CN112084019B - Simulated annealing based calculation unloading and resource allocation method in heterogeneous MEC calculation platform - Google Patents

Simulated annealing based calculation unloading and resource allocation method in heterogeneous MEC calculation platform Download PDF

Info

Publication number
CN112084019B
CN112084019B CN202010806146.0A CN202010806146A CN112084019B CN 112084019 B CN112084019 B CN 112084019B CN 202010806146 A CN202010806146 A CN 202010806146A CN 112084019 B CN112084019 B CN 112084019B
Authority
CN
China
Prior art keywords
mec
task
unloading
uei
offload
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.)
Active
Application number
CN202010806146.0A
Other languages
Chinese (zh)
Other versions
CN112084019A (en
Inventor
刘楠
张绪琰
潘志文
尤肖虎
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.)
Southeast University
Original Assignee
Southeast 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 Southeast University filed Critical Southeast University
Priority to CN202010806146.0A priority Critical patent/CN112084019B/en
Publication of CN112084019A publication Critical patent/CN112084019A/en
Application granted granted Critical
Publication of CN112084019B publication Critical patent/CN112084019B/en
Active 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
    • G06F9/5011—Allocation of resources, e.g. of the central processing unit [CPU] to service a request the resources being hardware resources other than CPUs, Servers and Terminals
    • G06F9/5022—Mechanisms to release resources
    • 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]
    • G06F9/5061—Partitioning or combining of resources
    • G06F9/5072—Grid computing
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Medical Informatics (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Computing Systems (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Mobile Radio Communication Systems (AREA)

Abstract

The invention discloses a calculation unloading and resource allocation method in a heterogeneous MEC calculation platform based on simulated annealing, which is based on a space-ground combined MEC network, considers three unloading modes of local unloading, MBS-MEC server unloading and UAV-MEC server unloading in the network, constructs a user side energy consumption minimization problem model, considers the maximum service number of a heterogeneous server and the constraint of maximum calculation resources in the problem, considers the calculation power constraint of a UAV, discovers an unloading algorithm based on simulated annealing through computer simulation, and effectively reduces the total energy consumption of users in the network compared with the traditional algorithm.

Description

Simulated annealing based calculation unloading and resource allocation method in heterogeneous MEC calculation platform
Technical Field
The invention relates to the technical field of mobile edges, in particular to a calculation unloading and resource allocation method in a heterogeneous MEC calculation platform based on simulated annealing.
Background
Mobile Edge Computing (MEC), an extension of cloud Computing, is receiving increasing attention from both academic and industrial circles to provide IT cloud services for User Equipment (UE) at the Edge of a network. The MEC research currently deploys the MEC server mainly in a ground network, and the combination of Unmanned Aerial Vehicles (UAVs) and MECs in the sky has a significant advantage compared to the ground MEC network. Firstly, the UAV has strong maneuverability, so the deployment is flexible, and the UAV can be deployed in a burst hot spot area or a place which is difficult to be covered by a traditional MEC network; secondly, because the UAV and the ground mobile device are generally in a non-shielding state, a line-of-sight communication mode is adopted, and therefore the UAV and the ground mobile device have the characteristic of large-capacity transmission. How to jointly consider the problems of computing offloading and computing resource allocation in the ground MEC network and the UAV-MEC network to achieve the minimization of user energy consumption within the network is a little consideration.
Disclosure of Invention
The technical problem to be solved by the invention is to provide a method for computing offloading and resource allocation in a heterogeneous MEC computing platform based on simulated annealing, which can effectively reduce the total energy consumption of users in a network.
In order to solve the above technical problem, the present invention provides a method for computation offload and resource allocation in a heterogeneous MEC computing platform based on Simulated Annealing (SA), which includes the following steps:
(1) initializing parameters s, E, Z, a, t0,tth,snext＝s,Enext＝s,sbest＝s,EbestE, where s denotes the unloaded state, i.e. s ═ { ω1,..,ωi,...,ωNWhere ω isiIndicating the offload decision of the UE, the set of offload decisions of the UE is indicated as
Figure BDA0002629189600000011
If UEi decides to offload a task to timeslot t of UAV-MEC or MBS-MEC platform m, this decision action is denoted as pmt,
Figure BDA0002629189600000012
If UEi decides to compute a task locally, the decision action is denoted as p00；
(2) Performing steps (3) to (6) on Z, wherein Z represents the length of a Markov chain in a simulated annealing algorithm;
(3) after the current state s is determined, the offloading platforms of all UEs are also determined, and then the optimization problem of the computational resource allocation matrix F is considered;
(4) by sbestRepresenting the optimal state in the simulated annealing process by s in each iterationbestComparing with the current state s to avoid losing the current optimum state during Metropolis execution if Enext＜EbestUpdate sbest＝snext,Ebest＝Enext；
(5) Calculating an evaluation function difference Δ E-Enext；
(6) When the UE's offload status s changes to snextWhen the corresponding energy consumption changes from E to EnextThe unload state accepts theThe probability of change is expressed as
Figure BDA0002629189600000021
If E < EnextThen receive snextAnd according to snextReallocating the minimum computing resource; otherwise, a random number p is generatedrandE (0,1), if p (s → s)next)＜prandThen reject snextIf p (s → s)next)≥prandThen receive snext；
(7) Updating the current temperature t0←a*t0Wherein alpha (0 < alpha < 1) is a temperature-annealing constant, if a temperature-termination condition t is satisfied0≤tthThen output the current sbestAnd Ebest。
Preferably, in the step (3), the optimization problem of the calculation resource allocation matrix F is specifically considered as follows: if UEi decides to offload a task onto MEC platform m, take offload decision action pmtThen to ensure the user delay constraint, the minimum computing resource allocated by the MEC platform m for the task is expressed as
Figure BDA0002629189600000022
If less computing resources are allocated than
Figure BDA0002629189600000023
The task cannot be completed in one time slot; furthermore, if UEi decides to take an offload decision action p00If the UE local equipment completes the task within the specified time delay, the minimum local resource allocated to the UE task is
Figure BDA0002629189600000031
For each UE in the network, not all actions in its offload decision action set C can satisfy the constraints,for example, the MEC platform can allocate the computing resource less than the minimum computing resource required by the UE task
Figure BDA0002629189600000032
Or UAV-MEC or MBS-MEC reaches the maximum service number constraint in a certain time slot, or UAVj reaches the maximum value of the total calculation power of the task in the time slot t, or the task completion time from UEi to MEC platform m is too long, so that the task cannot be completed in one time slot t; therefore, for the offload decision set C of the UEi in a certain state s, the above-mentioned invalid action needs to be removed from C, and finally the valid action set C of the UEi is formedi(ii) a Active action set C from UEiiIn the random selection of action pmtUpdating the state s, then distributing the minimum computing resource for the UE according to the minimum computing resource distribution process, and finally obtaining the next unloading decision state snext(ii) a Finally, according to the total energy consumption of UE in the network, the total energy consumption is used as a certain unloading state s and snextEvaluation function of (1), denoted as E and Enext。
The invention has the beneficial effects that: the invention is based on a space-ground combined MEC network, three unloading modes of local unloading, MBS-MEC server unloading and UAV-MEC server unloading are considered in the network, a user side energy consumption minimization problem model is constructed, the maximum service number and the maximum computing resource constraint of a heterogeneous server are considered in the problem, the computing power constraint of the UAV is considered, an unloading algorithm based on simulated annealing is found through computer simulation, and compared with the traditional algorithm, the invention effectively reduces the total energy consumption of users in the network.
Drawings
Fig. 1 is a schematic diagram of a heterogeneous MEC platform network model according to the present invention.
FIG. 2 is a diagram illustrating comparison of performance of different algorithms for a small number of users.
FIG. 3 is a schematic diagram of analysis of the type of energy consumption for unloading under the SA algorithm of the present invention.
FIG. 4 is a diagram illustrating comparison of performance of different algorithms for a large number of users.
Detailed Description
System for controlling a power supplyThe model is shown in figure 1. Consider the representation of the UE set of the present network as
Figure BDA0002629189600000033
n represents
Figure BDA0002629189600000034
Middle element, UAV set denoted as
Figure BDA0002629189600000035
j represents
Figure BDA0002629189600000036
Middle element, micro base station set is
Figure BDA0002629189600000037
k represents
Figure BDA0002629189600000038
And (5) medium element. By deploying MEC servers at UAVs and MBS, the two MEC platforms have the capability of unloading UE tasks. Consider the joint representation of two off-loadable platforms as
Figure BDA0002629189600000039
m represents
Figure BDA0002629189600000041
A middle element, if m is less than or equal to J, representing an MEC server deployed on a UAVm in the UAV set, and if m is more than J, representing an MBS set
Figure BDA0002629189600000042
And an MEC server deployed beside the MBSk in the system. Furthermore we define a set of augmented offload platforms
Figure BDA0002629189600000043
m-0 indicates that the UE performs local computation on the task. Consider a network in which UEi has a compute-intensive task to perform, denoted as
Figure BDA0002629189600000044
Wherein FiRepresenting the computational resources required to complete the task in units of number of computational cycles, DiRepresenting the amount of upstream data in bits if the UE offloads the task to the MEC server.
The invention assumes that each UAV in the MEC system performs circular flight at a fixed altitude, and for UAVj, the flight period can be dispersed into a time slot set
Figure BDA0002629189600000045
The elements in the set are UAVs for a certain flight time slot t, each of which can provide computational offload services for UEs on the ground. The invention assumes that the position of the UAVj in each time slot changes little relative to the UE, and the change in position between the two can be ignored, so that the UAV position is considered to be unchanged within the time slot t. In addition, the service period of the MEC server located at MBSk can be discretized into a set of service slots due to the change of channel characteristics
Figure BDA0002629189600000046
The element in the set is the service period t of the MBS-MECk. Combining the UAV flight time slot set and the MBS-MEC service time slot set, and defining the service time slot sets under different platforms as
Figure BDA0002629189600000047
If m is less than or equal to J, representing the UAVm flight time slot set in the UAV set
Figure BDA0002629189600000048
If m > J, it represents the service time slot set of MBS-MEC (m-J)
Figure BDA0002629189600000049
In addition, an extended service slot set is defined
Figure BDA00026291896000000410
If t is equal to 0, the process is repeated,
Figure BDA00026291896000000411
it means that the UE is computing the task locally. Considering the offloading problem in MEC networks, we define the offloading decision factor of UEi as
Figure BDA00026291896000000412
If a isimtWhen m is 0 and t is 0, the UEi selects the local computing task; if a isimtWhen m is less than or equal to J, indicating that the UEi selects the unloading task in the UAvm flight time slot t; if a isimtAnd if m is more than or equal to J +1, indicating that the UE i selects the service time slot t of the MBS-MEC (m-J) for task unloading.
The invention considers a calculation task full-off model, i.e. for UE task IiIt can only be completely offloaded to a certain time slot of a certain UAV, a certain time slot of a certain MBS-MEC or the task is calculated locally, so there is a constraint
Figure BDA0002629189600000051
(1) UE task offloading to UAV-MEC model
The invention contemplates the representation of the three-dimensional coordinates of the UAVj within the flight time slot t as [ X ]jt,Yjt,Hjt]The three-dimensional coordinate of UEi on the ground is expressed as [ xi,yi,0]The three-dimensional coordinate of MBS-MECk on the ground is expressed as [ Xk,Yk,0]. The horizontal distance between UEi and UAVj in the flight time slot t can be expressed as
Figure BDA0002629189600000052
If the UEi decides to offload a task into the time slot t of the UAVj, the upstream communication rate may be expressed as follows
Figure BDA0002629189600000053
Wherein B isURepresenting the channel bandwidth, P, between the UAV and the UEi TWhich represents the transmission power of the ue i,
Figure BDA0002629189600000054
Figure BDA0002629189600000055
in the formula g0Is the channel gain per unit distance (m),
Figure BDA0002629189600000056
representing the noise power at the UAV. The data uplink transmission time when the UEi unloads the task to the UAVj flight time slot t at this time is expressed as follows
Figure BDA0002629189600000057
At this time, define
Figure BDA0002629189600000058
For the computational resources that the UAVj can provide for the UEi in a unit time in the time slot t, the computation time of the UEi task in the time slot t of the UAVj can be expressed as
Figure BDA0002629189600000059
In the MEC network, the task execution time mainly includes the uplink transmission time of the data to be calculated, the task calculation time and the result downlink transmission time. Considering that most application types have the characteristic of less result bit number, the result downlink transmission time is omitted from the total task execution time, and the task execution time can be represented as
Figure BDA00026291896000000510
In addition, the UE energy consumption of the UEi task when calculated in the time slot t of the UAVj is mainly UE uplink data transmission energy consumption, which can be expressed as UE uplink data transmission energy consumption
Figure BDA0002629189600000061
The calculated power of all UE tasks in time slot t of UAVj can be expressed as
Figure BDA0002629189600000062
Wherein s isjIs the effective transition capacitance associated with the server chip architecture, w-3, considering that UAVs have limited computational power, this constraint can be expressed as follows
Figure BDA0002629189600000063
Wherein P isURepresents the maximum computational power that the UAV can withstand per time slot, and furthermore, considering the limited service capability of the UAV-MEC server, the constraint of the total number of service tasks in a certain time slot is expressed as
Figure BDA0002629189600000064
Wherein A isUIs the UAVj maximum service number constraint, however since MEC servers have limited computational resources and UAVs have less computational resource carrying capacity, the UAV-MBS computational resource constraint can be expressed as
Figure BDA0002629189600000065
Wherein
Figure BDA0002629189600000066
Is the maximum computational resource that the UAVj can provide in a slot unit of time.
(2) UE task offloading to MBS-MEC model
The horizontal distance between UEi and MBS-MECk can be expressed as
Figure BDA0002629189600000067
According to the Shannon formula, the uplink rate when the UEi task unloads the MBS-MECk can be expressed as
Figure BDA0002629189600000068
Wherein B isMIs the channel bandwidth between the UE and the MBS,
Figure BDA0002629189600000071
is the channel gain of UEi and MBSk in the service time slot t, where ζ represents the path loss factor, siktIndicating shadowing fading of UEi and MBSk at time slot t,
Figure BDA0002629189600000072
representing channel noise power at MBS
Similar to the UAV-MEC offload model, if the UE decides to offload a task in slot t of MBSk, the time of the uplink transmission of the task data at this time
Figure BDA0002629189600000073
And calculating time
Figure BDA0002629189600000074
Can be expressed as
Figure BDA0002629189600000075
Figure BDA0002629189600000076
At this time, the total execution time of the task in the MBS-MEC offload mode can be expressed as
Figure BDA0002629189600000077
In addition, the UE offloading energy consumption in a certain time slot, i.e. the uplink data transmission energy consumption, can be expressed as
Figure BDA0002629189600000078
Meanwhile, compared with the UAV-MEC server with smaller number of simultaneous services and less computing resources, the MBS-MEC has more simultaneous service capability and richer computing resources, and the two constraints can be expressed as
Figure BDA0002629189600000079
Figure BDA00026291896000000710
Wherein A isMIndicating the maximum number of simultaneous services of the MBSk,
Figure BDA00026291896000000711
indicating the maximum computational resources that MBSk can provide in a slot unit of time.
(3) Task local computation mode
If the UE decides to compute the task locally, the completion time of the task local computation may be expressed as
Figure BDA00026291896000000712
Wherein f isimtRepresenting the computing resources per unit time that the UE local device can allocate, as with the power computation on the MEC server, the UE computing power at the time of the local computation of the UEi task can be represented by ki(fimt)μAt this time UEi calculationEnergy consumption can be expressed as
Figure BDA0002629189600000081
(4) Task offloading and computing resource allocation problem
The energy consumption for the UE includes three modes: the energy consumption of UE transmitting to UAV, the energy consumption of UE transmitting to MBS-MEC and the calculation energy consumption of UE local unloading can be uniformly expressed as EimtAs shown below
Figure BDA0002629189600000082
For the UE task delay, if the UE selects local unloading, the delay is local calculation time; if the UE selects to offload the task to the UAV-MEC and the MBS-MEC, the calculation delay mainly comprises the uplink data transmission time and the calculation time of the task on the MEC platform, which is specifically expressed as follows
Figure BDA0002629189600000083
Then unifying the UAV certain time slot access number constraint and the MBS-MEC certain time slot access number constraint, which is expressed as follows
Figure BDA0002629189600000084
Then, the UAV-MEC and the MBS-MEC are used for computing resources distributed in the time slot t for the UEi task
Figure BDA0002629189600000085
And
Figure BDA0002629189600000086
is uniformly expressed as follows
Figure BDA0002629189600000087
Then unifying the maximum computing resources owned by UAV-MEC and MBS-MEC in a certain time slot as FmIs shown as follows
Figure BDA0002629189600000091
Finally defining UE unloading decision matrix
Figure BDA0002629189600000092
Computing resource allocation matrix
Figure BDA0002629189600000093
The UE energy consumption optimization of the MEC heterogeneous network based on the UAV and the MBS at the moment is expressed as follows
Figure BDA0002629189600000094
Where L represents the length of one slot. The invention considers the UE unloading problem and the calculation resource distribution problem in the MEC network at the same time, and considers that the elements in the UE unloading decision matrix A are integers of 0-1 and the elements in the calculation resource distribution matrix F are continuous variables, so the problem is the Mixed Integer Non-Linear Programming (MINLP) problem.
(5) Description of algorithms
The invention discusses the unloading algorithm based on simulated annealing, and the specific flow is as follows
Step1 initializes parameters s, E, Z, a, t0,tth,snext＝s,Enext＝s,sbest＝s,EbestE. Where s denotes the unloaded state of the invention, i.e. s ═ ω1,..,ωi,...,ωNWhere ω isiIndicating the offload decision of the UEi. In the present model, the set of offloading decisions for a UE can be expressed as
Figure BDA0002629189600000095
If UEi decides to offload a task to timeslot t of UAV-MEC or MBS-MEC platform m, we denote this decision action as pmt,
Figure BDA0002629189600000101
If UEi decides to compute a task locally, we denote this decision action as p00。
Step2 makes steps 3 through Step6 for Z1.., Z (Z denotes the markov chain length in the simulated annealing algorithm); step3 for UEi 1.., N, after the current state s is determined, the offload platforms of all UEs are also determined, at which time we consider the computational resource allocation matrix F optimization problem. If UEi decides to offload a task onto MEC platform m, take offload decision action pmtThen to guarantee the user delay constraint, the minimum computing resource allocated by the MEC platform m for the task can be expressed as
Figure BDA0002629189600000102
If less computing resources are allocated than
Figure BDA0002629189600000103
The task cannot be completed in one time slot; furthermore, if UEi decides to take an offload decision action p00If the UE local equipment completes the task within the specified time delay, the minimum local resource allocated to the UE task is
Figure BDA0002629189600000104
For each UE in the network, not all actions in the offload decision action set C can satisfy the constraint, e.g., the MEC platform can allocate less computing resources than the minimum computing resources required for the UE task
Figure BDA0002629189600000109
Or UAV-MEC orThe MBS-MEC reaches the maximum service number constraint in a certain time slot, or the UAVj reaches the maximum value of the total task calculation power in the time slot t, or the task completion time from the UEi to the MEC platform m is too long, so the task cannot be completed in one time slot t. Hence offload decision set for UEi in a certain state s
Figure BDA0002629189600000105
We need to get from
Figure BDA0002629189600000106
Removing the invalid action and finally forming the valid action set of UEi
Figure BDA0002629189600000107
We are working set of actions from UEi
Figure BDA0002629189600000108
In the random selection of action pmtUpdating the state s, then distributing the minimum computing resource for the UE according to the minimum computing resource distribution process, and finally obtaining the next unloading decision state snext. Finally, the total energy consumption of the UE in the network is taken as a certain unloading state s and s according to the objective function in the formula (29)nextEvaluation function of (1), denoted as E and Enext。
Step4 in the present invention, we use sbestShowing the optimal state in the simulated annealing process. During each iteration we use sbestCompared to the current state s to avoid losing the current optimal state during execution of Metropolis. If E isnext＜EbestUpdate sbest＝snext,Ebest＝Enext；
Step5 calculates the evaluation function difference Δ E-Enext；
Step6 when the UE's offload status s changes to snextWhen the corresponding energy consumption changes from E to Enext. The probability that the unloaded state will accept the change can be expressed as
Figure BDA0002629189600000111
If E < EnextThen receive snextAnd according to snextReallocating the minimum computing resource; otherwise, a random number p is generatedrandE (0,1), if p (s → s)next)＜prandThen reject snextIf p (s → s)next)≥prandThen receive snext。
Step7 updating the current temperature t0←a*t0Wherein α (0 < α < 1) is the annealing constant. If the temperature termination condition t is satisfied0≤tthThen output the current sbestAnd Ebest。
The invention researches and solves the unloading problem in the MEC network based on a simulated annealing algorithm, in the example, the coverage area of the MEC network of a user is considered to be a circular area with the radius of 200m, and UE is uniformly and randomly distributed in the network; the number K of MBS-MEC servers is 2, two UAVs exist in the network, the flying height of the UAV is 200m, the circle flying is carried out by the radius of 100m, and the coordinates of the flying circle centers of the two UAVs are [100,100,200 ]]And [ -100,200](ii) a The total computing resource of the MBS-MEC server is 30 GHz; the total computing resource of the UAV-MEC server is 10GHz, and the local computing resource of the UE device is 1 GHz. The user transmit power is 23dbm, the noise power spectral density at UAV is set to-169 dbm/Hz, and the noise power spectral density at MBS is set to-174 dbm/Hz. For bandwidth resources in the network, the invention considers that the average bandwidth from the UAV to the UE is 1MHz, and the average bandwidth from the MBS to the UE is 900 kHz. The maximum calculated power within one slot of the UAV is constrained to 10W. The UE to-be-calculated task is generated from a network service application set, and the communication data volume size is compliant to 100,1000]KB are evenly distributed, the size of the computational resource demand obeys [108,109]The cycels are uniformly distributed and the capacitance k is effectively convertedi＝10-27W/(cycles/s) ^ 3. The path loss factor between MBS and UE is taken as zeta 4, and the shadow fading standard deviation in time slot is
Figure BDA0002629189600000112
The maximum access number in the time slot of the UAV and MBS unit is respectively10 and 15, the slot length L is set to 1000 milliseconds.
The method comprises the following specific steps:
step1 initializes parameters s, E, Z, a, t0,tth,snext＝s,Enext＝s,sbest＝s,Ebest＝E；
Step2 for L ═ 1, …, L from Step3 to Step 6;
step3 forms a valid decision set C for UEi 1iAnd from CiIn select valid offload decision pmtAllocating the minimum computing resource for the UE to obtain the next state snextAnd the value of the objective function Enext；
Step4 if Enext＜EbestUpdate sbest＝snext,Ebest＝Enext；
Step5 calculates the difference value of the objective function Δ E-Enext；
Step6 according to metropolis criterion, if Δ E < 0, update s-snext,E＝EnextOtherwise, accept s with probability exp (- Δ E/T)nextAs a new solution;
step7 updating the current temperature t0←a*t0If the temperature termination condition t is satisfied0≤tthThen output the current sbestAnd Ebest。
In fig. 2 the invention compares the performance of four algorithms with a smaller number of UEs and tasks in the MEC network. RO represents a Random Offloading algorithm (RO), i.e., the UE randomly selects a platform and a timeslot to offload; RLUO denotes a UAV platform Reinforcement Learning offloading (RLUO), that is, UE either offloads the task into the UAV timeslot or selects local computation and learns the offloading strategy using the Reinforcement Learning algorithm; RLBO stands for UAV platform Reinforcement Learning offloading (RLBO), i.e., the UE either offloads tasks into the service slots of the MBS-MEC, or selects local computation, and learns offloading strategies using Reinforcement Learning algorithm. With the increase of the number of the UE, the SA algorithm achieves the energy consumption performance far better than the RO, RLUO and RLBO algorithms. This is because RLUO and RLBO can only provide UAV-MEC or MBS-UAV as a server platform for task offloading. Compared with a single MEC platform, the heterogeneous network combining the UAV-MEC and the MBS-UAV provided by the invention can provide more access resources and computing resources.
In fig. 3, the present invention analyzes the offloading type and offloading energy consumption at different UE numbers in the proposed SA algorithm result. With the increase of the number of UEs in the MEC network, the energy consumption of the UE to be offloaded to the UAV-MEC and the MBS-MEC is more and more, and the total energy consumption of the UE to select local computing is not increased, because the resources that the UAV-MEC and the MBS-MEC can provide are far richer than the computing resources that the UE device can provide compared with the local computing, and in addition, the good channel conditions of the UAV-MEC and the MBS-MEC cause the energy consumption of the UE for uplink data transmission to be generally better than the energy consumption of the local computing.
In fig. 4, the present invention compares the performance of four algorithms with large-scale UEs and tasks in the MEC network. With the increase of the number of the UE, the energy consumption of the four algorithms is increased; the RO performance is the worst, because the random UE decision does not optimize the overall energy consumption in the network; when the number of the UEs reaches 60, the RLUO cannot find a solution satisfying the optimization problem with the increase of the UEs because the computational power constraint of the UAV is not considered, and compared with the RLBO and the RLUO, the SA algorithm of the present invention still has a great performance advantage when having large-scale UEs in the network.

Claims (1)

1. The method for calculating unloading and resource allocation in the heterogeneous MEC calculation platform based on simulated annealing is characterized by comprising the following steps of:
(1) initializing parameters s, E, Z, a, t0,tth,snext＝s,Enext＝s,sbest＝s,EbestE, where s denotes the unloaded state, i.e. s ═ ω1,..,ωi,...,ωNWhere ω isiIndicating the offload decision of the UE, the set of offload decisions of the UE is indicated as
Figure FDA0003522834910000011
If UEi decides to offload a task to a time slot t of UAV-MEC or MBS-MEC platform m, this decision action is expressed as
Figure FDA0003522834910000012
If UEi decides to compute a task locally, the decision action is denoted as p00；
(2) Performing steps (3) to (6) on Z, wherein Z represents the length of a Markov chain in a simulated annealing algorithm;
(3) after the current state s is determined, the offloading platforms of all UEs are also determined, and then the optimization problem of the computational resource allocation matrix F is considered; the method specifically comprises the following steps: if UEi decides to offload a task onto MEC platform m, take offload decision action pmtThen to ensure the user delay constraint, the minimum computing resource allocated by the MEC platform m for the task is expressed as
Figure FDA0003522834910000013
If less computing resources are allocated than
Figure FDA0003522834910000014
The task cannot be completed in one time slot; furthermore, if UEi decides to take an offload decision action p00If the UE local equipment completes the task within the specified time delay, the minimum local resource allocated to the UE task is
Figure 1
Offload decision set for UEi in a certain state s
Figure FDA0003522834910000016
Is required to be driven from
Figure FDA0003522834910000017
Removing invalid actions and finally forming an effective action set of UEi
Figure FDA0003522834910000018
Active action set from UEi
Figure FDA0003522834910000019
In the random selection of action pmtUpdating the state s, then distributing the least computing resources to the UE according to the least computing resource distribution process, and finally obtaining the next unloading decision state snext(ii) a Finally, according to the total energy consumption of UE in the network, the total energy consumption is used as a certain unloading state s and snextEvaluation function of (1), denoted as E and Enext；
(4) By sbestRepresenting the optimal state in the simulated annealing process by s in each iterationbestComparing with the current state s to avoid losing the current optimum state during Metropolis execution if Enext＜EbestUpdate sbest＝snext,Ebest＝Enext；
(5) Calculating an evaluation function difference Δ E-Enext；
(6) When the UE's offload status s changes to snextWhen the corresponding energy consumption changes from E to EnextThen the probability that the unloaded state accepts the change is expressed as
Figure FDA0003522834910000021
If E < EnextThen receive snextAnd according to snextReallocating the least computing resources; otherwise, a random number p is generatedrandE (0,1), if p (s → s)next)＜prandThen reject snextIf p (s → s)next)≥prandThen receive snext；
(7) Updating the current temperature t0←a*t0Wherein alpha is a temperature-annealing constant, 0 < alpha < 1, if a temperature-termination condition t is satisfied0≤tthThen output the current sbestAnd Ebest。
CN202010806146.0A 2020-08-12 2020-08-12 Simulated annealing based calculation unloading and resource allocation method in heterogeneous MEC calculation platform Active CN112084019B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010806146.0A CN112084019B (en) 2020-08-12 2020-08-12 Simulated annealing based calculation unloading and resource allocation method in heterogeneous MEC calculation platform

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010806146.0A CN112084019B (en) 2020-08-12 2020-08-12 Simulated annealing based calculation unloading and resource allocation method in heterogeneous MEC calculation platform

Publications (2)

Publication Number Publication Date
CN112084019A CN112084019A (en) 2020-12-15
CN112084019B true CN112084019B (en) 2022-05-10

Family

ID=73727886

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010806146.0A Active CN112084019B (en) 2020-08-12 2020-08-12 Simulated annealing based calculation unloading and resource allocation method in heterogeneous MEC calculation platform

Country Status (1)

Country Link
CN (1) CN112084019B (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107766135A (en) * 2017-09-29 2018-03-06 东南大学 Method for allocating tasks based on population and simulated annealing optimization in mobile cloudlet
US10037231B1 (en) * 2017-06-07 2018-07-31 Hong Kong Applied Science and Technology Research Institute Company Limited Method and system for jointly determining computational offloading and content prefetching in a cellular communication system
CN110418416A (en) * 2019-07-26 2019-11-05 东南大学 Resource allocation methods based on multiple agent intensified learning in mobile edge calculations system
CN110429973A (en) * 2019-08-05 2019-11-08 西北工业大学 A kind of unmanned plane and ground base station cooperated computing unloads and optimization method
CN110430593A (en) * 2019-08-17 2019-11-08 胡洋 Method for unloading tasks of edge computing user
CN111526526A (en) * 2020-04-03 2020-08-11 东南大学 Task unloading method in mobile edge calculation based on service mashup

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109814951B (en) * 2019-01-22 2021-09-28 南京邮电大学 Joint optimization method for task unloading and resource allocation in mobile edge computing network

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10037231B1 (en) * 2017-06-07 2018-07-31 Hong Kong Applied Science and Technology Research Institute Company Limited Method and system for jointly determining computational offloading and content prefetching in a cellular communication system
CN107766135A (en) * 2017-09-29 2018-03-06 东南大学 Method for allocating tasks based on population and simulated annealing optimization in mobile cloudlet
CN110418416A (en) * 2019-07-26 2019-11-05 东南大学 Resource allocation methods based on multiple agent intensified learning in mobile edge calculations system
CN110429973A (en) * 2019-08-05 2019-11-08 西北工业大学 A kind of unmanned plane and ground base station cooperated computing unloads and optimization method
CN110430593A (en) * 2019-08-17 2019-11-08 胡洋 Method for unloading tasks of edge computing user
CN111526526A (en) * 2020-04-03 2020-08-11 东南大学 Task unloading method in mobile edge calculation based on service mashup

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Joint Computation Offloading and Data Caching with Delay Optimization in Mobile-Edge Computing Systems;Haixia Wang等;《IEEE》;20171211;全文 *
MEC 中基于改进遗传模拟退火算法的虚拟网络功能部署策略;陈卓 等;《通信学报》;20200414;全文 *

Also Published As

Publication number Publication date
CN112084019A (en) 2020-12-15

Similar Documents

Publication Publication Date Title
CN111586720B (en) Task unloading and resource allocation combined optimization method in multi-cell scene
CN111405568B (en) Computing unloading and resource allocation method and device based on Q learning
CN112737837B (en) Method for allocating bandwidth resources of unmanned aerial vehicle cluster under high dynamic network topology
CN111970709B (en) Unmanned aerial vehicle relay deployment method and system based on particle swarm optimization algorithm
CN113377533A (en) Dynamic computation unloading and server deployment method in unmanned aerial vehicle assisted mobile edge computation
CN113359480A (en) Multi-unmanned aerial vehicle and user cooperative communication optimization method based on MAPPO algorithm
Gan et al. Trajectory optimization and computing offloading strategy in UAV-assisted MEC system
Yang et al. UAV-assisted 5G/6G networks: Joint scheduling and resource allocation based on asynchronous reinforcement learning
CN114363803B (en) Energy-saving multitasking distribution method and system for mobile edge computing network
CN114866979A (en) Unmanned aerial vehicle-assisted edge calculation time delay minimization method
CN114153515B (en) Highway internet of vehicles task unloading algorithm based on 5G millimeter wave communication
CN112084019B (en) Simulated annealing based calculation unloading and resource allocation method in heterogeneous MEC calculation platform
CN110602718B (en) Heterogeneous cellular network power distribution method and system based on alternative direction multiplier method
CN116723548A (en) Unmanned aerial vehicle auxiliary calculation unloading method based on deep reinforcement learning
CN116528250A (en) Unmanned aerial vehicle auxiliary MEC resource optimization method based on NOMA
Lee et al. Multi-Agent Reinforcement Learning in Controlling Offloading Ratio and Trajectory for Multi-UAV Mobile Edge Computing
CN115664486A (en) Energy efficiency optimization method for wireless energy supply in RIS (RIS) assisted UAV (unmanned aerial vehicle) edge computing system
Wu et al. ITS: Improved Tabu Search Algorithm for Path Planning in UAV-Assisted Edge Computing Systems
Liu et al. Joint optimization of resource scheduling and mobility for UAV-assisted vehicle platoons
Lee et al. MUSK-DQN: Multi-UBS selective-K deep Q-network for maximizing energy-efficiency
Cui et al. UAV-enabled MEC system: Locations and coverage optimization
CN117715117B (en) Task unloading method for multi-UAV-RIS auxiliary mobile edge calculation
CN117119534B (en) Unmanned aerial vehicle dynamic alliance-assisted unloading strategy and resource optimization method
Li et al. Game-Based Task Offloading and Resource Allocation in Multi-UAV-Assisted MEC Networks
CN118042529A (en) Method for assisting user edge calculation and wireless charging based on intelligent reflecting surface carried by unmanned aerial vehicle

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