IEEE Xplore:Energy Saving in Heterogeneous Wireless Rechargeable Sensor Networks(异质无线可充电传感器网络的节能问题)

一.Abstract:

移动充电器 (MC) ;无线可充电传感器网络 (WRSN)

Goal:min ---- energy cost which is due to the wireless charging and the MC’s movement

Subject:each sensor obtains its required amount of energy. 

Method: computational geometry-based algorithm

             shortest Hamiltonian cycle

二.System Model:

   Fig. 1. Network Model

The heterogeneous network model(异质性网络模型):

异质性网络模型是一种网络模型,在这个模型中,网络中的节点和边有不同的类型,节点之间和边之间的连接方式也会因为节点和边的类型而不同。这种模型常常用于描述现实社会、技术网络和生物网络等具有异构特性的复杂系统。

例如,在社交网络中,人们之间可以建立各种关系,比如朋友、家人、同事、陌生人等,也可以通过不同的方式进行联系,比如文字聊天、语音通话、视频聊天等。在这种网络中,人们就可以被划分为不同的节点类型,比如利用不同的联系方式进行沟通的人可以划分为不同的类型,而不同类型的节点之间的连边方式也可能有所不同。

异质性网络模型与同质性网络模型不同,同质性网络模型中的所有节点和边都是同一种类型。异质性网络模型能够更准确地反映真实世界中存在的复杂关系,并且可以更好地预测网络的演化和行为。

在实际应用中,异质性网络模型被广泛应用于各种领域,包括社交网络、生物网络、交通网络、金融网络等,以帮助人们更好地理解网络系统的结构和行为规律,以及预测未来的发展趋势

The two coupling NP-hard problems(两个耦合的NP-hard问题):

NP-hard是指非确定性多项式时间(NP)问题的一类,这些问题具有以下特征:

  1. 解决该问题可能需要枚举所有可能的情况,因此算法的复杂度至少是指数级别的。
  2. 该问题可以被转化为其他 NP 问题,即如果我们能够在多项式时间内解决该问题,我们可以在多项式时间内解决其他 NP 问题。

由于这些特征,NP-hard 问题通常非常难以处理,很难在可接受的时间内找到最优解。而将多个 NP-hard 问题组合成一个更大的问题时,问题的复杂度通常会变得更高,这就是所谓的“耦合”。

例如,在物理学中,蛋白质结构预测问题被认为是一个 NP-hard 问题,这意味着找到最优的蛋白质结构需要枚举所有可能的构象状态,因此计算量非常大。而当需要同时预测多个蛋白质结构时,这些问题之间可能会发生耦合,因为蛋白质结构之间可能会相互影响,而找到全局最优解可能需要考虑所有可能的结构,并且需要进行大量的计算。

在复杂系统中,耦合效应是普遍存在的。当多个问题之间发生耦合时,我们需要考虑问题之间的相互关系,并且需要使用更为复杂的算法来解决这些问题。模型简单、难处理的 NP-hard 问题经常涉及到耦合,并且需要特殊方法才能解决,比如贪心算法、遗传算法、模拟退火等优化算法

 divided into two sub-problems,

sub.1:deploying multiple SPs 

sub.2:constructing  all the deployed SPs charging path ( TSP)

Solutions:

 jointly optimize the charging efficiency and the moving cost while carefully bal-
ancing the tradeoff between them. (共同优化充电效率和移动成本,同时仔细平衡它们之间的权衡)

First, identify the optimal location of a single SP( 确定单个SP的最佳位置)

Second, design an edge-based selection algorithm (ESA) to deploy multiple SPs(基于边缘的选择算法(ESA),利用计算几何方法,在整个网络区域内部署多个SP。)

ESA的基本思想是反复选择能够覆盖离网络中心最远的传感器的SP,同时保证这些被覆盖的传感器有最大的充电效率。

Final: calculate all the deployed SPs the corresponding shortest Hamiltonian cycle, i.e., the MC’s charging path(给定所有部署的SP,我们计算出相应的最短哈密顿周期,即MC的充电路径)

Definition 1.charging demand cn of sensor sn

 cn = B − bn and n = 1, …,N.

B:All sensors  same battery capacity(所有传感器电池容量)

bn (n = 1, …,N.):different residual battery levels(不同的电池剩余水平)

三.NETWORK MODEL

A. Mobility Model and Charging Path:

 particular deployment of SPs greatly affects both of the path length and the charging efficiency, 
optimizing the deployment of SPs within the network is the main issue 

B. Wireless Charging Model

WISP-reader charging model

Definition 2. power receiving threshold (PRT)功率接收阈值

pc:源功率(即本工作中MC的充电功率)

pr:传感器的接收功率。

β:调整短距离功率传输的弗里斯方程

a:代表其他恒定的系统参数,参考[6]以了解更多细节

dth:表示最大充电范围,它由充电功率pc和PRT pthr决。

Definition 3.Effective charging area (ECA)

 sn located within the ECA zm can achieve the receiving power pmr (n)(任何位于ECA zm内的传感器sn都可以实现接受功率)

dmn代表sn和om之间的距离,我们有dmn <= dth来成功传输电力。

tm表示充电延迟(即MC在om的停留时间),直到zm范围内的所有传感器的充电需求都满足时为止。

pctm1:相应的充电成本

tm受到各种因素的共同影响,包括充电功率pc,传感器在zm内的分布和充电需求在zm内的分布。在一次充电过程中,总的充电成本可以通过将每个SP的充电成本相加来计算

C. Problem Formulation

Minimized energy cOst (DEMO)

ljk:表示oj和ok之间的欧氏距离,

wjk:是一个决策变量,表示是否存在从oj到ok的路径。

约束条件(4)意味着每个传感器的充电需求必须在一次充电行程中得到满足。

约束条件(5)意味着对每个SP来说,只存在一条来自另一个SP的入站路径和一条到另一个SP的出站路径。

Step 1-Deployment of SPs within the whole network

 a NP-hard problem proved in [8] 

Step 2-Charging path construction:

minimize the path length: 

(TSP)

minimize the MC’s total energy cost:

Fig. 2. The tradeoff between charging efficiency and moving cost.

Assume  the value of the MC’s moving cost per unit distance is  larger than  the charging power.

III. SINGLE SP DEPLOYMENT WITHIN AN ECA

Homogeneous case:

smallest enclosing circle (SEC) :最小包围圈

minimizes the maximum distance between any sensor and the SP 

Optimal SP:at the center of the SEC

同构或均质(Homogeneous)通常被用来描述一种网络拓扑结构的性质。在同构图中,所有节点的度数和相互之间的连接方式都是相同的。

在同构网络中,所有节点之间的边权重或属性都相同。这意味着同构网络中的任何两个节点之间的关系都是相同的,没有节点对之间的差异。同构网络通常包含具有相同特征和相似行为的节点,比如社交网络中的用户、生物网络中的基因等。

同构网络相对于异构网络而言,异构网络中具有不同属性和特征的节点和边被赋予不同的权重和属性,节点之间和边之间的连接方式也会因为节点和边的类型而不同。

Heterogeneous case:

Optimal SP:located within the minimum convex hull created by employing the Graham - Scan algorithm. (Graham-Scan算法创建的最小凸壳内)

 Identify the optimal SP within the minimum convex hull :

sn的充电延迟:

 Cn:每个传感器sn的充电需求为cn,其中n = 1, . .N

 dn: the distance between sn and the SP

 all the N sensors  charging delay T:

 Target :identify the optimal SP to minimize T(minimizes the charging cost)

 a min-max optimization problem as
 

Subject:

(xn, yn)和(xo, yo)分别表示sn和SP的二维坐标

 Traditional optimization methods cannot solve the above problem  

since there is no explicit expression for the objective function in Equation (8)
and multiple optimization targets coexist, which need to be optimized in parallel.
 

Soliutions:

PCES (parallel circles expanding searching )

 the intuition of PCES 

 solving the min-max problem is to suppress the most outstanding element among all the
optimization targets. Based on this idea, the problem of minimizing T can be converted to the problem of minimizing the ratio rt

Corresponding detailed algorithm:

1.reducing the original searching region C to a common covered region Cr 

2.reducing Cr to a single point 

Fig. 3. Geometrical illustration of identifying the optimal SP within an ECA

Algorithm 1: Parallel Circles Expanding Searching (PCES)

 IV . DEPLOYMENT OF MULTIPLE SPS W I T H I N T H E W H O L E NETWORK AND CHARGING PAT H CONSTRUCTION

Edge-basedselection algorithm (ESA) :

First: employ the Graham-Scan algorithm to derive the minimum convex hull C and its
boundary. 

Second: draw a circle Cx which centers at s1 with the radius of R, where R = dth.  draw another circle Cy which centers at s1 with the radius of 2R.

Third: define the node set S3 containing all sensors in the boundary of C while being
covered by Cx, i.e., S3 = {s1} in Fig. 4

Fourth: define the node set S4 containing all sensors covered by Cy, i.e., S4 =
{s1, s2, s3, s4} in Fig. 4. For each sensor sj in S4 \ S3, we calculate the optimal SP based on each set {S3, sj} via PCES,i.e., the three candidate SPs shown in Fig. 4. 

Fifth: identify one of these candidate SPs which covers the largest number of sensors, i.e., 

Finally,:we remove s1, s2, s3, s4 from the current network topology and obtain the new network topology shown in  Fig. 4

Repeat the above process until we identify all the SPs required to cover the whole network.

Fig. 5. Multiple ESA iterations for generating multiple S

V. PERFORMANCE EV ALUA TION

A. Evaluation Settings

B. Existing Algorithms for Performance Comparison

C. Evaluation Results and Implications

VI.Related works

Fig. 6.

The impact of number of sensors on the total energy cost, the moving energy cost and the charging energy cost.

Fig. 7.

The impact of charging power on the total energy cost, the moving energy cost and the charging energy cost.

VII. CONCLUSION
In this work, we minimized the MC’s total energy cost for both charging and movement, through the charging path op-timization of the MC in a heterogeneous network setting. Wedesigned an efficient edge-based selection algorithm (ESA) tocarefully balance the tradeoff between the charging efficiencyand the moving cost, which has the approximation ratio ofO (ln N) and the time complexity of O (N ln N). We validatethe effectiveness of the proposed path design via extensive
evaluations, which outperforms the existing main algorithmsin terms of the MC’s energy cost minimization.

猜你喜欢

转载自blog.csdn.net/m0_62690134/article/details/130115132
今日推荐