CN110012039B - ADMM-based task allocation and power control method in Internet of vehicles - Google Patents

ADMM-based task allocation and power control method in Internet of vehicles Download PDF

Info

Publication number
CN110012039B
CN110012039B CN201810006519.9A CN201810006519A CN110012039B CN 110012039 B CN110012039 B CN 110012039B CN 201810006519 A CN201810006519 A CN 201810006519A CN 110012039 B CN110012039 B CN 110012039B
Authority
CN
China
Prior art keywords
roadside unit
task
user equipment
computing
optimal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
CN201810006519.9A
Other languages
Chinese (zh)
Other versions
CN110012039A (en
Inventor
周振宇
刘朋矩
许晨
冯俊豪
唐良瑞
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
North China Electric Power University
Original Assignee
North China Electric Power University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by North China Electric Power University filed Critical North China Electric Power University
Priority to CN201810006519.9A priority Critical patent/CN110012039B/en
Publication of CN110012039A publication Critical patent/CN110012039A/en
Application granted granted Critical
Publication of CN110012039B publication Critical patent/CN110012039B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F9/00—Arrangements for program control, e.g. control units
    • G06F9/06—Arrangements for program control, e.g. control units using stored programs, i.e. using an internal store of processing equipment to receive or retain programs
    • G06F9/46—Multiprogramming arrangements
    • G06F9/50—Allocation of resources, e.g. of the central processing unit [CPU]
    • G06F9/5005—Allocation of resources, e.g. of the central processing unit [CPU] to service a request
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/50—Network services
    • H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
    • H04L67/62—Establishing a time schedule for servicing the requests
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/50—Network services
    • H04L67/60—Scheduling or organising the servicing of application requests, e.g. requests for application data transmissions using the analysis and optimisation of the required network resources
    • H04L67/63—Routing a service request depending on the request content or context
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W52/00—Power management, e.g. Transmission Power Control [TPC] or power classes
    • H04W52/02—Power saving arrangements
    • H04W52/0209—Power saving arrangements in terminal devices
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

本发明涉及一种车联网场景中的移动边缘计算方案，该方法在满足时延要求的前提下，对车内用户设备的计算任务分配和传输功率控制问题进行了优化。把设备在计算任务分配率加权下的能源损耗作为目标函数，使用排队论方法获得用户设备和边缘计算节点的数据传输模型，通过非线性分式优化和交替方向乘子法的迭代来解决该优化问题。在每一轮循环中，外层循环解决非线性分式规划问题，内层循环更新初始值和变量，直到迭代的结果满足设定的阈值，确定各用户设备的任务量分配比率并得到最小化的能耗。本发明提供的技术方案可以有效降低用户设备的能耗并满足时延的要求，提高整个网络的计算能力。

Figure 201810006519

The invention relates to a mobile edge computing solution in a car networking scenario. The method optimizes the problems of computing task allocation and transmission power control of in-vehicle user equipment on the premise of satisfying time delay requirements. Taking the energy consumption of the device under the weighting of the computing task allocation rate as the objective function, the data transmission model of the user equipment and edge computing nodes is obtained by using the queuing theory method, and the optimization is solved by nonlinear fractional optimization and iteration of the alternate direction multiplier method. question. In each round of loops, the outer loop solves the nonlinear fractional programming problem, and the inner loop updates the initial values and variables until the iterative result meets the set threshold, and determines the task allocation ratio of each user equipment and minimizes it. energy consumption. The technical solution provided by the present invention can effectively reduce the energy consumption of the user equipment, meet the requirement of time delay, and improve the computing capability of the entire network.

Figure 201810006519

Description

一种车联网中基于ADMM的任务分配与功率控制方法A task allocation and power control method based on ADMM in the Internet of Vehicles

技术领域technical field

本发明涉及无线通信领域的移动边缘计算方案，具体涉及一种车联网中基于ADMM的任务分配与功率控制方法。The invention relates to a mobile edge computing scheme in the field of wireless communication, in particular to an ADMM-based task allocation and power control method in the Internet of Vehicles.

背景技术Background technique

作为物联网在交通运输领域的典型应用，车联网实现了在很少或没有人为干预的车辆中进行无所不在的信息共享，这对于实现未来智能交通系统至关重要。一方面，车联网会刺激道路安全、旅行援助和自动驾驶等领域的一系列具有严格时效性要求的应用程序快速发展；另一方面，增强现实、流媒体视频和在线游戏等丰富的多媒体物联网应用迅速发展，导致极大的工作负载数据需要被缓存和处理，这需求大量的计算、通信和存储资源。在传统的云计算模型中，云服务器的位置远离需求边，而且回传路径和主干网络能力有限，这就造成了不可预测的延迟，无法保证物联网提供可靠的服务质量和体验质量。As a typical application of the Internet of Things in transportation, the Internet of Vehicles enables ubiquitous information sharing in vehicles with little or no human intervention, which is crucial for the realization of future intelligent transportation systems. On the one hand, the Internet of Vehicles will stimulate the rapid development of a series of applications with strict timeliness requirements in areas such as road safety, travel assistance and autonomous driving; on the other hand, rich multimedia IoT such as augmented reality, streaming video and online games The rapid development of applications results in enormous workload data that needs to be cached and processed, which requires a lot of computing, communication, and storage resources. In the traditional cloud computing model, the location of cloud servers is far from the demand side, and the backhaul paths and backbone network capabilities are limited, which results in unpredictable delays and cannot guarantee the reliable quality of service and experience provided by the IoT.

而作为物联网中快速的任务处理方法，车辆的边缘计算VEC将计算模式从远端的中心分布构架扩展到了分布式边缘服务器。在车联网中，计算、通信和存储资源被分配到接近用户的地方，并且分散在网络边缘。车联网可以被视为传统云计算的一种有益补充。在网络边缘处理较低计算需求和有严格时效性限制的任务可以消除过多的网络越点，这不仅减少了计算响应时间，也缓解了能力有限的回程链路的信号拥挤问题。进一步来说，车联网通过将耗能过多的工作负载转移到具有更高计算能力和持续能量供应的VEC节点上，大大提高了电池容量有限的智能手机和可穿戴设备等车内用户设备的续航时间。通过适宜的任务分配策略，本地计算的能量损耗以增加数据传输的能量消耗，以及由数据传输、边缘服务器上的工作负载处理及跨区引起的延迟为代价而被减少。As a fast task processing method in the Internet of Things, the edge computing VEC of the vehicle extends the computing mode from the remote central distributed architecture to the distributed edge server. In the Internet of Vehicles, computing, communication and storage resources are allocated close to the user and dispersed at the network edge. The Internet of Vehicles can be seen as a useful complement to traditional cloud computing. Handling tasks with lower computational demands and strict timeliness at the network edge can eliminate excessive network crossing points, which not only reduces computational response time, but also alleviates the problem of signal congestion on capacity-limited backhaul links. Further, the Internet of Vehicles greatly improves the performance of in-vehicle user devices such as smartphones and wearables with limited battery capacity by offloading energy-intensive workloads to VEC nodes with higher computing power and continuous energy supply. battery life. With an appropriate task allocation strategy, the energy consumption of local computing is reduced at the expense of increasing the energy consumption of data transmission, as well as the delay caused by data transmission, workload processing on edge servers, and cross-region.

因此，实现VEC场景下的计算任务分配和功率控制是重要的问题。首先，由于车辆的快速移动造成信道状况和网络拓扑快速改变，在不同延迟约束下决定最优的任务分配比率是很困难的。而车辆也可能在数据传输或者任务处理期间离开路边单元的服务范围。其次，不同问题的任务分配变量由于VEC节点计算能力有限而相互耦合，从能量效率的角度出发，任务分配比率必须与功率控制进行联合优化。最后，由于在用户设备和VEC节点上的工作任务随机变化，无法得到确定的计算和通信资源的最佳利用方案。Therefore, it is an important issue to realize the computing task allocation and power control in the VEC scenario. First, it is difficult to decide the optimal task allocation ratio under different delay constraints due to the rapid change of channel conditions and network topology due to the rapid movement of vehicles. The vehicle may also leave the service area of the roadside unit during data transfer or task processing. Second, the task assignment variables of different problems are coupled with each other due to the limited computing power of VEC nodes. From the perspective of energy efficiency, the task assignment ratio must be jointly optimized with power control. Finally, due to random changes of work tasks on user equipment and VEC nodes, it is impossible to obtain a definite optimal utilization scheme of computing and communication resources.

发明内容SUMMARY OF THE INVENTION

为解决上述现有技术中的不足，本发明的目的是提供一种车联网中基于ADMM的任务分配与功率控制方法。通过使用本发明的算法，可以在保证时延限制的前提下，合理分配待计算任务和传输功率，有效降低用户移动设备的能量消耗。In order to solve the above deficiencies in the prior art, the purpose of the present invention is to provide a task allocation and power control method based on ADMM in the Internet of Vehicles. By using the algorithm of the present invention, the tasks to be calculated and the transmission power can be reasonably allocated under the premise of ensuring the time delay limit, and the energy consumption of the user's mobile equipment can be effectively reduced.

一种车联网中基于ADMM的任务分配与功率控制方法，包括如下步骤：A method for task allocation and power control based on ADMM in the Internet of Vehicles, comprising the following steps:

1)确定能否在车辆离开路边单元服务范围之前完成数据传输；1) Determine whether the data transfer can be completed before the vehicle leaves the service range of the roadside unit;

2)通过非线性分式优化和交替方向乘子法的迭代过程进行优化，获得使能量损耗

Figure GDA0002685470600000021
最小的计算任务分配比率和传输功率；2) Optimized through the iterative process of nonlinear fractional optimization and alternating direction multiplier method to obtain the energy loss
Figure GDA0002685470600000021
Minimum computing task allocation ratio and transmission power;

3)路边单元的服务器对分配的任务进行计算，并在中心控制器的控制下，将计算的结果通过路边单元发送给车内移动设备；3) The server of the roadside unit calculates the assigned tasks, and under the control of the central controller, sends the calculation result to the in-vehicle mobile device through the roadside unit;

从用户设备

Figure GDA0002685470600000022
到路边单元m的任务量分配比率服从平均到达率为
Figure GDA0002685470600000023
的泊松过程，在确定能否于车辆离开路边单元m的服务范围之前完成数据传输的过程中，进一步包括：from user device
Figure GDA0002685470600000022
The distribution ratio of tasks to roadside unit m obeys the average arrival rate
Figure GDA0002685470600000023
In the process of determining whether the data transmission can be completed before the vehicle leaves the service area of the roadside unit m, the Poisson process further includes:

1)在车辆进入路边单元m的服务范围时，由车速

Figure GDA0002685470600000024
和车辆与路边单元边缘的距离
Figure GDA0002685470600000025
确定最大容忍时间
Figure GDA0002685470600000026
当数据传输时间
Figure GDA0002685470600000027
小于该值时，可以进行数据传输；1) When the vehicle enters the service area of the roadside unit m, it is determined by the vehicle speed
Figure GDA0002685470600000024
and the distance between the vehicle and the edge of the roadside unit
Figure GDA0002685470600000025
Determine the maximum tolerance time
Figure GDA0002685470600000026
when data transfer time
Figure GDA0002685470600000027
When it is less than this value, data transmission can be performed;

2)在数据传输时间

Figure GDA0002685470600000028
满足要求后，若整个边缘计算过程的执行时间
Figure GDA0002685470600000029
不大于车辆离开该路段的时间
Figure GDA00026854706000000210
则按一定任务分配比率
Figure GDA00026854706000000211
将计算任务发送至路边单元；2) At the time of data transfer
Figure GDA0002685470600000028
After meeting the requirements, if the execution time of the entire edge computing process
Figure GDA0002685470600000029
Not greater than the time the vehicle leaves the road section
Figure GDA00026854706000000210
Then according to a certain task distribution ratio
Figure GDA00026854706000000211
send computing tasks to roadside units;

上述的判断过程，依据任务量分配比率，时延和能耗由本地计算过程和数据传输及边缘计算过程两部分构成，进一步包括：The above judgment process is composed of two parts: the local computing process and the data transmission and edge computing process according to the task volume allocation ratio, and further includes:

1)分配的任务首先从车内移动设备转发至车内转发器，然后车内转发器用最大传输功率将此任务发送到路边单元，全过程为两跳传输；两跳的信噪比分别表示为：1) The assigned task is first forwarded from the in-vehicle mobile device to the in-vehicle transponder, and then the in-vehicle transponder sends this task to the roadside unit with the maximum transmission power, and the whole process is two-hop transmission; the signal-to-noise ratio of the two hops respectively represents for:

Figure GDA00026854706000000212
Figure GDA00026854706000000212

Figure GDA00026854706000000213
Figure GDA00026854706000000213

其中，

Figure GDA00026854706000000214
和
Figure GDA00026854706000000215
分别代表了移动设备和转发器的传输功率
Figure GDA00026854706000000216
和
Figure GDA00026854706000000217
表示从移动设备到转发器和从转发器到路边单元的信道增益，用N0表示高斯白噪声的单边功率谱密度，并得到两跳总信噪比：in,
Figure GDA00026854706000000214
and
Figure GDA00026854706000000215
represent the transmission power of the mobile device and the transponder, respectively
Figure GDA00026854706000000216
and
Figure GDA00026854706000000217
Denote the channel gain from the mobile device to the repeater and from the repeater to the roadside unit, denote the single-sided power spectral density of white Gaussian noise by N0 , and obtain the two-hop total signal-to-noise ratio:

Figure GDA0002685470600000031
Figure GDA0002685470600000031

进而对于所传输的大小为

Figure GDA0002685470600000032
的数据包，当信道带宽为
Figure GDA0002685470600000033
时，传输时间
Figure GDA0002685470600000034
通过下式得到：And then for the transmitted size of
Figure GDA0002685470600000032
packets, when the channel bandwidth is
Figure GDA0002685470600000033
time, transmission time
Figure GDA0002685470600000034
It is obtained by the following formula:

Figure GDA0002685470600000035
Figure GDA0002685470600000035

2)对于在本地计算的任务，本地计算时间

Figure GDA0002685470600000036
由待计算任务对计算资源的需求
Figure GDA0002685470600000037
移动设备的本地计算能力
Figure GDA0002685470600000038
待计算任务对CPU资源的占有率
Figure GDA0002685470600000039
平均到达率为
Figure GDA00026854706000000310
和任务量分配比率
Figure GDA00026854706000000311
导出：2) For tasks that are computed locally, the local computation time
Figure GDA0002685470600000036
Demand for computing resources by tasks to be computed
Figure GDA0002685470600000037
The local computing power of the mobile device
Figure GDA0002685470600000038
The occupancy rate of the CPU resource of the task to be calculated
Figure GDA0002685470600000039
The average arrival rate
Figure GDA00026854706000000310
and task allocation ratio
Figure GDA00026854706000000311
Export:

Figure GDA00026854706000000312
Figure GDA00026854706000000312

3)对于被分配到路边单元的服务器进行计算的任务，等待被服务器计算的来自于不同移动设备的任务量有总到达率

Figure GDA00026854706000000313
路边单元m具有c个等同的服务器，每个服务器的计算能力为
Figure GDA00026854706000000314
在M/M/c队列模型和Erlang公式的基础上，得到计算任务在路边单元m的平均处理时间：3) For tasks that are calculated by the server assigned to the roadside unit, the amount of tasks from different mobile devices waiting to be calculated by the server has a total arrival rate
Figure GDA00026854706000000313
The roadside unit m has c equivalent servers, each with a computing power of
Figure GDA00026854706000000314
On the basis of the M/M/c queue model and Erlang formula, the average processing time of the computing task in the roadside unit m is obtained:

Figure GDA00026854706000000315
Figure GDA00026854706000000315

其中in

Figure GDA00026854706000000316
Figure GDA00026854706000000316

Figure GDA00026854706000000317
Figure GDA00026854706000000317

由于路边单元m的处理能力有限，待计算任务在必须队列中等待，然后被路边单元m处理并将结果发送给用户设备

Figure GDA00026854706000000318
因此每一个计算结果在路边单元m处的平均等待时间为：Due to the limited processing capacity of the roadside unit m, the task to be calculated waits in the necessary queue, and then is processed by the roadside unit m and the result is sent to the user equipment
Figure GDA00026854706000000318
Therefore, the average waiting time at roadside unit m for each calculation result is:

Figure GDA00026854706000000319
Figure GDA00026854706000000319

其中，

Figure GDA00026854706000000320
为路边单元m的传输处理速度，由于计算结果的数据长度远小于计算任务，计算结果从路边单元m到用户设备
Figure GDA00026854706000000321
的时延可以忽略；当准备发送计算结果时，如果车辆
Figure GDA00026854706000000322
已经运动到路边单元m的覆盖范围之外了，计算结果将首先被发送到中心控制器，然后被转发至车辆
Figure GDA0002685470600000041
所在的路边单元m'；此过程中的传输延时
Figure GDA0002685470600000042
在中心控制器的平均等待时间
Figure GDA0002685470600000043
和在路边单元m'的等待时间
Figure GDA0002685470600000044
可以认为是常量，因此跨区的延时可以表达为：in,
Figure GDA00026854706000000320
is the transmission processing speed of the roadside unit m. Since the data length of the calculation result is much smaller than the calculation task, the calculation result is transmitted from the roadside unit m to the user equipment.
Figure GDA00026854706000000321
The delay can be ignored; when preparing to send the calculation results, if the vehicle
Figure GDA00026854706000000322
has moved out of the coverage area of roadside unit m, the calculation result will first be sent to the central controller and then forwarded to the vehicle
Figure GDA0002685470600000041
The roadside unit m' where it is located; the transmission delay in this process
Figure GDA0002685470600000042
Average wait time at the central controller
Figure GDA0002685470600000043
and the waiting time at roadside unit m'
Figure GDA0002685470600000044
It can be considered as a constant, so the delay across regions can be expressed as:

Figure GDA0002685470600000045
Figure GDA0002685470600000045

4)对整个移动边缘计算过程的执行时间

Figure GDA0002685470600000046
有
Figure GDA0002685470600000047
4) Execution time for the entire mobile edge computing process
Figure GDA0002685470600000046
Have
Figure GDA0002685470600000047

用户设备

Figure GDA0002685470600000048
的能量损耗应包括本地计算的能量消耗和传输数据的能量消耗；定义
Figure GDA0002685470600000049
为本地计算功率，它取决于CPU的固有特性和工作负载的复杂性，在任务执行期间可以被视为常量；User equipment
Figure GDA0002685470600000048
The energy consumption should include the energy consumption of local computing and the energy consumption of transmitting data; define
Figure GDA0002685470600000049
for local computing power, which depends on the inherent characteristics of the CPU and the complexity of the workload, and can be treated as constant during task execution;

通过下得到用户设备

Figure GDA00026854706000000410
的本地计算能量消耗：Get user equipment by
Figure GDA00026854706000000410
The local computing energy consumption of:

Figure GDA00026854706000000411
Figure GDA00026854706000000411

通过下式得到用户设备

Figure GDA00026854706000000412
向车内转发器发送数据的能量损耗：Get the user equipment by the following formula
Figure GDA00026854706000000412
Energy consumption for sending data to the in-vehicle transponder:

Figure GDA00026854706000000413
Figure GDA00026854706000000413

通过下式得到用户设备

Figure GDA00026854706000000414
的总能量损耗：Get the user equipment by the following formula
Figure GDA00026854706000000414
The total energy loss of:

Figure GDA00026854706000000415
Figure GDA00026854706000000415

能耗优化方案为基于ADMM的计算任务分配和功率控制方案，其目标为最小化路边单元m服务范围内mk辆车的整体能耗，定义优化变量集合

Figure GDA00026854706000000416
其中
Figure GDA00026854706000000417
则优化问题为：The energy consumption optimization scheme is a computing task allocation and power control scheme based on ADMM, and its goal is to minimize the overall energy consumption of m k vehicles within the service range of roadside unit m, and define a set of optimization variables.
Figure GDA00026854706000000416
in
Figure GDA00026854706000000417
Then the optimization problem is:

P1:

Figure GDA00026854706000000418
P1:
Figure GDA00026854706000000418

s.t.s.t.

Figure GDA00026854706000000419
Figure GDA00026854706000000419

Figure GDA00026854706000000420
Figure GDA00026854706000000420

Figure GDA00026854706000000421
Figure GDA00026854706000000421

Figure GDA00026854706000000422
Figure GDA00026854706000000422

Figure GDA00026854706000000423
Figure GDA00026854706000000423

Figure GDA00026854706000000424
Figure GDA00026854706000000424

C1和C2为限制了工作负载的到达率

Figure GDA00026854706000000425
和
Figure GDA00026854706000000426
分别不能超过用户设备
Figure GDA00026854706000000427
和路边单元m的处理速率,C3确保了传输功率不超过用户设备的最大传输功率,C4和C5分别为数据传输和任务计算过程的延迟限制,C6为任务分配比率
Figure GDA0002685470600000051
的边界限制； C1 and C2 limit the arrival rate of the workload
Figure GDA00026854706000000425
and
Figure GDA00026854706000000426
respectively cannot exceed the user equipment
Figure GDA00026854706000000427
and the processing rate of the roadside unit m , C3 ensures that the transmission power does not exceed the maximum transmission power of the user equipment, C4 and C5 are the delay limits of the data transmission and task calculation process, respectively, and C6 is the task allocation ratio
Figure GDA0002685470600000051
boundary limits;

在P1中，因为不同的用户设备

Figure GDA0002685470600000052
的任务分配变量是耦合的，因此优化目标是不可分离的；为了解决该问题，进一步包括以下步骤：In P1, because different user equipment
Figure GDA0002685470600000052
The task assignment variables of are coupled, so the optimization objective is inseparable; to solve this problem, the following steps are further included:

1)引入最优资源分配策略的本地副本；使用一组新的变量来表示局部优化变量，定义

Figure GDA0002685470600000053
和
Figure GDA0002685470600000054
分别作为
Figure GDA0002685470600000055
和
Figure GDA0002685470600000056
的本地变量，则本地优化变量的集合被定义为
Figure GDA0002685470600000057
其中
Figure GDA0002685470600000058
1) Introduce a local copy of the optimal resource allocation strategy; use a new set of variables to represent local optimization variables, define
Figure GDA0002685470600000053
and
Figure GDA0002685470600000054
respectively as
Figure GDA0002685470600000055
and
Figure GDA0002685470600000056
, the set of local optimization variables is defined as
Figure GDA0002685470600000057
in
Figure GDA0002685470600000058

则P1的次优问题可以表达为：Then the suboptimal problem of P1 can be expressed as:

P2:

Figure GDA0002685470600000059
P2:
Figure GDA0002685470600000059

s.t.s.t.

Figure GDA00026854706000000510
Figure GDA00026854706000000510

Figure GDA00026854706000000511
Figure GDA00026854706000000511

Figure GDA00026854706000000512
Figure GDA00026854706000000512

Figure GDA00026854706000000513
Figure GDA00026854706000000513

Figure GDA00026854706000000514
Figure GDA00026854706000000514

Figure GDA00026854706000000515
Figure GDA00026854706000000515

2)P2通过引入局部变量使目标函数

Figure GDA00026854706000000516
可分离，将目标函数分解为mK个可以被并行解决的子问题，这些分散的联合优化问题可以被表达为：2) P2 makes the objective function by introducing local variables
Figure GDA00026854706000000516
Separable, decomposing the objective function into m K sub-problems that can be solved in parallel, these decentralized joint optimization problems can be expressed as:

P3:

Figure GDA00026854706000000517
P3:
Figure GDA00026854706000000517

s.t.

Figure GDA00026854706000000518
st
Figure GDA00026854706000000518

目标函数P3依然是一个非凸问题，将P3的分子和分母分别定义为：The objective function P3 is still a non-convex problem, and the numerator and denominator of P3 are defined as:

Figure GDA00026854706000000519
Figure GDA00026854706000000519

Figure GDA0002685470600000061
Figure GDA0002685470600000061

并定义

Figure GDA0002685470600000062
作为P3的最优目标函数值：and define
Figure GDA0002685470600000062
As the optimal objective function value of P3:

Figure GDA0002685470600000063
Figure GDA0002685470600000063

其中

Figure GDA0002685470600000064
和
Figure GDA0002685470600000065
分别代表了最优本地计算任务分配比率和功率控制策略；in
Figure GDA0002685470600000064
and
Figure GDA0002685470600000065
represent the optimal local computing task allocation ratio and power control strategy, respectively;

3)根据非线性分式优化问题，获得最优目标值

Figure GDA0002685470600000066
的充分必要条件是：当且仅当方程3) According to the nonlinear fractional optimization problem, the optimal target value is obtained
Figure GDA0002685470600000066
The necessary and sufficient conditions are: if and only if the equation

Figure GDA0002685470600000067
Figure GDA0002685470600000067

成立，即通过解决下面的问题得到最优的本地优化变量

Figure GDA0002685470600000068
和
Figure GDA0002685470600000069
：is established, that is, the optimal local optimization variables are obtained by solving the following problems
Figure GDA0002685470600000068
and
Figure GDA0002685470600000069
:

P3:

Figure GDA00026854706000000610
P3:
Figure GDA00026854706000000610

s.t.

Figure GDA00026854706000000611
st
Figure GDA00026854706000000611

4)为每个用户设备定义本地变量集合

Figure GDA00026854706000000612
并定义函数：4) Define a set of local variables for each user device
Figure GDA00026854706000000612
and define the function:

Figure GDA00026854706000000613
Figure GDA00026854706000000613

由此，关于P2的凸优化问题可以表达为：From this, the convex optimization problem on P2 can be expressed as:

P5:

Figure GDA00026854706000000614
P5:
Figure GDA00026854706000000614

s.t.

Figure GDA00026854706000000615
st
Figure GDA00026854706000000615

5)定义关联于P5的最优变量集合

Figure GDA00026854706000000616
在权利要求1步骤2)迭代算法的每一次迭代过程中，下面的问题被解决：5) Define the optimal set of variables associated with P5
Figure GDA00026854706000000616
During each iteration of the iterative algorithm of claim 1, step 2), the following problems are solved:

P6:

Figure GDA00026854706000000617
P6:
Figure GDA00026854706000000617

s.t.

Figure GDA0002685470600000071
st
Figure GDA0002685470600000071

其中最优解

Figure GDA0002685470600000072
在前一步迭代中获得，当限制条件
Figure GDA0002685470600000073
被满足时
Figure GDA0002685470600000074
是所求优化问题P1的一组最优解；The optimal solution
Figure GDA0002685470600000072
obtained in the previous iteration, when the constraints
Figure GDA0002685470600000073
when satisfied
Figure GDA0002685470600000074
is a set of optimal solutions to the optimization problem P1;

对于迭代过程，定义对应于方程P6的拉格朗日乘子集合

Figure GDA0002685470600000075
定义正常数ρ调整收敛速度，则P6的增广拉格朗日公式可以被表达为：For the iterative process, define the set of Lagrange multipliers corresponding to equation P6
Figure GDA0002685470600000075
Define the constant ρ to adjust the convergence rate, then the augmented Lagrangian formula of P6 can be expressed as:

Figure GDA0002685470600000076
Figure GDA0002685470600000076

该迭代过程包含两层循环，外循环为非线性分式优化问题，用n来指示迭代次数；内循环为原始变量和对偶变量的更新，用t来指示迭代次数，The iterative process consists of two layers of loops. The outer loop is a nonlinear fractional optimization problem, and n is used to indicate the number of iterations; the inner loop is the update of the original variable and the dual variable, and t is used to indicate the number of iterations.

进一步包括：Further includes:

1)对工作任务分配比率

Figure GDA0002685470600000077
传输功率
Figure GDA0002685470600000078
和最优解
Figure GDA0002685470600000079
初始化，设置终止条件ε；1) Assignment ratio to work tasks
Figure GDA0002685470600000077
Transmission power
Figure GDA0002685470600000078
and the optimal solution
Figure GDA0002685470600000079
Initialize, set the termination condition ε;

2)更新优化变量集合

Figure GDA00026854706000000710
给定第n次外循环的最优解
Figure GDA00026854706000000711
进而获得每个用户设备的传输功率
Figure GDA00026854706000000712
本地变量
Figure GDA00026854706000000713
和
Figure GDA00026854706000000714
的更新可以被分解为能够并行解决的mK个子问题；根据下式计算用户设备
Figure GDA00026854706000000715
在第t次内循环时的获得的最优任务分配比率
Figure GDA00026854706000000716
和传输功率
Figure GDA00026854706000000717
2) Update the set of optimization variables
Figure GDA00026854706000000710
The optimal solution given the nth outer loop
Figure GDA00026854706000000711
And then obtain the transmission power of each user equipment
Figure GDA00026854706000000712
local variable
Figure GDA00026854706000000713
and
Figure GDA00026854706000000714
The update of can be decomposed into m K sub-problems that can be solved in parallel; the user equipment is calculated according to the following formula
Figure GDA00026854706000000715
The obtained optimal task allocation ratio at the t-th inner loop
Figure GDA00026854706000000716
and transmission power
Figure GDA00026854706000000717

Figure GDA00026854706000000718
Figure GDA00026854706000000718

3)更新

Figure GDA00026854706000000719
根据下式获得第t+1次内循环时的全局最优任务分配比率
Figure GDA00026854706000000720
3) Update
Figure GDA00026854706000000719
The global optimal task allocation ratio at the t+1th inner loop is obtained according to the following formula
Figure GDA00026854706000000720

Figure GDA00026854706000000721
Figure GDA00026854706000000721

根据下式获得第t+1次内循环时的拉格朗日乘子

Figure GDA00026854706000000722
The Lagrange multiplier at the t+1th inner loop is obtained according to the following formula
Figure GDA00026854706000000722

Figure GDA00026854706000000723
Figure GDA00026854706000000723

4)更新最优解

Figure GDA0002685470600000081
在ADMM的初始变量和对偶变量的迭代过程中，当t趋于下确界时，满足目标函数收敛，残差收敛和对偶变量收敛条件；第n次迭代的内循环终止时得到
Figure GDA0002685470600000082
和
Figure GDA0002685470600000083
则第n+1次迭代的最优解
Figure GDA0002685470600000084
按下式得到：4) Update the optimal solution
Figure GDA0002685470600000081
In the iterative process of the initial variable and dual variable of ADMM, when t tends to the infimum, the objective function convergence, residual convergence and dual variable convergence conditions are satisfied; when the inner loop of the nth iteration terminates, the
Figure GDA0002685470600000082
and
Figure GDA0002685470600000083
Then the optimal solution of the n+1th iteration
Figure GDA0002685470600000084
Get it as follows:

Figure GDA0002685470600000085
Figure GDA0002685470600000085

5)循环终止；当第n次外层循环满足

Figure GDA0002685470600000086
时，通过下式获得最优任务分配比率
Figure GDA0002685470600000087
最优传输功率
Figure GDA0002685470600000088
和最优解
Figure GDA0002685470600000089
5) The loop terminates; when the nth outer loop satisfies
Figure GDA0002685470600000086
When , the optimal task allocation ratio is obtained by the following formula
Figure GDA0002685470600000087
optimal transmission power
Figure GDA0002685470600000088
and the optimal solution
Figure GDA0002685470600000089

Figure GDA00026854706000000810
Figure GDA00026854706000000810

Figure GDA00026854706000000811
Figure GDA00026854706000000811

Figure GDA00026854706000000812
Figure GDA00026854706000000812

与最接近的现有技术相比，本发明提供的技术方案具有的有益效果是：Compared with the closest prior art, the beneficial effects of the technical solution provided by the present invention are:

本发明介绍了怎样去实现有较高能效的车联网边缘计算方法，通过交替方向乘子法和非线性分式优化解决该能量消耗最小化问题，并考虑到包括本地计算和数据传输在内的能量消耗及由本地计算、数据传输、在VEC节点和路边单元的等待时间以及跨区造成的延迟。在VEC节点计算能力的约束条件下，提出分数形式的目标函数和耦合的优化变量，形成NP难问题。The invention introduces how to realize the edge computing method of the Internet of Vehicles with higher energy efficiency, solves the problem of minimizing energy consumption through the alternate direction multiplier method and nonlinear fractional optimization, and takes into account the problems including local computing and data transmission. Energy consumption and delays caused by local computation, data transfer, latency at VEC nodes and roadside units, and cross-region. Under the constraints of the computing power of VEC nodes, a fractional objective function and coupled optimization variables are proposed to form NP-hard problems.

为了更好的搭建多任务多服务器计算模式，引入排队论。在考虑队列异质性的情况下，推导出在用户设备和VEC节点处的动态传输模型。并假设每个用户设备产生的工作量服从泊松分布，且任何一个用户设备和VEC节点的服务时间遵循指数分布，由此在用户设备和VEC节点的任务传输模型可以分别被视为M/M/1队列和M/M/c队列。In order to better build a multi-task and multi-server computing model, queuing theory is introduced. Taking into account queue heterogeneity, a dynamic transmission model at user equipment and VEC nodes is derived. And it is assumed that the workload generated by each user equipment obeys the Poisson distribution, and the service time of any user equipment and VEC node follows the exponential distribution, so the task transmission model in the user equipment and the VEC node can be regarded as M/M respectively. /1 queue and M/M/c queue.

附图说明Description of drawings

图1是本发明提供的车联网边缘计算系统图；1 is a diagram of an edge computing system for the Internet of Vehicles provided by the present invention;

图2是本发明提供的不同功率下归一化能量消耗随任务分配比率变化图；Fig. 2 is the variation diagram of normalized energy consumption with task allocation ratio under different power provided by the present invention;

图3是本发明提供的不同任务分配比率下归一化能量消耗随传输功率变化图；Fig. 3 is the change diagram of normalized energy consumption with transmission power under different task allocation ratios provided by the present invention;

图4是本发明提供的不同算法中能量消耗与用户设备数量关系图；Fig. 4 is the relation diagram of energy consumption and the quantity of user equipment in different algorithms provided by the present invention;

图5是本发明提供的不同功率下归一化能量消耗随路边单元服务半径变化图；FIG. 5 is a graph showing the variation of normalized energy consumption with the service radius of roadside units under different powers provided by the present invention;

图6是本发明提供的算法收敛性与迭代次数关系图；Fig. 6 is the relation diagram of algorithm convergence and iteration times provided by the present invention;

图7是本发明提供的不同用户设备数量下归一化能量消耗随任务分配比率变化图。FIG. 7 is a graph showing the variation of normalized energy consumption with task allocation ratio under different numbers of user equipment provided by the present invention.

具体实施方式Detailed ways

下面结合附图对本发明的具体实施方式作进一步的详细说明。The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

以下描述和附图充分地示出本发明的具体实施方案，以使本领域的技术人员能够实践它们。其他实施方案可以包括结构的、逻辑的、电气的、过程的以及其他的改变。实施例仅代表可能的变化。除非明确要求，否则单独的组件和功能是可选的，并且操作的顺序可以变化。一些实施方案的部分和特征可以被包括在或替换其他实施方案的部分和特征。本发明的实施方案的范围包括权利要求书的整个范围，以及权利要求书的所有可获得的等同物。在本文中，本发明的这些实施方案可以被单独地或总地用术语“发明”来表示，这仅仅是为了方便，并且如果事实上公开了超过一个的发明，不是要自动地限制该应用的范围为任何单个发明或发明构思。The following description and drawings sufficiently illustrate specific embodiments of the invention to enable those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, process, and other changes. The examples are only representative of possible variations. Unless explicitly required, individual components and functions are optional and the order of operations may vary. Portions and features of some embodiments may be included in or substituted for those of other embodiments. The scope of embodiments of the invention includes the full scope of the claims, along with all available equivalents of the claims. These embodiments of the invention may be referred to herein by the term "invention," individually or collectively, for convenience only and not to automatically limit the application if more than one invention is in fact disclosed. The scope is any single invention or inventive concept.

实施例一Example 1

本发明模拟在车联网场景下多任务多服务器的场景，考虑到车辆较快的移动速度，待计算任务可能无法在一个路边单元的服务范围传输完毕，而接收计算结果时也会存在跨区问题。通过判断能否将任务分派到路边单元进行计算，以及协调任务分配的比率及传输功率，可以在保证时延要求的条件下降低用户设备能耗。通过中心控制器的调控，确定车辆所在位置属于的路边单元，完成计算结果的回传。同时需要考虑到由于多用户产生多任务，而路边单元的计算能力及存储能力有限而造成的等待时间。其系统模型图如图1所示，整个过程包括以下内容：The present invention simulates the scenario of multi-task and multi-server in the scenario of the Internet of Vehicles. Considering the relatively fast moving speed of the vehicle, the task to be calculated may not be transmitted within the service range of a roadside unit, and there will also be cross-area when receiving the calculation result. question. By judging whether the task can be assigned to the roadside unit for calculation, and by coordinating the task assignment ratio and transmission power, the energy consumption of the user equipment can be reduced under the condition of ensuring the delay requirement. Through the regulation of the central controller, the roadside unit to which the vehicle location belongs is determined, and the return of the calculation result is completed. At the same time, it is necessary to take into account the waiting time caused by the limited computing power and storage capacity of the roadside unit due to multi-tasking by multiple users. Its system model diagram is shown in Figure 1, and the whole process includes the following:

1、判断能否在路边单元的服务范围内完成数据传输。用户设备

Figure GDA0002685470600000091
分配到路边单元m的任务分配比率服从平均到达率为
Figure GDA0002685470600000092
的泊松过程，车辆以速度
Figure GDA0002685470600000093
运动，距路边单元边缘的距离为
Figure GDA0002685470600000094
时，数据传输时间
Figure GDA0002685470600000095
需不大于
Figure GDA0002685470600000096
且整个边缘计算过程的执行总时间
Figure GDA0002685470600000097
不大于车辆离开该路段的时间
Figure GDA0002685470600000098
满足上述条件时，以任务分配比率
Figure GDA0002685470600000101
将计算任务发送至路边单元。1. Determine whether data transmission can be completed within the service range of the roadside unit. User equipment
Figure GDA0002685470600000091
The assignment ratio of tasks assigned to roadside unit m obeys the average arrival rate
Figure GDA0002685470600000092
The Poisson process of the vehicle at speed
Figure GDA0002685470600000093
movement, the distance from the edge of the roadside unit is
Figure GDA0002685470600000094
time, data transfer time
Figure GDA0002685470600000095
no more than
Figure GDA0002685470600000096
And the total execution time of the entire edge computing process
Figure GDA0002685470600000097
Not greater than the time the vehicle leaves the road section
Figure GDA0002685470600000098
When the above conditions are met, the task distribution ratio is
Figure GDA0002685470600000101
Send computing tasks to roadside units.

2、算数执行执行过程中的时延主要由传输时间、等待时间和计算时间构成。2. The delay in the execution of arithmetic is mainly composed of transmission time, waiting time and calculation time.

1)计算任务从车内用户设备发送到路边单元为两跳传输。其中从车内移动设备到车内转发器的信噪比为

Figure GDA0002685470600000102
从车内转发器大到路边单元的信噪比为
Figure GDA0002685470600000103
故而总信噪比为
Figure GDA0002685470600000104
在所传输数据包的大小为
Figure GDA0002685470600000105
信道带宽为
Figure GDA0002685470600000106
时，传输时间
Figure GDA0002685470600000107
通过下式得到：1) The computing task is sent from the in-vehicle user equipment to the roadside unit as a two-hop transmission. Among them, the signal-to-noise ratio from the in-vehicle mobile device to the in-vehicle transponder is
Figure GDA0002685470600000102
The signal-to-noise ratio from the in-vehicle transponder to the roadside unit is
Figure GDA0002685470600000103
So the total signal-to-noise ratio is
Figure GDA0002685470600000104
The size of the transmitted packet is
Figure GDA0002685470600000105
The channel bandwidth is
Figure GDA0002685470600000106
time, transmission time
Figure GDA0002685470600000107
It is obtained by the following formula:

Figure GDA0002685470600000108
Figure GDA0002685470600000108

2)有比例为

Figure GDA0002685470600000109
的任务在本地计算，其对计算资源的需求为
Figure GDA00026854706000001010
对CPU资源的占有率
Figure GDA00026854706000001011
用户设备的本地计算能力
Figure GDA00026854706000001012
则得到本地计算时间：2) There is a ratio of
Figure GDA0002685470600000109
The tasks are calculated locally, and their demand for computing resources is
Figure GDA00026854706000001010
Occupancy of CPU resources
Figure GDA00026854706000001011
The local computing power of the user equipment
Figure GDA00026854706000001012
Then get the local computation time:

Figure GDA00026854706000001013
Figure GDA00026854706000001013

3)有比例为

Figure GDA00026854706000001014
的任务被分配到路边单元计算。路边单元配备有c个等同的计算能力为
Figure GDA00026854706000001015
的服务器。由于一个路边单元m的服务范围内会有多个用户设备传输计算任务，其总的到达率为
Figure GDA00026854706000001016
在M/M/c队列模型和Erlang公式的基础上，得到待计算任务在路边单元m的平均计算时间:3) There is a ratio of
Figure GDA00026854706000001014
The tasks are assigned to the roadside unit for computation. The roadside unit is equipped with c equivalent computing power for
Figure GDA00026854706000001015
server. Since there will be multiple user equipments transmitting computing tasks within the service range of a roadside unit m, the total arrival rate is
Figure GDA00026854706000001016
On the basis of the M/M/c queue model and Erlang formula, the average computation time of the task to be computed in the roadside unit m is obtained:

Figure GDA00026854706000001017
Figure GDA00026854706000001017

其中

Figure GDA00026854706000001018
in
Figure GDA00026854706000001018

4)由于路边单元m的处理能力有限，待计算任务必须在队列中等待。路边单元m的传输处理速度为

Figure GDA00026854706000001019
则每一个计算结果在路边单元m处的平均等待时间为：4) Due to the limited processing capacity of the roadside unit m, the task to be calculated must wait in the queue. The transmission processing speed of the roadside unit m is
Figure GDA00026854706000001019
Then the average waiting time of each calculation result at the roadside unit m is:

Figure GDA00026854706000001020
Figure GDA00026854706000001020

5)当准备发送计算结果时，如果车辆

Figure GDA0002685470600000111
已经运动到路边单元m的服务范围之外了，计算结果将首先被发送到中心控制器，然后被转发至车辆
Figure GDA0002685470600000112
所在的路边单元m'。此过程中的传输延时
Figure GDA0002685470600000113
在中心控制器的平均等待时间
Figure GDA0002685470600000114
和在路边单元m'的等待时间
Figure GDA0002685470600000115
可以认为是常量，由于计算结果的数据长度远小于计算任务，计算结果从路边单元m到用户设备
Figure GDA0002685470600000116
的时延可以忽略。因此跨区的延时可以表达为：5) When preparing to send the calculation result, if the vehicle
Figure GDA0002685470600000111
has moved out of the service range of roadside unit m, the calculation result will first be sent to the central controller and then forwarded to the vehicle
Figure GDA0002685470600000112
The roadside unit m' where it is located. Transmission delay in this process
Figure GDA0002685470600000113
Average wait time at the central controller
Figure GDA0002685470600000114
and the waiting time at roadside unit m'
Figure GDA0002685470600000115
It can be considered as a constant. Since the data length of the calculation result is much smaller than the calculation task, the calculation result goes from the roadside unit m to the user equipment.
Figure GDA0002685470600000116
delay can be ignored. Therefore, the delay across regions can be expressed as:

Figure GDA0002685470600000117
Figure GDA0002685470600000117

6)移动边缘计算过程的全部执行时间

Figure GDA0002685470600000118
可以表达为：6) The full execution time of the mobile edge computing process
Figure GDA0002685470600000118
can be expressed as:

Figure GDA0002685470600000119
Figure GDA0002685470600000119

3、计算过程中，用户设备的能量损耗主要包括本地计算的能量消耗和数据传输的能量消耗。3. During the calculation process, the energy consumption of the user equipment mainly includes the energy consumption of local computing and the energy consumption of data transmission.

1)本地计算功率

Figure GDA00026854706000001110
由CPU的固有特性和工作负载的复杂性决定，在任务计算期间可以视作常量，则本地计算的能量消耗为：1) Local computing power
Figure GDA00026854706000001110
Determined by the inherent characteristics of the CPU and the complexity of the workload, it can be regarded as a constant during the task calculation, and the energy consumption of the local calculation is:

Figure GDA00026854706000001111
Figure GDA00026854706000001111

2)用户设备

Figure GDA00026854706000001112
的数据传输功率为
Figure GDA00026854706000001113
则用户设备向车内转发器发送数据的能量损耗为：2) User equipment
Figure GDA00026854706000001112
The data transmission power is
Figure GDA00026854706000001113
Then the energy consumption of the user equipment sending data to the in-vehicle transponder is:

Figure GDA00026854706000001114
Figure GDA00026854706000001114

3)在边缘计算执行过程中，用户设备的总能量损耗为：3) During the execution of edge computing, the total energy consumption of the user equipment is:

Figure GDA00026854706000001115
Figure GDA00026854706000001115

实施例二、Embodiment two,

本发明的优化算法分为两层迭代过程，外层迭代过程解决非线性分式优化问题，内层迭代过程对变量进行更新。其目标为最小化路边单元m服务范围内mk辆车的整体能耗。该问题被表达为：The optimization algorithm of the present invention is divided into two layers of iterative processes, the outer layer iterative process solves the nonlinear fractional optimization problem, and the inner layer iterative process updates variables. Its goal is to minimize the overall energy consumption of m k vehicles within the service area of roadside unit m. The problem is expressed as:

Figure GDA0002685470600000121
Figure GDA0002685470600000121

s.t.s.t.

Figure GDA0002685470600000122
Figure GDA0002685470600000122

Figure GDA0002685470600000123
Figure GDA0002685470600000123

Figure GDA0002685470600000124
Figure GDA0002685470600000124

Figure GDA0002685470600000125
Figure GDA0002685470600000125

Figure GDA0002685470600000126
Figure GDA0002685470600000126

Figure GDA0002685470600000127
Figure GDA0002685470600000127

由于不同的用户设备

Figure GDA0002685470600000128
的任务分配变量是耦合的，因此优化目标不可分离。为了解决该问题，引入最优资源分配策略的本地副本并定义局部优化变量，使目标函数是可分离的：Due to different user equipment
Figure GDA0002685470600000128
The task assignment variables of are coupled, so the optimization objectives are not separable. To solve this problem, a local copy of the optimal resource allocation strategy is introduced and local optimization variables are defined to make the objective function separable:

Figure GDA0002685470600000129
Figure GDA0002685470600000129

s.t.s.t.

Figure GDA00026854706000001210
Figure GDA00026854706000001210

Figure GDA00026854706000001211
Figure GDA00026854706000001211

Figure GDA00026854706000001212
Figure GDA00026854706000001212

Figure GDA00026854706000001213
Figure GDA00026854706000001213

Figure GDA00026854706000001214
Figure GDA00026854706000001214

Figure GDA00026854706000001215
Figure GDA00026854706000001215

因此目标函数可以被分解为mK能并行解决的子问题，该问题为非凸问题。通过进一步数学变换，并定义目标函数值为

Figure GDA00026854706000001216
可以将该问题转化为凸优化问题，进而可以在迭代过程中被优化。加入限制条件
Figure GDA00026854706000001217
后在每一次迭代时，下面的问题Therefore, the objective function can be decomposed into sub-problems that can be solved in parallel by m K , which is a non-convex problem. Through further mathematical transformation, and define the objective function as
Figure GDA00026854706000001216
This problem can be transformed into a convex optimization problem, which in turn can be optimized in an iterative process. Add restrictions
Figure GDA00026854706000001217
After each iteration, the following question

Figure GDA0002685470600000131
Figure GDA0002685470600000131

被解决：solved:

当限制条件

Figure GDA0002685470600000132
被满足时，所得结果即为该优化问题的最优解。通过解决下面的增广拉格朗日问题获得每一次内层迭代更新的变量及此次外层循环过程的最优解：when restrictive
Figure GDA0002685470600000132
When satisfied, the result obtained is the optimal solution of the optimization problem. The variables updated in each inner iteration and the optimal solution of the outer loop process are obtained by solving the following augmented Lagrangian problem:

Figure GDA0002685470600000133
Figure GDA0002685470600000133

对变量进行初始化之后，在对偶变量的迭代过程中，满足目标函数收敛，残差收敛和对偶变量收敛条件后可以得到此次外层循环的最优解，并在满足设定的循环终止条件后，获得所求目标函数的最优解

Figure GDA0002685470600000134
及最佳任务分配比率
Figure GDA0002685470600000135
和最佳传输功率
Figure GDA0002685470600000136
After the variables are initialized, in the iterative process of the dual variables, the optimal solution of the outer loop can be obtained after satisfying the objective function convergence, residual convergence and dual variable convergence conditions, and after satisfying the set loop termination conditions , to obtain the optimal solution of the desired objective function
Figure GDA0002685470600000134
and optimal task allocation ratio
Figure GDA0002685470600000135
and optimal transmission power
Figure GDA0002685470600000136

对于本发明，我们进行了大量仿真实验。如图2，随着

Figure GDA0002685470600000137
的增长，即更多的任务被分配到边缘计算节点进行计算，能量损耗先降低后增加。这是由于在
Figure GDA0002685470600000138
较小时，数据传输的能量要少于本地计算的能量，而后随着数据传输消耗的能量多于本地计算的能量。图3显示了随着传输功率的增加，传输速率在增大，然而数据传输所消耗能量的增长速度要快于传输速率的增加，因此表现为传输能量损耗的单调增加。图4反映了用户设备数量对三种不同优化方案下的能量损耗的影响。图5显示出随着路边单元覆盖范围的增加，能量损耗逐渐减少并趋于稳定。在图6所示的过程中，表明该算法的迭代可以在6～7次迭代中迅速收敛，即可以快速获得最优解。图7为不同用户设备数量下，能量损耗与任务分配比率的关系。该结果与图2、图4的结论一致。For the present invention, we have conducted a large number of simulation experiments. As shown in Figure 2, with
Figure GDA0002685470600000137
With the increase of , that is, more tasks are allocated to edge computing nodes for computing, and the energy consumption first decreases and then increases. This is due to the
Figure GDA0002685470600000138
When it is small, the energy of the data transmission is less than the energy of the local calculation, and then the energy consumed by the data transmission is more than the energy of the local calculation. Figure 3 shows that as the transmission power increases, the transmission rate increases, however, the energy consumed by data transmission increases faster than the transmission rate, thus showing a monotonous increase in transmission energy loss. Figure 4 reflects the effect of the number of user equipments on the energy consumption under three different optimization schemes. Figure 5 shows that the energy loss decreases gradually and stabilizes as the coverage of the roadside unit increases. In the process shown in Figure 6, it is shown that the iteration of the algorithm can quickly converge in 6 to 7 iterations, that is, the optimal solution can be obtained quickly. Figure 7 shows the relationship between energy consumption and task allocation ratio under different numbers of user equipment. This result is consistent with the conclusions of Figures 2 and 4.

以上实施例仅用以说明本发明的技术方案而非对其限制，尽管参照上述实施例对本发明进行了详细的说明，所属领域的普通技术人员依然可以对本发明的具体实施方式进行修改或者等同替换，这些未脱离本发明精神和范围的任何修改或者等同替换，均在申请待批的本发明的权利要求保护范围之内。The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art can still modify or equivalently replace the specific embodiments of the present invention. , any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention are all within the protection scope of the claims of the present invention for which the application is pending.

Claims (1)

1.一种车联网中基于ADMM的任务分配与功率控制方法，其特征在于，包括如下步骤：1. a task distribution and power control method based on ADMM in the Internet of Vehicles, is characterized in that, comprises the steps: 1)确定能否在车辆离开路边单元服务范围之前完成数据传输；1) Determine whether the data transfer can be completed before the vehicle leaves the service range of the roadside unit; 2)通过非线性分式优化和交替方向乘子法的迭代过程进行优化，获得使能量损耗
Figure FDA0002810302750000011
最小的计算任务分配比率和传输功率；2) Optimized through the iterative process of nonlinear fractional optimization and alternating direction multiplier method to obtain the energy loss
Figure FDA0002810302750000011
Minimum computing task allocation ratio and transmission power; 3)路边单元的服务器对分配的任务进行计算，并在中心控制器的控制下，将计算的结果通过路边单元发送给车内移动设备；3) The server of the roadside unit calculates the assigned tasks, and under the control of the central controller, sends the calculation result to the in-vehicle mobile device through the roadside unit; 从用户设备
Figure FDA0002810302750000012
到路边单元m的任务量分配比率服从平均到达率为
Figure FDA0002810302750000013
的泊松过程，在确定能否于车辆离开路边单元m的服务范围之前完成数据传输的过程中，进一步包括：from user device
Figure FDA0002810302750000012
The distribution ratio of tasks to roadside unit m obeys the average arrival rate
Figure FDA0002810302750000013
In the process of determining whether the data transmission can be completed before the vehicle leaves the service area of the roadside unit m, the Poisson process further includes: 1.1在车辆进入路边单元m的服务范围时，由车速
Figure FDA0002810302750000014
和车辆与路边单元边缘的距离
Figure FDA0002810302750000015
确定最大容忍时间
Figure FDA0002810302750000016
当数据传输时间
Figure FDA0002810302750000017
小于该值时，进行数据传输；1.1 When the vehicle enters the service area of the roadside unit m, it is determined by the vehicle speed
Figure FDA0002810302750000014
and the distance between the vehicle and the edge of the roadside unit
Figure FDA0002810302750000015
Determine the maximum tolerance time
Figure FDA0002810302750000016
when data transfer time
Figure FDA0002810302750000017
When it is less than this value, data transmission is performed; 1.2在数据传输时间
Figure FDA0002810302750000018
满足要求后，若整个边缘计算过程的执行时间
Figure FDA0002810302750000019
不大于车辆离开该路段的时间
Figure FDA00028103027500000110
则按一定任务分配比率
Figure FDA00028103027500000111
将计算任务发送至路边单元；1.2 At data transfer time
Figure FDA0002810302750000018
After meeting the requirements, if the execution time of the entire edge computing process
Figure FDA0002810302750000019
Not greater than the time the vehicle leaves the road section
Figure FDA00028103027500000110
Then according to a certain task distribution ratio
Figure FDA00028103027500000111
send computing tasks to roadside units; 依据任务量分配比率，时延和能耗由本地计算过程和数据传输及边缘计算过程两部分构成，进一步包括：According to the task volume distribution ratio, the delay and energy consumption are composed of two parts: the local computing process and the data transmission and edge computing process, further including: 1.21分配的任务首先从车内移动设备转发至车内转发器，然后车内转发器用最大传输功率将此任务发送到路边单元，全过程为两跳传输；两跳的信噪比分别表示为：1.21 The assigned task is first forwarded from the in-vehicle mobile device to the in-vehicle transponder, and then the in-vehicle transponder sends this task to the roadside unit with the maximum transmission power. The whole process is two-hop transmission; the signal-to-noise ratio of the two hops is expressed as :
Figure FDA00028103027500000112
Figure FDA00028103027500000112
Figure FDA00028103027500000113
Figure FDA00028103027500000113
 其中，
Figure FDA00028103027500000114
和
Figure FDA00028103027500000115
分别代表了移动设备和转发器的传输功率
Figure FDA00028103027500000116
和
Figure FDA00028103027500000117
表示从移动设备到转发器和从转发器到路边单元的信道增益，用N0表示高斯白噪声的单边功率谱密度，并得到两跳总信噪比：in,
Figure FDA00028103027500000114
and
Figure FDA00028103027500000115
represent the transmission power of the mobile device and the transponder, respectively
Figure FDA00028103027500000116
and
Figure FDA00028103027500000117
Denote the channel gain from the mobile device to the repeater and from the repeater to the roadside unit, denote the single-sided power spectral density of white Gaussian noise by N0 , and obtain the two-hop total signal-to-noise ratio:
Figure FDA00028103027500000118
Figure FDA00028103027500000118
 进而对于所传输的大小为
Figure FDA00028103027500000119
的数据包，当信道带宽为
Figure FDA00028103027500000120
时，传输时间
Figure FDA00028103027500000121
通过下式得到：And then for the transmitted size of
Figure FDA00028103027500000119
packets, when the channel bandwidth is
Figure FDA00028103027500000120
time, transmission time
Figure FDA00028103027500000121
It is obtained by the following formula:
Figure FDA00028103027500000123
Figure FDA00028103027500000123
 1.22对于在本地计算的任务，本地计算时间
Figure FDA00028103027500000122
由待计算任务对计算资源的需求
Figure FDA0002810302750000021
移动设备的本地计算能力
Figure FDA0002810302750000022
待计算任务对CPU资源的占有率
Figure FDA0002810302750000023
平均到达率为
Figure FDA0002810302750000024
和任务量分配比率
Figure FDA0002810302750000025
导出：1.22 For tasks that are computed locally, the local computation time
Figure FDA00028103027500000122
Demand for computing resources by tasks to be computed
Figure FDA0002810302750000021
The local computing power of the mobile device
Figure FDA0002810302750000022
The occupancy rate of the CPU resource of the task to be calculated
Figure FDA0002810302750000023
average arrival rate
Figure FDA0002810302750000024
and task allocation ratio
Figure FDA0002810302750000025
Export:
Figure FDA0002810302750000026
Figure FDA0002810302750000026
 1.23对于被分配到路边单元的服务器进行计算的任务，等待被服务器计算的来自于不同移动设备的任务量有总到达率
Figure FDA0002810302750000027
路边单元m具有c个等同的服务器，每个服务器的计算能力为
Figure FDA0002810302750000028
在M/M/c队列模型和Erlang公式的基础上，得到计算任务在路边单元m的平均处理时间：1.23 For tasks that are calculated by the server assigned to the roadside unit, the total arrival rate of the tasks from different mobile devices waiting to be calculated by the server
Figure FDA0002810302750000027
The roadside unit m has c equivalent servers, each with a computing power of
Figure FDA0002810302750000028
On the basis of the M/M/c queue model and Erlang formula, the average processing time of the computing task in the roadside unit m is obtained:
Figure FDA0002810302750000029
Figure FDA0002810302750000029
 其中in
Figure FDA00028103027500000210
Figure FDA00028103027500000210
Figure FDA00028103027500000211
Figure FDA00028103027500000211
 由于路边单元m的处理能力有限，待计算任务在必须队列中等待，然后被路边单元m处理并将结果发送给用户设备
Figure FDA00028103027500000212
因此每一个计算结果在路边单元m处的平均等待时间为：Due to the limited processing capacity of the roadside unit m, the task to be calculated waits in the necessary queue, and then is processed by the roadside unit m and the result is sent to the user equipment
Figure FDA00028103027500000212
Therefore, the average waiting time at roadside unit m for each calculation result is:
Figure FDA00028103027500000213
Figure FDA00028103027500000213
 其中，
Figure FDA00028103027500000214
为路边单元m的传输处理速度，由于计算结果的数据长度远小于计算任务，计算结果从路边单元m到用户设备
Figure FDA00028103027500000215
的时延忽略；当准备发送计算结果时，如果车辆
Figure FDA00028103027500000216
已经运动到路边单元m的覆盖范围之外了，计算结果将首先被发送到中心控制器，然后被转发至车辆
Figure FDA00028103027500000217
所在的路边单元m'；此过程中的传输延时
Figure FDA00028103027500000218
在中心控制器的平均等待时间
Figure FDA00028103027500000219
和在路边单元m'的等待时间
Figure FDA00028103027500000220
认为是常量，因此跨区的延时表达为：in,
Figure FDA00028103027500000214
is the transmission processing speed of the roadside unit m. Since the data length of the calculation result is much smaller than the calculation task, the calculation result is transmitted from the roadside unit m to the user equipment.
Figure FDA00028103027500000215
The delay is ignored; when preparing to send the calculation results, if the vehicle
Figure FDA00028103027500000216
has moved out of coverage of roadside unit m, the calculation result will first be sent to the central controller and then forwarded to the vehicle
Figure FDA00028103027500000217
The roadside unit m' where it is located; the transmission delay in this process
Figure FDA00028103027500000218
Average wait time at the central controller
Figure FDA00028103027500000219
and the waiting time at roadside unit m'
Figure FDA00028103027500000220
It is considered to be a constant, so the delay across regions is expressed as:
Figure FDA00028103027500000221
Figure FDA00028103027500000221
 1.24对整个移动边缘计算过程的执行时间
Figure FDA00028103027500000222
有1.24 Execution time of the entire mobile edge computing process
Figure FDA00028103027500000222
Have
Figure FDA00028103027500000223
Figure FDA00028103027500000223
 用户设备
Figure FDA00028103027500000224
的能量损耗应包括本地计算的能量消耗和传输数据的能量消耗；定义
Figure FDA00028103027500000225
为本地计算功率，它取决于CPU的固有特性和工作负载的复杂性，在任务执行期间被视为常量；User equipment
Figure FDA00028103027500000224
The energy consumption should include the energy consumption of local computing and the energy consumption of transmitting data; define
Figure FDA00028103027500000225
for the local computing power, which depends on the inherent characteristics of the CPU and the complexity of the workload, and is treated as constant during the execution of the task; 通过下得到用户设备
Figure FDA0002810302750000031
的本地计算能量消耗：Get user equipment by
Figure FDA0002810302750000031
The local computing energy consumption of:
Figure FDA0002810302750000032
Figure FDA0002810302750000032
 通过下式得到用户设备
Figure FDA0002810302750000033
向车内转发器发送数据的能量损耗：Get the user equipment by the following formula
Figure FDA0002810302750000033
Energy consumption for sending data to the in-vehicle transponder:
Figure FDA0002810302750000034
Figure FDA0002810302750000034
 通过下式得到用户设备
Figure FDA0002810302750000035
的总能量损耗：Get the user equipment by the following formula
Figure FDA0002810302750000035
The total energy loss of:
Figure FDA0002810302750000036
Figure FDA0002810302750000036
 能耗优化方案为基于ADMM的计算任务分配和功率控制方案，其目标为最小化路边单元m服务范围内mk辆车的整体能耗，定义优化变量集合
Figure FDA0002810302750000037
其中
Figure FDA0002810302750000038
则优化问题为：The energy consumption optimization scheme is a computing task allocation and power control scheme based on ADMM, and its goal is to minimize the overall energy consumption of m k vehicles within the service range of roadside unit m, and define a set of optimization variables.
Figure FDA0002810302750000037
in
Figure FDA0002810302750000038
Then the optimization problem is:
Figure FDA0002810302750000039
Figure FDA0002810302750000039
 s.t.s.t.
Figure FDA00028103027500000310
Figure FDA00028103027500000310
Figure FDA00028103027500000311
Figure FDA00028103027500000311
Figure FDA00028103027500000312
Figure FDA00028103027500000312
Figure FDA00028103027500000313
Figure FDA00028103027500000313
Figure FDA00028103027500000314
Figure FDA00028103027500000314
Figure FDA00028103027500000315
Figure FDA00028103027500000315
 C1和C2为限制了工作负载的到达率
Figure FDA00028103027500000316
和
Figure FDA00028103027500000317
分别不能超过用户设备
Figure FDA00028103027500000318
和路边单元m的处理速率,C3确保了传输功率不超过用户设备的最大传输功率,C4和C5分别为数据传输和任务计算过程的延迟限制,C6为任务分配比率
Figure FDA00028103027500000319
的边界限制； C1 and C2 limit the arrival rate of the workload
Figure FDA00028103027500000316
and
Figure FDA00028103027500000317
respectively cannot exceed the user equipment
Figure FDA00028103027500000318
and the processing rate of the roadside unit m , C3 ensures that the transmission power does not exceed the maximum transmission power of the user equipment, C4 and C5 are the delay limits of the data transmission and task calculation process, respectively, and C6 is the task allocation ratio
Figure FDA00028103027500000319
boundary limits; 在P1中，因为不同的用户设备
Figure FDA00028103027500000320
的任务分配变量是耦合的，因此优化目标是不可分离的；为了解决该问题，进一步包括以下步骤：In P1, because different user equipment
Figure FDA00028103027500000320
The task assignment variables of are coupled, so the optimization objective is inseparable; to solve this problem, the following steps are further included: 2.1引入最优资源分配策略的本地副本；使用一组新的变量来表示局部优化变量，定义
Figure FDA00028103027500000321
和
Figure FDA00028103027500000322
分别作为
Figure FDA00028103027500000323
和
Figure FDA00028103027500000324
的本地变量，则本地优化变量的集合被定义为
Figure FDA00028103027500000325
其中
Figure FDA00028103027500000326
2.1 Introduce a local copy of the optimal resource allocation strategy; use a new set of variables to represent local optimization variables, define
Figure FDA00028103027500000321
and
Figure FDA00028103027500000322
respectively as
Figure FDA00028103027500000323
and
Figure FDA00028103027500000324
, the set of local optimization variables is defined as
Figure FDA00028103027500000325
in
Figure FDA00028103027500000326
 则P1的次优问题表达为：Then the suboptimal problem of P1 is expressed as:
Figure FDA00028103027500000327
Figure FDA00028103027500000327
 s.t.s.t.
Figure FDA0002810302750000041
Figure FDA0002810302750000041
Figure FDA0002810302750000042
Figure FDA0002810302750000042
Figure FDA0002810302750000043
Figure FDA0002810302750000043
Figure FDA0002810302750000044
Figure FDA0002810302750000044
Figure FDA0002810302750000045
Figure FDA0002810302750000045
Figure FDA0002810302750000046
Figure FDA0002810302750000046
 2.2 P2通过引入局部变量使目标函数
Figure FDA0002810302750000047
可分离，将目标函数分解为mK个被并行解决的子问题，这些分散的联合优化问题被表达为：2.2 P2 makes the objective function by introducing local variables
Figure FDA0002810302750000047
Separable, decomposing the objective function into mK sub-problems that are solved in parallel, these decentralized joint optimization problems are expressed as:
Figure FDA0002810302750000048
Figure FDA0002810302750000048
Figure FDA0002810302750000049
Figure FDA0002810302750000049
 目标函数P3依然是一个非凸问题，将P3的分子和分母分别定义为：The objective function P3 is still a non-convex problem, and the numerator and denominator of P3 are defined as:
Figure FDA00028103027500000410
Figure FDA00028103027500000410
Figure FDA00028103027500000411
Figure FDA00028103027500000411
 并定义
Figure FDA00028103027500000412
作为P3的最优目标函数值：and define
Figure FDA00028103027500000412
As the optimal objective function value of P3:
Figure FDA00028103027500000413
Figure FDA00028103027500000413
 其中
Figure FDA00028103027500000414
和
Figure FDA00028103027500000415
分别代表了最优本地计算任务分配比率和功率控制策略；in
Figure FDA00028103027500000414
and
Figure FDA00028103027500000415
represent the optimal local computing task allocation ratio and power control strategy, respectively; 2.3根据非线性分式优化问题，获得最优目标值
Figure FDA00028103027500000416
的充分必要条件是：当且仅当方程2.3 According to the nonlinear fractional optimization problem, obtain the optimal target value
Figure FDA00028103027500000416
The necessary and sufficient conditions are: if and only if the equation
Figure FDA00028103027500000417
Figure FDA00028103027500000417
 成立，即通过解决下面的问题得到最优的本地优化变量
Figure FDA00028103027500000418
和
Figure FDA00028103027500000419
is established, that is, the optimal local optimization variables are obtained by solving the following problems
Figure FDA00028103027500000418
and
Figure FDA00028103027500000419
Figure FDA0002810302750000051
Figure FDA0002810302750000051
Figure FDA0002810302750000052
Figure FDA0002810302750000052
 2.4为每个用户设备定义本地变量集合
Figure FDA0002810302750000053
并定义函数：2.4 Define a local set of variables for each user device
Figure FDA0002810302750000053
and define the function:
Figure FDA0002810302750000054
Figure FDA0002810302750000054
 由此，关于P2的凸优化问题表达为：Thus, the convex optimization problem on P2 is expressed as:
Figure FDA0002810302750000055
Figure FDA0002810302750000055
Figure FDA0002810302750000056
Figure FDA0002810302750000056
 2.5定义关联于P5的最优变量集合
Figure FDA0002810302750000057
步骤2)迭代算法的每一次迭代过程中，下面的问题被解决：2.5 Define the optimal set of variables associated with P5
Figure FDA0002810302750000057
Step 2) During each iteration of the iterative algorithm, the following problems are solved:
Figure FDA0002810302750000058
Figure FDA0002810302750000058
Figure FDA0002810302750000059
Figure FDA0002810302750000059
 其中最优解
Figure FDA00028103027500000510
在前一步迭代中获得，当限制条件
Figure FDA00028103027500000511
被满足时
Figure FDA00028103027500000512
是所求优化问题P1的一组最优解；The optimal solution
Figure FDA00028103027500000510
obtained in the previous iteration, when the constraints
Figure FDA00028103027500000511
when satisfied
Figure FDA00028103027500000512
is a set of optimal solutions to the optimization problem P1; 对于迭代过程，定义对应于方程P6的拉格朗日乘子集合μm＝{μ1,...,μmk,...,μmK}，定义正常数ρ调整收敛速度，则P6的增广拉格朗日公式被表达为：For the iterative process, define the set of Lagrangian multipliers μ m ={μ 1 ,...,μ mk ,...,μ mK } corresponding to equation P6, and define a constant ρ to adjust the convergence rate, then the The augmented Lagrangian formula is expressed as:
Figure FDA00028103027500000513
Figure FDA00028103027500000513
 该迭代过程包含两层循环，外循环为非线性分式优化问题，用n来指示迭代次数；内循环为原始变量和对偶变量的更新，用t来指示迭代次数，The iterative process consists of two layers of loops. The outer loop is a nonlinear fractional optimization problem, and n is used to indicate the number of iterations; the inner loop is the update of the original variable and the dual variable, and t is used to indicate the number of iterations. 进一步包括：Further includes: 2.51对工作任务分配比率
Figure FDA00028103027500000514
传输功率
Figure FDA00028103027500000515
和最优解
Figure FDA00028103027500000516
初始化，设置终止条件ε；2.51 Assignment ratio to work tasks
Figure FDA00028103027500000514
Transmission power
Figure FDA00028103027500000515
and optimal solution
Figure FDA00028103027500000516
Initialize, set the termination condition ε; 2.52更新优化变量集合
Figure FDA00028103027500000517
给定第n次外循环的最优解
Figure FDA00028103027500000518
进而获得每个用户设备的传输功率
Figure FDA00028103027500000519
本地变量
Figure FDA00028103027500000520
和
Figure FDA00028103027500000521
的更新被分解为能够并行解决的mK个子问题；根据下式计算用户设备
Figure FDA00028103027500000522
在第t次内循环时的获得的最优任务分配比率
Figure FDA0002810302750000061
和传输功率
Figure FDA0002810302750000062
2.52 Update optimization variable set
Figure FDA00028103027500000517
The optimal solution given the nth outer loop
Figure FDA00028103027500000518
and then obtain the transmission power of each user equipment
Figure FDA00028103027500000519
local variable
Figure FDA00028103027500000520
and
Figure FDA00028103027500000521
The update of is decomposed into m K sub-problems that can be solved in parallel; the user equipment is calculated according to
Figure FDA00028103027500000522
The obtained optimal task allocation ratio at the t-th inner loop
Figure FDA0002810302750000061
and transmission power
Figure FDA0002810302750000062
Figure FDA0002810302750000063
Figure FDA0002810302750000063
 2.53更新
Figure FDA0002810302750000064
根据下式获得第t+1次内循环时的全局最优任务分配比率
Figure FDA0002810302750000065
2.53 Update
Figure FDA0002810302750000064
The global optimal task allocation ratio at the t+1th inner loop is obtained according to the following formula
Figure FDA0002810302750000065
Figure FDA0002810302750000066
Figure FDA0002810302750000066
 根据下式获得第t+1次内循环时的拉格朗日乘子
Figure FDA0002810302750000067
The Lagrange multiplier at the t+1th inner loop is obtained according to the following formula
Figure FDA0002810302750000067
Figure FDA0002810302750000068
Figure FDA0002810302750000068
 2.54更新最优解
Figure FDA0002810302750000069
在ADMM的初始变量和对偶变量的迭代过程中，当t趋于下确界时，满足目标函数收敛，残差收敛和对偶变量收敛条件；第n次迭代的内循环终止时得到
Figure FDA00028103027500000610
和
Figure FDA00028103027500000611
则第n+1次迭代的最优解
Figure FDA00028103027500000612
按下式得到：2.54 Update the optimal solution
Figure FDA0002810302750000069
In the iterative process of the initial variable and dual variable of ADMM, when t tends to the infimum, the objective function convergence, residual convergence and dual variable convergence conditions are satisfied; when the inner loop of the nth iteration terminates, the
Figure FDA00028103027500000610
and
Figure FDA00028103027500000611
Then the optimal solution of the n+1th iteration
Figure FDA00028103027500000612
Get it as follows:
Figure FDA00028103027500000613
Figure FDA00028103027500000613
 2.55循环终止；当第n次外层循环满足
Figure FDA00028103027500000614
时，通过下式获得最优任务分配比率
Figure FDA00028103027500000615
最优传输功率
Figure FDA00028103027500000616
和最优解
Figure FDA00028103027500000617
2.55 The loop terminates; when the nth outer loop satisfies
Figure FDA00028103027500000614
When , the optimal task allocation ratio is obtained by the following formula
Figure FDA00028103027500000615
optimal transmission power
Figure FDA00028103027500000616
and optimal solution
Figure FDA00028103027500000617
Figure FDA00028103027500000618
Figure FDA00028103027500000618
Figure FDA00028103027500000619
Figure FDA00028103027500000619
Figure FDA00028103027500000620
Figure FDA00028103027500000620
CN201810006519.9A 2018-01-04 2018-01-04 ADMM-based task allocation and power control method in Internet of vehicles Expired - Fee Related CN110012039B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810006519.9A CN110012039B (en) 2018-01-04 2018-01-04 ADMM-based task allocation and power control method in Internet of vehicles

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810006519.9A CN110012039B (en) 2018-01-04 2018-01-04 ADMM-based task allocation and power control method in Internet of vehicles

Publications (2)

Publication Number Publication Date
CN110012039A CN110012039A (en) 2019-07-12
CN110012039B true CN110012039B (en) 2021-01-05

Family

ID=67164159

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810006519.9A Expired - Fee Related CN110012039B (en) 2018-01-04 2018-01-04 ADMM-based task allocation and power control method in Internet of vehicles

Country Status (1)

Country Link
CN (1) CN110012039B (en)

Families Citing this family (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110717300B (en) * 2019-09-27 2022-10-21 云南电网有限责任公司 Edge calculation task allocation method for real-time online monitoring service of power internet of things
CN110856045B (en) * 2019-09-30 2021-12-07 咪咕视讯科技有限公司 Video processing method, electronic device, and storage medium
CN110650457B (en) * 2019-10-14 2020-07-14 河海大学 Joint optimization method for task unloading calculation cost and time delay in Internet of vehicles
CN110798858B (en) * 2019-11-07 2023-04-25 华北电力大学（保定） Distributed task unloading method based on cost efficiency
CN111090522B (en) * 2019-12-13 2023-07-28 南京邮电大学 A scheduling system and decision-making method for service deployment and migration in a mobile edge computing environment
CN111475301B (en) * 2020-04-09 2021-06-11 清华大学 Satellite resource allocation method and device and electronic equipment
CN111585615B (en) * 2020-04-17 2021-06-15 华北电力大学（保定） Direct current energy supply method
CN111935205B (en) * 2020-06-19 2022-08-26 东南大学 Distributed resource allocation method based on alternating direction multiplier method in fog computing network
CN112218351B (en) * 2020-10-27 2022-07-12 中国联合网络通信集团有限公司 Data transmission method, device and system
CN113590708B (en) * 2021-06-17 2024-02-20 贝壳找房(北京)科技有限公司 Adaptive delay consuming method, program product and storage medium
CN114264220B (en) * 2021-12-23 2022-11-22 湖南大学 A method for accurate sensing and detection of relative displacement of mobile devices
CN115941479B (en) * 2022-12-26 2024-12-27 北京城建智控科技股份有限公司 Collaborative task unloading method based on alternate direction multiplier method in intelligent video monitoring system

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105501216A (en) * 2016-01-25 2016-04-20 合肥工业大学 Internet of vehicles based hierarchical energy management control method for hybrid vehicle
CN106304290A (en) * 2016-08-12 2017-01-04 辛建芳 Internet of Things cooperative node Poewr control method based on N strategy
CN106502098A (en) * 2016-11-19 2017-03-15 合肥工业大学 A kind of optimum speed closed loop fast prediction control method and device based on car networking
CN106936892A (en) * 2017-01-09 2017-07-07 北京邮电大学 A kind of self-organizing cloud multi-to-multi computation migration method and system
CN106972898A (en) * 2017-03-15 2017-07-21 北京大学 Car networking data transmission scheduling method based on channel estimating

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10321409B2 (en) * 2013-10-28 2019-06-11 Huawei Technologies Co., Ltd. System and method for joint power allocation and routing for software defined networks

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105501216A (en) * 2016-01-25 2016-04-20 合肥工业大学 Internet of vehicles based hierarchical energy management control method for hybrid vehicle
CN106304290A (en) * 2016-08-12 2017-01-04 辛建芳 Internet of Things cooperative node Poewr control method based on N strategy
CN106502098A (en) * 2016-11-19 2017-03-15 合肥工业大学 A kind of optimum speed closed loop fast prediction control method and device based on car networking
CN106936892A (en) * 2017-01-09 2017-07-07 北京邮电大学 A kind of self-organizing cloud multi-to-multi computation migration method and system
CN106972898A (en) * 2017-03-15 2017-07-21 北京大学 Car networking data transmission scheduling method based on channel estimating

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Energy Efficient Optimization for Computation;Zheng Chang, Zhenyu Zhou, Tapani Ristaniemi, and Zhisheng Niu;《2017 IEEE Global Communications Conference》;20171208;第1-5页 *
Mobile Edge Computing Potential in;Tarik Taleb, Sunny Dutta, Adlen Ksentini等;《Enabling Mobile and Wireless Technologies for Smart Cities》;20170331;第38-43页 *

Also Published As

Publication number Publication date
CN110012039A (en) 2019-07-12

Similar Documents

Publication Publication Date Title
CN110012039B (en) ADMM-based task allocation and power control method in Internet of vehicles
CN109951821B (en) A task offloading scheme for minimizing vehicle energy consumption based on mobile edge computing
CN112351503B (en) Task prediction-based multi-unmanned aerial vehicle auxiliary edge computing resource allocation method
CN113286329B (en) Communication and computing resource joint optimization method based on mobile edge computing
CN113810233B (en) A Distributed Computing Offloading Method Based on Computational Network Collaboration in Random Networks
CN109951869B (en) Internet of vehicles resource allocation method based on cloud and mist mixed calculation
Zhang et al. User scheduling and task offloading in multi-tier computing 6G vehicular network
CN111010684B (en) A method for resource allocation of Internet of Vehicles based on MEC cache service
Zhou et al. Energy-efficient workload offloading and power control in vehicular edge computing
CN107708135A (en) A kind of resource allocation methods for being applied to mobile edge calculations scene
CN111930436A (en) Random task queuing and unloading optimization method based on edge calculation
CN107766135A (en) Method for allocating tasks based on population and simulated annealing optimization in mobile cloudlet
CN113064665B (en) Multi-server computing unloading method based on Lyapunov optimization
Shu et al. Joint offloading strategy based on quantum particle swarm optimization for MEC-enabled vehicular networks
CN112689296B (en) Edge calculation and cache method and system in heterogeneous IoT network
CN112650581A (en) Cloud-side cooperative task scheduling method for intelligent building
CN114205353B (en) A Computational Offloading Method Based on Hybrid Action Space Reinforcement Learning Algorithm
CN116233926B (en) Task unloading and service cache joint optimization method based on mobile edge calculation
CN111130911A (en) Calculation unloading method based on mobile edge calculation
Li et al. Distributed task offloading strategy to low load base stations in mobile edge computing environment
CN117715117B (en) Task unloading method for multi-UAV-RIS auxiliary mobile edge calculation
CN115002799A (en) A task offloading and resource allocation method for industrial hybrid networks
CN115022332A (en) A dynamic service placement method based on deep reinforcement learning in edge computing
CN116405493A (en) A MOGWO strategy-based edge cloud collaborative task offloading method
CN113573363A (en) MEC calculation unloading and resource allocation method based on deep reinforcement learning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20210105

CF01 Termination of patent right due to non-payment of annual fee