CN116405979A - Millimeter wave mobile edge computing networking resource allocation method - Google Patents

Millimeter wave mobile edge computing networking resource allocation method Download PDF

Info

Publication number
CN116405979A
CN116405979A CN202310237010.6A CN202310237010A CN116405979A CN 116405979 A CN116405979 A CN 116405979A CN 202310237010 A CN202310237010 A CN 202310237010A CN 116405979 A CN116405979 A CN 116405979A
Authority
CN
China
Prior art keywords
edge computing
users
server
optimization
resource allocation
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.)
Pending
Application number
CN202310237010.6A
Other languages
Chinese (zh)
Inventor
王宁
李泽
高梓涵
申凌峰
杨守义
屈凌波
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.)
Zhengzhou University
Original Assignee
Zhengzhou 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 Zhengzhou University filed Critical Zhengzhou University
Priority to CN202310237010.6A priority Critical patent/CN116405979A/en
Publication of CN116405979A publication Critical patent/CN116405979A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W28/00—Network traffic management; Network resource management
    • H04W28/02—Traffic management, e.g. flow control or congestion control
    • H04W28/08—Load balancing or load distribution
    • H04W28/09—Management thereof
    • H04W28/0925—Management thereof using policies
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04W—WIRELESS COMMUNICATION NETWORKS
    • H04W28/00—Network traffic management; Network resource management
    • H04W28/02—Traffic management, e.g. flow control or congestion control
    • H04W28/08—Load balancing or load distribution
    • H04W28/09—Management thereof
    • H04W28/0958—Management thereof based on metrics or performance parameters
    • H04W28/0967—Quality of Service [QoS] parameters
    • H04W28/0975—Quality of Service [QoS] parameters for reducing delays
    • 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)
  • Quality & Reliability (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The invention provides a millimeter wave mobile edge computing networking resource allocation method, which considers user pairing and resource allocation schemes among users, edge cloud computing servers and a central cloud server in terms of system constitution, uses a Gership-saprolimus algorithm-based system for resource allocation optimization, pairing scheme optimization and task unloading optimization of the users, the edge computing servers and the central cloud server as a whole, flexibly allocates tasks with different computation complexity to a proper server for processing, and assists in OFDMA technology and millimeter wave communication, so that the servers can serve more users simultaneously and effectively reduce communication delay, thereby reducing the energy time cost of the system to the maximum extent.

Description

Millimeter wave mobile edge computing networking resource allocation method
Technical Field
The invention belongs to the technical field of communication, and particularly relates to a millimeter wave mobile edge computing networking resource allocation method.
Background
With the development of fifth generation mobile communication technology and internet of things, wireless communication devices and wireless data traffic have grown geometrically; in particular, in recent years, with the advent of various computationally intensive and delay sensitive applications, such as intelligent transportation, smart cities, etc., it is important to provide a low-delay, ultra-reliable, high-robustness network.
In the prior art, the problem of computing resource allocation between a single-edge computing server and multiple users is mostly considered, and a central cloud server with stronger computing capacity is not considered; most consider a single-sided favorites list when dealing with pairing problems of users and edge computing servers; pairing changes of users and edge computing servers, which may occur during optimization of the offload policies, are not considered in processing the offload policies, thereby affecting the final optimization result.
To address the above challenges, the concept of mobile edge computing has been proposed to address the above issues; the mobile edge calculation aims to deploy a server with calculation capability to a network edge which is closer to terminal equipment, such as a road side base station and an unmanned aerial vehicle, and further assists in further reducing communication delay by millimeter wave technology, so that the millimeter wave mobile edge calculation networking resource allocation method has important significance for achieving the aims of reducing time delay, improving efficiency, saving power consumption and the like.
Disclosure of Invention
In the aspect of system constitution, the invention considers the user pairing and resource allocation scheme among users, the edge cloud computing servers and the central cloud server, uses the Gership-saprolimus algorithm to carry out resource allocation optimization, pairing scheme optimization and task unloading optimization on the users, the edge cloud computing servers and the central cloud server as a whole system, flexibly distributes tasks with different computation complexity to suitable servers for processing, and assists in using OFDMA technology and millimeter wave communication, so that the servers can serve more users simultaneously and effectively reduce communication delay, thereby reducing the energy time cost of the system to the maximum extent.
The purpose of the invention is realized in the following way: a millimeter wave mobile edge computing networking resource allocation method is characterized by comprising the following steps:
step 1, setting M users, N edge computing servers and 1 central cloud server to form a system model, wherein the users are in wireless connection with the edge computing servers, and the edge computing servers are in wired connection with the central cloud servers;
the computing tasks generated by the users are executed locally or by the collaboration cloud, and consumed resources are divided into time consumption and energy consumption;
with OFDM technology millimeter wave communication, the time consumption and energy consumption of data transmission are affected by load decisions and transmission power.
Step 2, constructing the minimization problem of the energy time cost ETC by jointly optimizing the resource allocation, the moving edge matching strategy and the load decision according to the system model in the step 1 in order to measure the system performance; the above problem of joint optimization is expressed as:
P：
Figure BDA0004122751120000022
Figure BDA0004122751120000023
Figure BDA0004122751120000024
Figure BDA0004122751120000025
Figure BDA0004122751120000026
Figure BDA0004122751120000027
Figure BDA0004122751120000028
step 3, solving the resource allocation problem, firstly optimizing the transmission power transmitted to the edge computing server n by the optimized user m under the condition that the application program needs to be unloaded to the specific edge computing server
Figure BDA0004122751120000029
And the edge calculation server n assigns the calculation frequency of user m +.>
Figure BDA00041227511200000210
So that each +.>
Figure BDA00041227511200000211
Minimizing; will->
Figure BDA00041227511200000212
And->
Figure BDA00041227511200000213
Carry-in
Figure BDA00041227511200000214
The minimization problem of (1) can be expressed as P1:
P1:
Figure BDA0004122751120000031
s.t.C 4 ,C 5
step 4, optimizing a pairing problem of a user and an edge computing server, wherein the mobile edge computing server matching policy optimization problem is an optimization algorithm based on a parameter b of specific resource allocation, wherein one edge computing server can serve a plurality of users, one user can only match one edge computing server, and the problem can be expressed as:
P4:
Figure BDA0004122751120000032
s.t.C 2 ,C 3 ,C 6 .
since a is known, the present optimization problem is an optimization algorithm for parameter b of a particular resource allocation;
step 5, optimizing the optimal unloading strategy optimization problem, and obtaining the optimal unloading decision by exhausting all possible unloading decisions, but leading to higher computational complexity; determining a suboptimal offloading decision to optimize offloading policy parameters a using a minimum offloading algorithm based on a guerre-saproli algorithm, the goal of the algorithm being to find an application that can be executed locally by judging whether ETC can be reduced, the algorithm will converge to a system where ETC cannot be reduced further; the problem can be expressed as:
P5:
Figure BDA0004122751120000033
s.t.C 1 ,C 6 .
step 6, combining the optimization results in the steps 1-5:
the above-mentioned solved resource allocation problem, optimization problem of mobile edge computing server matching strategy and optimal unloading strategy optimization problem respectively optimize the parameters set forth in problem P
Figure BDA0004122751120000034
A) m Combining them together results in a minimized system energy time cost.
Further, in step 3
Figure BDA0004122751120000035
And->
Figure BDA0004122751120000036
Are independent of each other, are->
Figure BDA0004122751120000037
And->
Figure BDA0004122751120000038
The optimization can be performed by a dichotomy and a polynomial analysis method respectively;
transmission frequency
Figure BDA0004122751120000039
The optimization of (c) can be expressed as solving the minimization problem of P2:
P2:
Figure BDA0004122751120000041
s.t.C 4 .
this is a quasi-convex problem, so that the solution can be performed by a dichotomy;
computing frequency of edge computing server
Figure BDA0004122751120000042
Can be translated into a minimization problem of solution P3:
P3:
Figure BDA0004122751120000043
s.t.C 5 .
combining dichotomy and quadratic polynomial analysis, the frequency can be calculated by optimizing the edge calculation server
Figure BDA0004122751120000049
The smallest P3 is obtained.
Further, the optimization algorithm of the parameter b in the step 4 is essentially based on the bilateral preference minimization problem of the guerre-saproli algorithm, and comprises the following specific steps:
1) Each edge computing server N e N is based on
Figure BDA0004122751120000044
m∈M off The user m who uninstalls the data is subjected to preference ordering according to ascending order, and a preference list is obtained>
Figure BDA0004122751120000045
2) The user mE M needing data unloading is according to the best
Figure BDA0004122751120000046
The edge computing server is strictly arranged in ascending order of preference degree to obtain a preference degree list +.>
Figure BDA0004122751120000047
3) Each user M e M off The most preferred edge computing server at present is proposed in each iteration, namely, the preferred list of the edge computing server is proposed, and the edge computing server is deleted from the preferred list after the edge computing server is proposed;
4) After all users put forward the current preferred edge computing servers, each edge computing server checks the suggestions received by the edge computing servers;
4.1 The edge computing server that did not get any suggestion remains unpaired;
4.2 Edge computing server that obtains preferences of multiple users based on its own list of preferences
Figure BDA0004122751120000048
The user accepted in the previous iteration is temporarily used as the most preferred user and accepted the preference suggestion until the edge computing server list reaches the maximum capacity, and the rest user preference is refused;
5) The above matching process will loop until all users are paired.
Further, the optimization problem of optimizing the optimal unloading strategy in the step 5 specifically comprises the following steps:
1) Assuming that all users need to perform edge calculation server calculation, i.e., a= {1,..once, 1}, the result in step 2 is input
Figure BDA0004122751120000051
Initial set M off =m, local computation set +.>
Figure BDA0004122751120000052
Δ k ＝ETC(a * )-ETC(a)；
2) Transferring the operation tasks of users 1-M, namely k=1, 2, & gt, M-1, M to local execution in sequence;
2.1 Calculating new when user k=1, 2..m-1, M's tasks are offloaded to the local
Figure BDA0004122751120000053
At this time M epsilon M off /{k},n∈N；
2.2 Updating at this time according to step 2)
Figure BDA0004122751120000054
2.3 According to the new
Figure BDA0004122751120000055
And->
Figure BDA0004122751120000056
Calculate new +.>
Figure BDA0004122751120000057
2.4 Delta calculated from the 3:2.3 result k For example delta k If not more than 0, the task of user k is performed locally and a is taken k =0, updating the unloading policy a * ；
3) Updating the offload policy parameter a=a according to the result of step 2) * 。
The invention has the beneficial effects that:
1. the invention adopts a multi-user multi-edge computing server and a resource allocation and unloading scheme of a central cloud server; the resource allocation scheme describes the energy time cost through the energy consumption and the calculation time in the local calculation and the collaborative cloud calculation respectively; the energy time cost minimization problem is then formulated by jointly optimizing the offloading decisions, communication and computational resources, and moving edge matching strategies under hard constraints.
2. In the invention, the edge cloud and the central cloud work in parallel, and the central cloud distributes complete computing resources to all users for parallel computing; therefore, the optimal task segmentation ratio can be obtained according to the calculation delay of the central cloud and the calculation delay of the edge cloud, and the subsequent operation complexity is greatly simplified.
3. The unloading scheme based on the Gerr-saprolide algorithm runs in two stages; firstly, optimizing the transmission power and the calculation resources of each possible user-edge calculation server matching strategy under the current unloading decision through a bilateral favorites list; then, based on the current offloading decision and the optimized resources, further optimizing the mobile edge matching policy by optimizing the offloading decision; if the unloading decision is updated in the optimization process, the updated unloading decision is re-brought into the user-edge computing server matching strategy so as to ensure that the time and energy cost of the system can be further optimized.
4. Because the line-of-sight communication is between the user and the edge computing server, the invention uses millimeter wave communication to further reduce delay.
Drawings
Fig. 1 is a schematic flow chart of a millimeter wave mobile edge computing networking resource allocation method provided by the invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to fall within the scope of the invention.
The invention provides a millimeter wave mobile edge computing networking resource allocation method, which splits a joint optimization problem into resource allocation optimization, mobile edge computing decision optimization and unloading decision optimization problems of a system to be solved respectively, so that the resource time cost of the system is minimized.
And (3) system model:
the invention centers on a high-power 5G base station with a coverage radius of R, a 5G core network adopts a CUPS architecture, the coverage area of the 5G core network comprises M users and N uniformly distributed ground Edge Computing Servers (MECs), all MEC servers are connected to a central cloud server (MCC) through the base station, each user can only be served by one MEC device, each MEC device can serve k users at most, and the number of load users of the system is as follows: m=nk.
The system optimization problem is that of a multi-user, multi-edge computing server and a single central cloud server.
A data processing model:
for computationally intensive applications for user m, the present invention is represented in two dimensions: l (L) m (bit) represents the size of the input data, including system settings, program code, and input parameters. C (C) m Representing the CPU processing density (cycles/bit) required to complete the application.
1. And (3) local calculation:
the time consumption and energy consumption calculated locally per user in the present system can be expressed as:
Figure BDA0004122751120000071
Figure BDA0004122751120000072
wherein the method comprises the steps of
Figure BDA0004122751120000073
Representing the local computing power of user m, κ l Indicating the effective switching capacity.
2. Collaborative cloud computing
The collaborative cloud computing combines an edge computing server and a central cloud server, and gives two reasonable assumptions for scene analysis in the present invention:
(1) Each computing application may be arbitrarily partitioned without regard to inherent content, which corresponds to video compression and speech recognition scenarios.
(2) The edge computing server and the central cloud server work in parallel, and the central cloud server distributes complete computing resources to all users for parallel computing. Each edge computing server determines whether the uninstalled application is executed by itself or by the edge computing server and the central cloud server in concert. For the latter, the edge computing server should determine the proportion of each application that needs to be executed on the central cloud server after full receipt
Figure BDA0004122751120000074
Since the edge cloud and the central cloud process the application in parallel, the total time consumption of the nth edge computing server to execute the user m application can be expressed as:
Figure BDA0004122751120000081
wherein the method comprises the steps of
Figure BDA0004122751120000082
Is the computational delay in the edge cloud n performing the user m computational task,
Figure BDA0004122751120000083
is the transmission delay of the backhaul link, where W is the communication capacity of the backhaul link.
Figure BDA0004122751120000084
Is the computation delay of MCC, where F cent Is the maximum operating frequency of the MCC.
Due to
Figure BDA0004122751120000085
Thus->
Figure BDA0004122751120000086
When monotonously increasing->
Figure BDA0004122751120000087
Also monotonically increase, so when
Figure BDA0004122751120000088
When (I)>
Figure BDA0004122751120000089
In addition->
Figure BDA00041227511200000810
Therefore->
Figure BDA00041227511200000811
At->
Figure BDA00041227511200000812
Upper following->
Figure BDA00041227511200000813
Increasing and decreasing. In order to minimize the overall delay, by calculation, when +.>
Figure BDA00041227511200000814
When (I)>
Figure BDA00041227511200000815
A minimum value can be taken where the optimal proportion of applications that need to execute on the MCC server after complete reception is:
Figure BDA00041227511200000816
wherein the method comprises the steps of
Figure BDA00041227511200000817
Is the computing frequency that edge server n assigns to user m. At the optimal split ratio, the overall delay and energy consumption of collaborative cloud computing can be expressed as:
Figure BDA00041227511200000818
Figure BDA00041227511200000819
wherein kappa is e Is the effective sample volume.
(II) communication model:
the computationally intensive applications of the present invention are loaded via millimeter wave wireless channels, here using a m E {0,1} to represent the offloading decision of user m, a m =0 means that user m selects local calculation, a m =1 denotes collaborative cloud computing. Thus, we can formulate the load decision profile as a= (a) 1 ,a 2 ,...,a m ) And (2) and
Figure BDA0004122751120000091
representing the total number of collaborative cloud computing users. Based on OFDMA techniques, the total bandwidth B is equally divided by K (a) users. Based on shannon's principle, the wireless transmission rate can be expressed as:
Figure BDA0004122751120000092
wherein the method comprises the steps of
Figure BDA0004122751120000093
Transmission power representing user m to edge cloud server n, +.>
Figure BDA0004122751120000094
Representing millimeter wave wireless channel gain between user m and edge cloud n, where the communication between the user and the edge computing server each takes into account a linear antenna array (ULA). B and N 0 Representing the total channel bandwidth and noise power spectral density, respectively. The transmission energy consumption and the time consumption are affected by a, and can be expressed as:
Figure BDA0004122751120000095
Figure BDA0004122751120000096
the above equation shows that the time consumption and the energy consumption of data transmission are affected by the load decision and the transmission power.
Construction joint optimization problem
To measure system performance, the present invention constructs an energy-time cost (ETC) minimization problem by jointly optimizing resource allocation, mobile edge matching policies, and load decisions.
The time energy cost in the present invention is defined as a weighted sum of energy consumption and time consumption to perform certain operations (such as transmission, local computation, and collaborative cloud computation), expressed as:
Figure BDA0004122751120000097
Figure BDA0004122751120000098
and->
Figure BDA0004122751120000099
Is a weight factor for measuring the preference of user m, < >>
Figure BDA00041227511200000910
The weighting factors can balance time consumption and energy consumption in the case of a large amount of loss. The ETC of the local computation, the transmission process, and the edge cloud computation can be expressed as:
Figure BDA00041227511200000911
and +.>
Figure BDA00041227511200000912
Thus, the collaborative cloud energy time cost for user m is:
Figure BDA00041227511200000913
wherein the method comprises the steps of
Figure BDA0004122751120000101
Representing a mobile edge matching policy, if the computing task of user m is performed by edge cloud server n
Figure BDA0004122751120000102
Other cases->
Figure BDA0004122751120000103
Considering the offloading decision, the ETC of user m can be expressed as:
Figure BDA0004122751120000104
the optimization objective of the invention is to optimize
Figure BDA0004122751120000105
b and a are the sum of ETCs for all users:
Figure BDA0004122751120000106
taking the minimum, the overall optimization problem can be expressed as:
P:
Figure BDA0004122751120000107
Figure BDA0004122751120000108
Figure BDA0004122751120000109
Figure BDA00041227511200001010
Figure BDA00041227511200001011
Figure BDA00041227511200001012
Figure BDA00041227511200001013
wherein M is off User set representing all data offloading, cap n Is the maximum computing capacity of the edge cloud server, C 1 The explanation can be executed locally or unloaded to the collaboration cloud for processing, C 2 ～C 6 Is a set M of pairs off And uninstall the constraints of the user.
Optimization problem solving
Since the formulation problem P is a mixed integer nonlinear programming problem that is difficult to solve directly, it is decomposed into three sub-problems: resource allocation, moving edge matching problems, and offloading decisions.
Resource allocation problem:
resource allocation is effective in situations where an application needs to be offloaded to a particular edge cloud, such as a m ＝1，
Figure BDA00041227511200001014
When (I)>
Figure BDA00041227511200001015
The minimization problem of (2) can be converted into by optimizing +.>
Figure BDA00041227511200001016
And->
Figure BDA00041227511200001017
Conversion to Each +.>
Figure BDA00041227511200001018
Minimization of problems, will->
Figure BDA00041227511200001019
And->
Figure BDA00041227511200001020
Carry in->
Figure BDA00041227511200001021
The minimization problem of (1) can be expressed as P1:
Figure BDA0004122751120000111
due to
Figure BDA0004122751120000112
And->
Figure BDA0004122751120000113
Are independent of each other, are->
Figure BDA0004122751120000114
And->
Figure BDA0004122751120000115
The optimization can be performed by a dichotomy and a polynomial analysis method, respectively.
For communication resources
Figure BDA0004122751120000116
The optimization of (c) can be expressed as solving the minimization problem of P2:
Figure BDA0004122751120000117
because (11) is a quasi-convex problem, the solution can be performed by a dichotomy.
Computing frequencies for edge clouds
Figure BDA0004122751120000118
The optimization problem of (2) can be translated into a minimization problem of solution P3:
Figure BDA0004122751120000119
combining dichotomy and quadratic polynomial analysis, the smallest P3 can be obtained by optimizing the edge cloud computing frequency.
(II) optimization problem of mobile edge cloud matching strategy:
the mobile edge matching algorithm provides a parameter b optimization algorithm based on specific resource allocation by combining
Figure BDA00041227511200001110
Taking equation (9), the moving edge matching problem can be expressed as P4:
Figure BDA00041227511200001111
p4 will be converted to a many-to-one matching problem by the mobile edge matching algorithm, where a and resource allocation schemes are known, the multiplex is matched to its preferred edge computing server, and there is bi-directional choice between the user and the edge computing server.
1) Preference list: the invention designs a bilateral preference list of an edge computing server and a user, which are respectively expressed as follows:
Figure BDA0004122751120000121
and->
Figure BDA0004122751120000122
Each edge computing server N e N is according to ∈ ->
Figure BDA0004122751120000123
m∈M off And (5) carrying out preference ordering on the users m unloading the data according to ascending order. Similarly, each user mε M that needs data offloading is according to the best
Figure BDA0004122751120000124
And (5) strictly arranging the ascending favorites of the edge cloud. Thus, if->
Figure BDA0004122751120000125
Edge cloud n pairs user m 1 Preference of user m is greater than that of user m 2 。
2) Moving edge matching algorithm: in each iteration, each user mε M off The edge computing server whose current most preference is presented in each iteration, i.e., the edge computing server's preferred list, and the edge computing server is removed from its preferred list after it is presented. After all users have proposed their current preferred edge computing servers, each edge computing server examines the suggestions it receives. Some edge computing servers may not get any suggestions, while other edge computing servers may get one or more suggestions. The edge computing servers that did not get any proposal remain unpaired. The edge computing server that obtained the preferences of the plurality of users temporarily treats the users that it accepted in the previous iteration as the most preferred users and accepts their preference suggestions until the list of edge computing servers reaches maximum capacity and rejects the remaining user preferences. The above matching process will loop until all users are paired.
3) Stability of pairing
If the matching set satisfies the following condition:
(1)：m i in the pairing set Pair (m i ) N is preferred in (a) j
(2)：n j In the pairing set Pair (n j ) M is preferred in i
Then a pair of paired users and edge computing server (m i ,n j ) For a Matching Set (MS) is a pair ofThe pair is occluded. If user m does not allow a blocking pair to appear, it is said to be stable. Because each user prefers to select their favorite edge computing server, the edge computing server will consider whether to replace paired users based on their preferences and capabilities, and thus there is no blocking pairing. For example, user 2 matches with edge computing server 2, if user 1 prefers edge computing server 2, and edge computing server 2 also prefers user 1 over user 2, edge computing server 2 will accept user 1 preferentially according to its capacity and delay accepting or rejecting user 2 so there is no blocking pairing. Thus, P4 may translate into a bilateral preference minimization problem based on the guerre-saproli algorithm.
(III) optimal offload strategy optimization problem:
minimum offloading algorithm
By exhausting all possible offloading decisions, an optimal offloading decision can be obtained, but this leads to a higher computational complexity, so that a minimum offloading algorithm based on the guerre-saproli algorithm is proposed in the present invention to determine a suboptimal offloading decision. Bringing equation (1 b) into equation (9), the constructed problem P5 can be expressed as:
Figure BDA0004122751120000131
in order to solve the problem P5, the following settings need to be satisfied:
(1) If and only if the system ETC (a) is smaller than the system ETC (a * ) When unloading decision a is better than decision a * When expressed as
Figure BDA0004122751120000137
(2) If and only if |M local Either |=m or there is no satisfaction
Figure BDA0004122751120000132
Local computing set M of (2) local The local computation set is not scalable when { k }.
(3) With tasks added to the userThe local processor, when processing, reduces the ETC of the system: pi (II) k ＝ETC(a * )-ETC(a)。
(4) If and only if a is the best offload decision when the number of users is fixed, a is called locally optimal, the decision of whether a is locally optimal requires comparison of the entire set of local calculations M local 。
The goal of the minimum offload algorithm is to find an application that can be executed locally by determining if ETC can be reduced, and the algorithm will converge to a system ETC that cannot be reduced further. In each iteration, it is assumed that each application executing on the collaborative cloud is computed locally, corresponding to a different M off I { k }, first calculate its latest
Figure BDA0004122751120000133
After which is M off Update edge matching policy +.>
Figure BDA0004122751120000134
And according to the new->
Figure BDA0004122751120000135
Calculate new +.>
Figure BDA0004122751120000136
Finally according to pi k ＝ETC(a * ) ETC (a) calculating n k . The negative number result in the iteration is pi k Results < 0 will be summarized to set pi l o cal Middle-position pi local ＝{Π k ' set pi l o cal Possessing the smallest pi k Will perform the computing task locally. The above process is circularly carried out until pi local Becomes an empty set.
The above-mentioned solved resource allocation problem, optimization problem of mobile edge cloud matching strategy and optimal unloading strategy optimization problem respectively optimize the parameters set forth in problem P
Figure BDA0004122751120000141
A) m Combining them together results in a minimized system energy time cost.

Claims (4)

1. A millimeter wave mobile edge computing networking resource allocation method is characterized by comprising the following steps:
step 1, setting M users, N edge computing servers and 1 central cloud server to form a system model, wherein the users are in wireless connection with the edge computing servers, and the edge computing servers are in wired connection with the central cloud servers;
the computing tasks generated by the users are executed locally or by the collaboration cloud, and consumed resources are divided into time consumption and energy consumption;
by adopting the OFDM technology millimeter wave communication, the time consumption and the energy consumption of data transmission are influenced by load decision and transmission power;
step 2, constructing the minimization problem of the energy time cost ETC by jointly optimizing the resource allocation, the moving edge matching strategy and the load decision according to the system model in the step 1 in order to measure the system performance; the above problem of joint optimization is expressed as:
P:
Figure FDA0004122751110000011
s.t.C 1 :
Figure FDA0004122751110000012
C 2 :
Figure FDA0004122751110000013
C 3 :
Figure FDA0004122751110000014
C 4 :
Figure FDA0004122751110000015
C 5 :
Figure FDA0004122751110000016
C 6 :
Figure FDA0004122751110000017
step 3, solving the resource allocation problem, firstly optimizing the transmission power transmitted to the edge computing server n by the optimized user m under the condition that the application program needs to be unloaded to the specific edge computing server
Figure FDA0004122751110000018
And the edge calculation server n assigns the calculation frequency of user m +.>
Figure FDA0004122751110000019
So that each +.>
Figure FDA00041227511100000110
Minimizing; will->
Figure FDA00041227511100000111
And->
Figure FDA00041227511100000112
Carry-in
Figure FDA00041227511100000113
The minimization problem of (1) can be expressed as P1:
P1:
Figure FDA0004122751110000021
s.t.C 4 ,C 5
step 4, optimizing a pairing problem of a user and an edge computing server, wherein the mobile edge computing server matching policy optimization problem is an optimization algorithm based on a parameter b of specific resource allocation, wherein one edge computing server can serve a plurality of users, one user can only match one edge computing server, and the problem can be expressed as:
P4:
Figure FDA0004122751110000022
s.t.C 2 ,C 3 ,C 6 .
since a is known, the present optimization problem is an optimization algorithm for parameter b of a particular resource allocation;
step 5, optimizing the optimal unloading strategy optimization problem, and obtaining the optimal unloading decision by exhausting all possible unloading decisions, but leading to higher computational complexity; determining a suboptimal offloading decision to optimize offloading policy parameters a using a minimum offloading algorithm based on a guerre-saproli algorithm, the goal of the algorithm being to find an application that can be executed locally by judging whether ETC can be reduced, the algorithm will converge to a system where ETC cannot be reduced further; the problem can be expressed as:
P5:
Figure FDA0004122751110000023
s.t.C 1 ,C 6 .
step 6, combining the optimization results in the steps 1-5:
the above-mentioned solved resource allocation problem, optimization problem of mobile edge computing server matching strategy and optimal unloading strategy optimization problem respectively optimize the parameters set forth in problem P
Figure FDA0004122751110000024
A) m Combining them together results in a minimized system energy time cost.
2. The millimeter wave mobile edge computing networking resource allocation method according to claim 1, wherein the method comprises the following steps: in step 3
Figure FDA0004122751110000031
And->
Figure FDA0004122751110000032
Are independent of each other, are->
Figure FDA0004122751110000033
And->
Figure FDA0004122751110000034
The optimization can be performed by a dichotomy and a polynomial analysis method respectively;
transmission frequency
Figure FDA0004122751110000035
The optimization of (c) can be expressed as solving the minimization problem of P2:
P2:
Figure FDA0004122751110000036
s.t.C 4 .
this is a quasi-convex problem, so that the solution can be performed by a dichotomy;
computing frequency of edge computing server
Figure FDA0004122751110000037
Can be translated into a minimization problem of solution P3:
P3:
Figure FDA0004122751110000038
s.t.C 5 .
combining dichotomy and quadratic polynomial analysis, the frequency can be calculated by optimizing the edge calculation server
Figure FDA0004122751110000039
The smallest P3 is obtained.
3. The millimeter wave mobile edge computing networking resource allocation method according to claim 1, wherein the method comprises the following steps: the optimization algorithm of the parameter b in the step 4 is essentially based on the bilateral preference minimization problem of the guerre-saproli algorithm, and comprises the following specific steps:
1) Each edge computing server N e N is based on
Figure FDA00041227511100000310
m∈M off The user m who uninstalls the data is subjected to preference ordering according to ascending order, and a preference list is obtained>
Figure FDA00041227511100000311
2) The user mE M needing data unloading is according to the best
Figure FDA00041227511100000312
n=1, 2., N performs a strict ascending preference arrangement on the edge calculation server, and obtains a preference list +.>
Figure FDA00041227511100000313
3) Each user M e M off The most preferred edge computing server at present is proposed in each iteration, namely, the preferred list of the edge computing server is proposed, and the edge computing server is deleted from the preferred list after the edge computing server is proposed;
4) After all users put forward the current preferred edge computing servers, each edge computing server checks the suggestions received by the edge computing servers;
4.1 The edge computing server that did not get any suggestion remains unpaired;
4.2 Edge computing server that obtains preferences of multiple users based on its own list of preferences
Figure FDA0004122751110000041
Forward itThe users accepted in one iteration are temporarily used as the most preferred users and accept preference suggestions until the edge computing server list reaches the maximum capacity, and the rest user preferences are refused;
5) The above matching process will loop until all users are paired.
4. The millimeter wave mobile edge computing networking resource allocation method according to claim 1, wherein the method comprises the following steps: the optimization problem of optimizing the optimal unloading strategy in the step 5 specifically comprises the following steps:
1) Assuming that all users need to perform edge calculation server calculation, i.e., a= {1,..once, 1}, the result in step 2 is input
Figure FDA0004122751110000042
Initial set M off =m, local computation set +.>
Figure FDA0004122751110000043
Δ k ＝ETC(a * )-ETC(a)；
2) Transferring the operation tasks of users 1-M, namely k=1, 2, & gt, M-1, M to local execution in sequence;
2.1 Calculating new when user k=1, 2..m-1, M's tasks are offloaded to the local
Figure FDA0004122751110000044
At this time M epsilon M off /{k},n∈N；
2.2 Updating at this time according to step 2)
Figure FDA0004122751110000045
2.3 According to the new
Figure FDA0004122751110000046
And->
Figure FDA0004122751110000047
Calculate new +.>
Figure FDA0004122751110000048
2.4 Delta calculated from the 3:2.3 result k For example delta k If not more than 0, the task of user k is performed locally and a is taken k =0, updating the unloading policy a * ；
3) Updating the offload policy parameter a=a according to the result of step 2) * 。
CN202310237010.6A 2023-03-13 2023-03-13 Millimeter wave mobile edge computing networking resource allocation method Pending CN116405979A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310237010.6A CN116405979A (en) 2023-03-13 2023-03-13 Millimeter wave mobile edge computing networking resource allocation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310237010.6A CN116405979A (en) 2023-03-13 2023-03-13 Millimeter wave mobile edge computing networking resource allocation method

Publications (1)

Publication Number Publication Date
CN116405979A true CN116405979A (en) 2023-07-07

Family

ID=87013344

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310237010.6A Pending CN116405979A (en) 2023-03-13 2023-03-13 Millimeter wave mobile edge computing networking resource allocation method

Country Status (1)

Country Link
CN (1) CN116405979A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116744261A (en) * 2023-08-16 2023-09-12 深圳市永达电子信息股份有限公司 Millimeter wave communication network and edge calculation fusion method

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116744261A (en) * 2023-08-16 2023-09-12 深圳市永达电子信息股份有限公司 Millimeter wave communication network and edge calculation fusion method
CN116744261B (en) * 2023-08-16 2023-11-28 深圳市永达电子信息股份有限公司 Millimeter wave communication network and edge calculation fusion method

Similar Documents

Publication Publication Date Title
CN109413724B (en) MEC-based task unloading and resource allocation scheme
CN111132191B (en) Method for unloading, caching and resource allocation of joint tasks of mobile edge computing server
CN112492626B (en) Method for unloading computing task of mobile user
CN111786839B (en) Calculation unloading method and system for energy efficiency optimization in vehicle-mounted edge calculation network
Liu et al. Economically optimal MS association for multimedia content delivery in cache-enabled heterogeneous cloud radio access networks
CN110087318B (en) Task unloading and resource allocation joint optimization method based on 5G mobile edge calculation
CN109951869B (en) Internet of vehicles resource allocation method based on cloud and mist mixed calculation
CN111182570B (en) User association and edge computing unloading method for improving utility of operator
CN112105062B (en) Mobile edge computing network energy consumption minimization strategy method under time-sensitive condition
CN111918311A (en) Vehicle networking task unloading and resource allocation method based on 5G mobile edge computing
CN109756912B (en) Multi-user multi-base station joint task unloading and resource allocation method
CN114051254B (en) Green cloud edge collaborative computing unloading method based on star-ground fusion network
CN107613556B (en) Full-duplex D2D interference management method based on power control
CN107708197B (en) high-energy-efficiency heterogeneous network user access and power control method
CN113727362B (en) Unloading strategy method of wireless power supply system based on deep reinforcement learning
CN110856259A (en) Resource allocation and offloading method for adaptive data block size in mobile edge computing environment
CN113064665A (en) Multi-server computing unloading method based on Lyapunov optimization
CN116405979A (en) Millimeter wave mobile edge computing networking resource allocation method
CN115801091A (en) Large-scale constellation network resource scheduling method for satellite-ground cooperative computing
Fan et al. Time-slotted task offloading and resource allocation for cloud-edge-end cooperative computing networks
CN111953730B (en) NOMA-based server cooperation edge computing method
CN107426775B (en) Distributed multi-user access method for high-energy-efficiency heterogeneous network
CN116761218A (en) Combined optimization method, system and storage medium in cloud edge end cooperative system
CN107333275B (en) robust power distribution method in uplink transmission femtocell heterogeneous network
CN114896039A (en) High-energy-efficiency edge computing unloading decision and resource allocation method

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