命，这对于许多物联网设备来说至关重要。
这篇研究论文为软件定义的物联网网络提供了一个创新的无线电资源分配策略，利用SMDP优化了资源管理，提高了网络效率和整体性能。这一方法对于未来物联网的发展具有重要的理论和实践意义，有望成为解决大规模物联网网络管理和控制挑战的关键技术之一。
7306 IEEE SENSORS JOURNAL, VOL. 16, NO. 20, OCTOBER 15, 2016
owns K radio resources (termed access resources (ARs)),
which are assigned to the MTCD-to-MTCG and MTCD-to-
eNB links. Each MTCD-to-MTCG link and MTCD-to-eNB
link can consume u ARs of the connected MTCG and eNB,
where u ∈{1,...,U} and U ≤ K . Moreover, data packets
transmitted from an MTCD follow a Poisson distribution with
the mean rate of λ
m
, and the MTC packet size follows an
exponential distribution with a mean size of ρ
m
. Therefore,
the total arrival rate of the MTC data packets to the network
is equal to Mλ
m
.
When an MTCD is connected to an MTCG, the MTCG
performs as a one-hop relay to forward data packets to the eNB
via the MTCG-to-eNB link. It is assumed that the MTCG-to-
eNB link employs the 3GPP LTE/LTE-A cellular communica-
tions technique, which supports high bandwidth transmission
and operates in a different frequency band from NB-IoT.
We suppose that the eNB owns L radio resources (termed
backhaul resources (BRs)) for the MTCG-to-eNB links, and
the eNB has many more BRs than ARs, i.e., K < L.
Each MTCG-to-eNB link occupies w BRs of the eNB, where
w ∈{1,...,W}, W ≤ L. Meanwhile, the eNB is also
responsible for supporting H2H communications. H HTC
devices (HTCDs) coexist with MTCDs in the serving areas of
the network, and interact with the eNB through the HTCD-to-
eNB links. The HTCD-to-eNB links and MTCG-to-eNB links
share the common BRs of the eNB, which allocates v BRs
for each HTCD-to-eNB link, where v ∈{1,...,V }, V ≤ L.
Assume that each HTCD sending data packets also follows
an independent Poisson process with a mean rate of λ
h
,and
the size of the HTC packets also follows an exponential
distribution with a mean size of ρ
h
. Each MTCG-to-eNB
link and HTCD-to-eNB link also can support l
h
transmission
modes with different transmission rates C
h
(·).Let{{θ
h,i
}
l
h
+1
i=1
}
be the set of SNR boundary points, where θ
h,l
h
+1
=∞.
When SNR ξ ∈[θ
h,i
,θ
h,i+1
), the transmission rate C
h
(ξ) of
transmission mode i is equal to μ
h,i
,wherei ∈{1,...,l
h
}.
To maximize network revenue, the network must decide the
access strategy for each MTCD (i.e., through the eNB or an
MTCG), and optimize radio resource allocation for each
wireless link. Thus, we formulate the radio resource allocation
model in the software-defined IoT network as an SMDP
problem. In SDN, all the decision-making procedures are
carried out in the SDN controller. When a new MTC or HTC
data packet arrives, the SDN controller first evaluates the
expected system gain and system expense based on the current
status information of the eNB and MTCGs. Then, the SDN
controller decides whether to accept or reject the data packets,
which MTCGs or eNB should be selected for MTC traffic
transmission, and how to allocate radio resources to each
wireless link. In the SMDP framework, the decisions adopted
by the SDN controller are called actions, and the moments
when decisions are made are termed decision epochs.The
action chosen is based on the current system state of the net-
work, which includes the current traffic load on each MTCG
and the eNB. To make the optimal decision for the reward
model, the SDN controller needs to obtain the system reward
for each action before making any decision. The long-term
expected average reward per unit time of the network is
considered as the optimal criterion for the SMDP.
B. Problem Formulation
This subsection aims to formulate the considered optimized
problem as an SMDP. The corresponding system states, actions
based on each state, and the reward model are described as
follows.
1) System States: The system state S of the software-
defined IoT network can be represented by the number of
current wireless links in the network with different numbers
of radio resources occupied, as well as an event occurred in
the system, which could be either the arrival or departure of
a data packet. The system state space S can be denoted as
follows, i.e.,
S =
s | s = (g
1
, g
2
,...,g
N
, s
1
, s
2
,...,s
N
, s
m
, s
h
, e)
,
(1)
where g
i
, s
i
, s
m
, s
h
, i ∈{1,...,N} are defined as
g
i
= (g
i,1
, g
i,2
,...,g
i,U
),
s
i
= (s
i,1
, s
i,2
,...,s
i,W
),
s
m
= (s
m,1
, s
m,2
,...,s
m,U
),
s
h
= (s
h,1
, s
h,2
,...,s
h,V
). (2)
The above symbols are explained in detail below
• g
i
: a vector of g
i,u
,whereu ∈{1,...,U}. g
i,u
represents
the number of wireless links between the MTCDs and
the ith MTCG that occupy u ARs. The total num-
ber of allocated ARs of the ith MTCG should satisfy
U
u=1
(ug
i,u
) ≤ K ;
• s
m
: a vector of s
m,u
,whereu ∈{1,...,U}. Similarly,
s
m,u
represents the number of MTCD-to-eNB links that
occupy u ARs. The total number of allocated ARs of the
eNB should also be subjected to
U
u=1
(us
m,u
) ≤ K ;
• s
i
: a vector of s
i,w
,wherew ∈{1,...,W}. s
i,w
is the
number of the wireless links between the ith MTCG and
the eNB that occupy w BRs;
• s
h
: a vector of s
h,v
,wherev ∈{1,...,V }. s
h,v
indi-
cates the number of HTCD-to-eNB links that occupy
v BRs. Thus, the total number of allocated BRs
of the eNB should satisfy
N
i=1
W
w=1
(ws
i,w
)
+
V
v=1
(vs
h,v
) ≤ L;
• e: An event in the event set E , i.e., e ∈ E .
The event set E is denoted by
E ={A
m
, A
h
}∪D
1
∪ D
2
∪ ...∪ D
N
∪ D
m
∪ D
h
, (3)
where D
i
, D
m
, D
h
, i ∈{1,...,N} are defined as
D
i
={D
i,u,w
| u ∈{1,...,U},w∈{1,...,W}},
D
m
={D
m,u
| u ∈{1,...,U}},
D
h
={D
h,v
| v ∈{1,...,V }}. (4)
Each event is detailed as follows
• A
m
: the network receives a data packet from an
MTCD. A
m
can be further denoted as A
m
=
(μ
1
,μ
2
,...,μ
N
,μ
N+1
),where{μ
i
}
N
i=1
represent the
transmission rate of the wireless link between the MTCD
剩余10页未读，继续阅读
2021-04-30 上传
2020-05-05 上传
2021-09-20 上传
2021-01-14 上传
2021-05-25 上传
2021-04-26 上传
2021-06-05 上传
2021-07-16 上传
weixin_38664556
- 粉丝: 5
- 资源: 958
我的内容管理
展开
最新资源
- Aspose资源包：转PDF无水印学习工具
- Go语言控制台输入输出操作教程
- 红外遥控报警器原理及应用详解下载
- 控制卷筒纸侧面位置的先进装置技术解析
- 易语言加解密例程源码详解与实践
- SpringMVC客户管理系统：Hibernate与Bootstrap集成实践
- 深入理解JavaScript Set与WeakSet的使用
- 深入解析接收存储及发送装置的广播技术方法
- zyString模块1.0源码公开-易语言编程利器
- Android记分板UI设计：SimpleScoreboard的简洁与高效
- 量子网格列设置存储组件：开源解决方案
- 全面技术源码合集：CcVita Php Check v1.1
- 中军创易语言抢购软件：付款功能解析
- Python手动实现图像滤波教程
- MATLAB源代码实现基于DFT的量子传输分析
- 开源程序Hukoch.exe：简化食谱管理与导入功能
