CN113518330B - Multi-user computing unloading resource optimization decision method based on D2D communication - Google Patents
Multi-user computing unloading resource optimization decision method based on D2D communication Download PDFInfo
- Publication number
- CN113518330B CN113518330B CN202110762459.5A CN202110762459A CN113518330B CN 113518330 B CN113518330 B CN 113518330B CN 202110762459 A CN202110762459 A CN 202110762459A CN 113518330 B CN113518330 B CN 113518330B
- Authority
- CN
- China
- Prior art keywords
- user
- task
- communication
- computing
- unloading
- 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
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W4/00—Services specially adapted for wireless communication networks; Facilities therefor
- H04W4/70—Services for machine-to-machine communication [M2M] or machine type communication [MTC]
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04W—WIRELESS COMMUNICATION NETWORKS
- H04W72/00—Local resource management
- H04W72/50—Allocation or scheduling criteria for wireless resources
- H04W72/53—Allocation or scheduling criteria for wireless resources based on regulatory allocation policies
-
- 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
Abstract
The invention discloses a multi-user computing unloading resource optimization decision method based on D2D communication, and belongs to the technical field of mobile communication. The method comprises the following steps: 1. establishing a data communication model based on D2D communication; 2. respectively establishing a calculation overhead model including time and energy overhead in a task transmission stage and a task execution stage; 3. establishing a global optimization problem for minimizing the total cost calculated by all users of the whole system; 4. establishing a preference sequence of the bilateral user by taking the size of the calculated total overhead as a sequencing basis; 5. and based on the established preference sequence, obtaining a resource optimization decision for multi-user D2D calculation unloading by using a stable matching algorithm. The D2D communication-based calculation unloading method is beneficial to reducing unloading time delay and energy expenditure, the resource optimization decision of calculation unloading is obtained by using the stable matching algorithm, the total calculation expenditure of the system can be effectively reduced compared with a random matching method, and the performance which is very close to the optimal exhaustive search method can be obtained with lower calculation complexity.
Description
Technical Field
The invention relates to the technical field of mobile communication, in particular to a resource optimization decision method for computation and unloading of multiple users through D2D communication, which is a resource optimization decision method capable of contributing to improving computing resource shortage of a mobile communication network and reducing computation and unloading time delay and energy consumption.
Background
With the development and commercialization of Fifth Generation mobile communication (5G), more and more mobile applications, such as internet of things, augmented reality, video streaming, etc., such new applications that not only require high-speed data transmission but also require delay sensitivity of powerful computing resources are rapidly becoming widespread. To be able to adequately support these applications with special demands on computational resources and latency, mobile Edge Computing (MEC) has come into play. MEC is a very promising emerging 5G service scheme, mainly utilizing wireless network edge facilities near users or a large number of customer premise equipment cooperation, instead of remote cloud computing to perform a large number of communication and computing services. Compared with cloud computing, the MEC can provide the capability equivalent to that of cloud computing at the network edge, namely the wireless access network end, so that a user can unload the computing task at a short distance, and the low-delay and flexible computing and communication extension service is realized.
For the offloading problem of user computing tasks, most methods focus on the MEC network resource management based on edge servers, i.e. users can offload computing tasks to the high-speed computing servers at the edge of the network under consideration of wireless communication and mobile computing criteria. Although the computing performance of the mobile network can be improved to some extent by using only the computation offload of the edge server in the MEC network, the limited computing resources of the edge server at the base station are not always sufficient to support all mobile devices in the coverage area. And due to the randomness of the distribution of users in the network, especially for users distributed at the edge of a cell, the distance between the edge server at the base station end and the user may be far, thereby resulting in long offloading delay. To remedy these deficiencies and challenges in MEC networks, it is a significant solution to incorporate Device-to-Device (D2D) into MEC networks that allow direct communication between terminals without going through a base station, and to utilize D2D communication to assist in offloading user computing tasks.
A large number of heterogeneous devices in a network, such as internet of things devices, smartphones, tablets, etc., can be leveraged to support collaborative computing task execution for multiple services with their diverse computing capabilities and multiplexing gains. A significant reduction in offload distance is a key that is often ignored when computing offload through inter-user collaboration. From a wireless communication perspective, a short transmission distance can achieve a higher data transmission rate with lower power consumption, thereby reducing offloading latency and energy consumption when tasks are offloaded over short distances between users. In consideration of the advantages of D2D offloading, offloading decision methods based on D2D communication assistance are concerned, but at present, most of the D2D-assisted computational offloading decisions have very high computational complexity of algorithms, and it is a single objective to optimize an objective to minimize a time delay, but energy consumption of D2D devices is also an important point to be considered, and some methods only utilize the concept of D2D offloading, and device-to-device communication still employs cellular communication.
Disclosure of Invention
In view of this, the present invention is to integrate D2D communication and D2D offloading into an MEC network to form a D2D-MEC network, so as to further reduce offloading latency and energy consumption of a user on the basis of minimizing the total overhead of system computation. The method comprises the steps of modeling the time delay and energy consumption overhead of a computing task in a D2D transmission and execution stage, taking the total overhead of system computing including time delay and energy consumption minimization as an optimization target, unloading a multi-user computing task and distributing multi-user computing resources, modeling as an integer planning problem of the total overhead of the system computing minimization, and then regarding the problem to be solved as a D2D pairing process, providing a multi-user computing unloading resource optimization decision method based on D2D communication by using a stable matching algorithm, iteratively solving an optimization distribution decision of the D2D unloading, realizing the minimization of the total overhead of the system computing with lower complexity, and improving user experience.
The object of the invention is achieved by a solution in which the multi-user communication system comprises d t A mobile user requiring D2D computation offload, using set D t ＝{1,2,...,i,...,d t Denotes with d r A neighboring mobile user with spare computing resources capable of providing D2D computing offload services, using set D r ＝{1,2,...,j,...,d r Denotes d is t Each mobile user has an independent and time-delay sensitive computing task and needs to request computing offloading service from neighboring mobile users, which is characterized in that the specific steps of D2D computing offloading are as follows:
1) Establishing a data communication model based on D2D communication to obtain a data rate which can be realized by the D2D communication;
2) Respectively establishing a calculation overhead model comprising time and energy overhead in a task transmission stage and a task execution stage;
3) Establishing a global optimization problem for minimizing the total cost required by the task unloading of all users of the whole system based on a time and energy overhead model;
4) Establishing a preference sequence of the bilateral user by taking the size of the calculated total overhead as a sequencing basis;
5) And based on the established preference sequence, obtaining a resource optimization decision for multi-user D2D calculation unloading by using a stable matching algorithm.
Further, in the step 1), the computation and the unloading among the users need to utilize D2D communication to transmit computation tasks, and the D2D communication is controlled by a cellular network to access a wireless channel by adopting an orthogonal frequency division multiple access mode; obtaining the user i belongs to D based on free space propagation path loss and Rayleigh fading t And user j ∈ D r Achievable data rate r for D2D communication between ij 。
Further, in the step 2), the computation overhead generated in the task transmission and task execution phases of computation and offloading among users includes time overhead and energy overhead; in the task transmission stage, the total calculation cost of the user i calculation task including time cost and energy cost is
Wherein the content of the first and second substances,
is input data B of the task to be calculated by the user i i The transmission time consumed for transmission to user j via D2D communication,
is the power P of user i i Will input data B i The energy consumed by the transmission to user j via D2D communication,
weights representing time overhead and energy overhead, respectively; in the task execution stage, the user i calculates the total calculation cost of the task including time cost and energy cost into
Wherein the content of the first and second substances,
is the computation time consumed by user j to compute the received computation task,
is the computational energy consumed by user j to compute the received computational task.
Further, in the step 3), in order to minimize the time delay in the whole task offloading process and minimize the energy consumption of the user, the user experience is improved, so as to minimize the total calculation overhead in the task offloading process of all the users in the whole system, that is, the total time overhead and energy overhead including the task transmission and task execution stages are used as the objective function, and a linear global optimization problem is established.
Further, in the step 4), in order to solve the established optimization problem with low complexity, the requested user set D is unloaded t And a set of users D of the computing offload service r The two sets are regarded as two sets participating in matching, the unloading process of the computing task can be regarded as bilateral matching of one user task matching one service user, and the final matching result is the optimization decision of multi-user computing unloading; the bilateral matching is carried out, firstly, a preference sequence of bilateral users is established according to different preferences of the users, and according to an optimization target, a request user and a service user establish the preference sequence by taking the size of the calculated total cost as a sequencing basis: the smaller the total cost, the higher the user's preference and the higher the ranking.
Further, in the step 5), after the preference sequence is established, a stable matching algorithm is used for performing bilateral matching, the requesting user initiates a matching request to the highest preference service user ranked first according to the preference sequence, the corresponding service user makes an acceptance or rejection selection according to the preference sequence of the corresponding service user, the algorithm is performed iteratively, the unmatched requesting user continues to make a new matching/task unloading request to the highest preference service user in the current preference sequence until the matching with the service user is successful or no service user can initiate a request in the preference sequence is available, the algorithm is terminated when no new matching request is available, and a final matching result θ is output, namely, a decision for optimizing the unloading resources by multi-user D2D computing is made.
Due to the adoption of the technical scheme, the invention at least has the following advantages:
the resource optimization decision of multi-user computing unloading based on D2D communication is realized by using a stable matching algorithm with lower computing complexity, and the total computing unloading cost of the whole system including time and energy consumption is minimized. According to the method, the D2D is merged into the MEC network, idle computing resources of adjacent users are utilized, computing task unloading is executed in a mode of requesting D2D unloading to the adjacent mobile users, and compared with the mode of unloading all computing tasks to a far-end MEC server, time delay and energy consumption of computing unloading are reduced, and computing unloading congestion caused by shortage of computing resources of the MEC server can be avoided. Compared with one-to-many, many-to-one or many-to-many D2D unloading, the method has the advantages that tasks of requesting users do not need to be divided after complex algorithms are executed, system processing complexity is low, and the condition that unloading delay and energy consumption are increased due to the fact that a plurality of tasks are queued at one service user to overload the service user can be avoided. Further, the method considers the one-to-one D2D calculation unloading process as bilateral matching of a user task matching with a service user, utilizes stable matching to perform optimized resource allocation decision, and can prove that the resource optimization decision obtained by the method is stable, weak pareto optimal and has lower complexity through theoretical analysis; compared with the traditional random matching method, the method can effectively reduce the calculation overhead of the system, and can obtain the performance which is very close to that of the optimal exhaustive search method with lower calculation complexity.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof.
Drawings
FIG. 1 is a block flow diagram of the present invention;
FIG. 2 is a schematic view of a system model according to an embodiment;
FIG. 3 is a block diagram of a flow chart based on a stable matching algorithm in an embodiment;
FIG. 4 is a graph of a comparative simulation of the example with two reference methods;
fig. 5 is a simulation diagram comparing the embodiment and MEC unloading method.
Detailed Description
The following description will further explain embodiments of the present invention by referring to the figures and examples.
In the present embodiment, in the context of a D2D-MEC network environment composed of a macro cell Base Station (BS) and a plurality of mobile users, referring to the system model shown in fig. 2, each mobile user in the model can not only establish a cellular communication link with the Base Station, but also establish a D2D link with other nearby mobile users in a D2D communication manner. The base station is provided with an MEC server to provide corresponding computing unloading service for unloading larger computing tasks, and smaller computing tasks or divided computing tasks of some mobile users can be directly provided with the computing unloading service by adjacent mobile users in a D2D unloading mode. Referring to FIG. 2, in the present embodiment, there is d in the network t A mobile user needing to request D2D computation offload and D r The neighboring mobile users with idle computing resources capable of providing D2D computing offload service respectively use the set D t ＝{1,2,...,i,...,d t And set D r ＝{1,2,...,j,...,d r Denotes wherein d is t Each mobile user has an independent and delay sensitive computing task that requires a computing offload service to be requested from neighboring mobile users. In order to avoid complex signaling interaction and consider that the task amount is small, the user i belongs to D in the embodiment t The computing task can be only unloaded to one adjacent user j E D r User j e D providing computing service r Will only allocate its computing resources to a user i e D requesting offloading t Providing services to it. Without loss of generality, referring to most research scenarios of mobile edge computing and mobile networks, in order to analyze feasibility and obtain practical results, the embodiment also adopts a semi-static computing unloading scenario, namely, an on-going-to-all scenarioUsually within one calculation offload time period (in the order of hundred milliseconds), the mobile user set D t And D r Remain unchanged and may change over two or more cycles.
D2D unloading among users needs to utilize a D2D link to transmit a calculation task and a calculation result, and D2D communication is controlled by a cellular network to access a wireless channel by adopting an orthogonal frequency division multiple access mode. Based on free space propagation path loss and Rayleigh fading, a user i belonging to D requesting to calculate unloading t And a user j ∈ D providing computation offload service r Received power P of inter-D2D link j Can be expressed as:
wherein, P i Represents the transmission power of user i; h is a total of ij Represents the channel gain between user i and user j; d ij Represents the distance between user i and user j; beta is a path fading index, and the value of beta is between 2 and 4 according to different communication environments and channel quality; h is 0 Are the rayleigh channel coefficients obeying a complex gaussian distribution CN (0, 1).
Further, the achievable data rate r of the D2D link between the user i and the user j is obtained ij Expressed as:
wherein W represents a bandwidth; n is a radical of 0 Representing additive white gaussian noise of the channel.
In this embodiment, a binary V is used i ＝{B i ,C i Represents the user i e D t Computing tasks requiring offloading, wherein B i A unit of kilobytes (Kb) indicating the size of the task data amount; c i Indicating the number of CPU cycles, in Megacycles, required to execute the task. A user's computing task is offloaded to an adjacent user for computation, typically through three stages. First, a task offload/transfer stageSegment, i.e. user i e D t The computing task needs to be transmitted to a neighboring user j e to D through D2D communication r (ii) a The next is the task execution phase, i.e. user j ∈ D r Executing a user i E D arranged for him t The computing task of (2); finally, result downloading stage, i.e. user j ∈ D r Returning the result of the calculation to the user i E D t . In these three phases of D2D offloading, the computational overhead resulting from offloading of the computational tasks all includes time overhead and energy overhead. However, since the size of the data amount of the final calculation result is usually very small compared to the input data amount, the download time and energy consumption of the calculation result are negligible in the present embodiment, and the calculation overhead of the task transmission and task execution stage is mainly analyzed.
Calculating an overhead model in a task transmission stage:
user i will calculate input data B of task i Transmitting to user j through D2D communication, and transmitting input data B based on data rate achievable by D2D link between user i and user j i Consumed transmission time
Can be calculated from the following formula:
in addition to the transmission time overhead, since users are usually energy limited, the energy consumed by users needs to be considered in order to extend the working time as much as possible to improve the user experience. User i with power P i Will input data B i The energy consumed for transmission to user j via D2D communication depends on the time of data transmission, so that input data B is transmitted i Consumed transmission energy
Can be calculated from the following formula:
integrating formula (3) and formula (4), task transmission stage, user i calculates task V i Computational overhead including time overhead and energy overhead
Comprises the following steps:
wherein the content of the first and second substances,
representing the weights of the time overhead and the energy overhead, respectively. Because different users may have different specific requirements on time and energy expenditure, the weight factors are increased to reflect different weight selections of different users, so that the flexible and practical use of the model is realized. This weight parameter may be dynamically changed depending on the type of program the user is executing or the user's requirements at different optimization cycles. For the convenience of modeling, the weights do not change within one optimization cycle in this embodiment.
Calculating a cost model in a task execution stage:
when the computing task of the user i is successfully transmitted to the user j, the user j replaces the user i to execute the related computing task, the computing power of different users may be different, and the computing power of the user j is set, that is, the CPU computing frequency is f j Then user j computes the received computing task V i Consumed computing time
Can be calculated from the following formula:
user j calculates the received calculation task V i Will consume self energyQuantity, set
Is the power that user j needs to consume to perform a computational task, where κ j 0 or more is an effective switched capacitor, set to k in this embodiment j ＝10 -11 ，ν j ≧ 1 is a normal number. Then user j computes the received computing task V i Consumed computing energy
Can be calculated from the following formula:
integrating equation (6) and equation (7), in task execution phase, user i calculates task V i Computational overhead including time overhead and energy overhead
Comprises the following steps:
in this embodiment, computing task offloading is performed between users in a D2D manner, so as to minimize the time delay in the whole task offloading process and minimize the energy consumption of the users, and improve the user experience, so as to minimize the computing overhead in the task offloading process of all the users in the whole system, that is, the computing overhead includes the total time overhead and energy overhead in the task transmission and task execution stages as objective functions, and establish the following linear global optimization problem:
wherein, the binary variable x = { x) of the constraint 1) ij Indicating between a user requesting task offloading and a user providing offloading serviceIn the case of pairing between x ij If =1, the computing task of user i is offloaded to user j for execution, and the computing resource of user j is allocated to user i, x ij If =0, it does not hold; constraint 2) ensures that a user requesting task offloading can only offload tasks to at most one user providing offloading services; constraint 3) ensures that a user providing an offload service can only allocate computing resources to at most one user requesting task offload. Since the value of the variable can only be a binary variable of 0 or 1, the established optimization problem is a special binary linear programming problem in the integer programming problem and is a non-deterministic polynomial problem. There are many possibilities for a solution that satisfies the problem conditions 1) -3), but there is only one optimal solution. Although the optimal solution can be found by exhaustively exhausting all possible solutions, the computational complexity is very high. Therefore, in this embodiment, a less complex, sub-optimal stable matching algorithm is used to solve the problem, which can achieve near-optimal performance.
Computing offload request user set D t Can only select a set D of users for the computation offload service r One user of (2), a set of computation offload service users D r Can only choose to compute the offload requesting user set D t The two user sets have exactly one-to-one relationship, which is very suitable for the structure of the matching theory. Matching theory is an effective way to model one-to-one, many-to-one, or many-to-many assignments between two independent sets of subjects. Set of users D who will offload requests t And computing a set of users D for the offload service r The two sets are regarded as two sets participating in matching, the unloading process of the computing task can be regarded as bilateral matching of one user task matching one service user, and the final matching result is the optimization decision of multi-user computing unloading.
For bilateral matching, a bilateral user preference sequence is established according to different preferences of users. When requesting user i e to D t When the system is matched with different service users, the communication distance between the system and the different service users is different, and the computing power of the different service users is also differentThen different pairings will result in different computational overheads. Similarly, when the service user j belongs to D r When pairing with different requesting users, different pairings may also result in different computational overheads. The optimization goal of this embodiment is to minimize the system computation overhead, so the requesting user i ∈ D t For service user j E D r And service user j ∈ D r For requesting user i e D t In the sum of the formulae (5) and (8)
Is measured as a utility function. That is, the requesting user and the service user establish the preference sequence by using the size of the calculated total cost as the sequencing basis: the smaller the total cost, the higher the user's preference and the higher the ranking.
Is calculated for each service user j e to D r Total overhead of
Then sorting the values in ascending order from small to large according to the total expense, and establishing a request user i belongs to D t J e to D for different service users r The preference sequence of (1). For the same reason, j ∈ D r Is computed with each requesting user i e D t Total overhead of
Then sorting the service users according to the ascending order of the numerical value of the total expense from small to large, and establishing the service user j belonging to the D r For different requesting users i e to D t The preference sequence of (1).
After the preference sequence is established, the requesting user initiates a matching request to the highest preference service user ranked first according to the preference sequence, and the corresponding service user makes a selection according to the preference sequence. In this embodiment, resource optimization of multi-user computing offloading is implemented by using stable matching, and a flow of a specific decision method is shown in fig. 3. First, the initialization systemAnd establishing a list of all unmatched user sets requesting computation uninstallation according to all the parameters of the users
Each requesting user i e D t And each service user j e D r To be provided with
And calculating preference values of each individual in the other side set for the utility function and arranging the preference values in ascending order to obtain respective preference sequences. Non-matching list
And under the condition that the preference sequence of a certain requesting user is not empty, initiating a task unloading/matching request to the service user with the highest preference according to the preference sequence of the certain requesting user. The service user who receives the task unloading request compares the request user who initiates the request with the current matching object according to the preference sequence, if the service user prefers the request user who initiates the request, the service user accepts that the request user who initiates the request rejects the current matching object, the request user who initiates the request is removed from the unmatched list, and the rejected matching object is added into the unmatched list; if the service user prefers the current matching object, the current matching is maintained, the request user initiating the request is refused, and the refused request user initiating the request removes the service user from the preference sequence of the service user and updates the preference sequence of the service user. The algorithm is iterated, and the unmatched requesting users continue to make new task unloading/matching requests to the service user with the highest preference in the current preference sequence until the matching with the service user is successful or no service user can initiate a request in the preference sequence. And when no new matching request exists, the algorithm is terminated, and a final matching result theta is output, namely the decision of the multi-user D2D calculation unloading resource optimization is made.
In the embodiment, when the preference sequence is established, each user in the bilateral set needs to calculate the preference value (calculation overhead) for each user in the other bilateral set, so the calculation for establishing the preference sequence is complexDegree of O (d) t d r ). The computational complexity of the ascending sort preference value to obtain the preference sequence is O (d) t d r log(d t d r )). In the process of matching, each requesting user can make d direction at most r Each service user initiates a match request, then d t A requesting user initiates d at most t ×d r Secondary request with computational complexity of O (d) t d r ). If an exhaustive search method is used to solve the established optimization problem, all possible pairs need to be enumerated, so that the total number of matching results (the number of exhaustive times) which can appear is d t ！×d r ! The computational complexity is O (d) t ！×d r | A ) (ii) a If the method solves the established optimization problem by adopting a special branch-and-bound method, namely a hidden enumeration method, the method can reduce the enumeration times by branch-and-bound method compared with an exhaustive search method, and the calculation complexity is
(ii) a If the random matching method is adopted to solve the established optimization problem, the algorithm complexity is increased linearly along with the increase of the number of the service users, and the calculation complexity is O (d) r ). As can be seen from the analysis and comparison of the complexity of the above different methods, the computational complexity of the embodiment using the stable matching algorithm is slightly higher than that of the random matching method, but is significantly lower than that of the exhaustive search method and the implicit enumeration method, and the advantage is more obvious when the number of participating individuals is more.
When performing a Matlab simulation experiment on the method described in this embodiment, an area with a radius of 100m, d, is set t A requesting user with independent computing task and d r The service users with idle computing resources are evenly distributed within the circle. And adopting orthogonal channel resources, and unloading the task of one requesting user to a corresponding service user in a D2D communication mode to execute according to the matched decision result. Setting the bandwidth W =0.5MHz, the maximum transmission distance D2D as 200m, and the transmitting power P of the user equipment i =23dBm, noise power N 0 = 174dBm/Hz, path fading index β =3, task V i Input data volume B i 10-100 Kb, task V i Number of required CPU cycles C i 1-10 Megacycles, user CPU computing power f j 1 to 3.1GHz, weight
Constant v j =1. Under the setting, the comparison result of the total calculation overhead of the method, the random matching method and the exhaustive search method according to the embodiment with the change of the number of the requested unloading users is shown in fig. 4, and the set number d of the service users at this time r And =4. It can be seen that the calculated total cost obtained by the method of the present embodiment is very close to the optimal exhaustive search method, and is much lower than that obtained by the random matching method. Although the exhaustive search can obtain the optimal performance, the calculation complexity is very high, and the method of the embodiment can obtain the performance similar to the optimal performance with lower complexity. Similarly, the comparison result of the overhead of the D2D offloading method and the MEC offloading method according to the present embodiment varying with the number of requested offloading users is shown in fig. 5, where the number D of service users set at this time is r Is equal to the number d of the requested users t . The MEC unloading means that all tasks of the requesting users are unloaded to an MEC server which is far away from 200-700 m for execution, and the CPU calculation frequency of the MEC server is 5GHz. Because the MEC server usually has ac power supply conditions, the energy consumption overhead when the MEC offloads to execute the computation task is not considered in this embodiment, and other overhead is calculated in the same way as the D2D offload. It can be seen that the total overhead, the total delay overhead and the total energy consumption overhead of the D2D offloading method described in this embodiment are all lower than those of the MEC offloading method, and especially, the gap increases as the number of users requesting offloading increases.
Finally, the above embodiments are only intended to illustrate the technical solutions of the present invention and not to limit the present invention, and although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions, and all of them should be covered by the claims of the present invention.
Claims (6)
1. Multi-user computing offload based on D2D communicationResource optimization decision method is carried out, and the multi-user communication system comprises a device d t A mobile user requiring D2D computation offload, using set D t ＝{1,2,...,i,...,d t Denotes with d r A neighboring mobile user with free computing resources capable of providing D2D computing offload services, using set D r ＝{1,2,...,j,...,d r Denotes d is t Each mobile user has an independent and time-delay sensitive computing task and needs to request computing offloading service from neighboring mobile users, which is characterized in that the specific steps of D2D computing offloading are as follows:
1) Establishing a data communication model based on D2D communication to obtain a data rate which can be realized by the D2D communication;
2) Respectively establishing a calculation overhead model comprising time and energy overhead in a task transmission stage and a task execution stage;
3) Establishing a global optimization problem for minimizing the total cost required by the task unloading of all users of the whole system based on a time and energy overhead model;
4) Establishing a preference sequence of the bilateral user by taking the size of the calculated total overhead as a sequencing basis;
5) And based on the established preference sequence, obtaining a resource optimization decision for multi-user D2D calculation unloading by using a stable matching algorithm.
2. The multi-user computing offload resource optimization decision method based on D2D communication of claim 1, wherein: in the step 1), the computation and unloading among users need to transmit computation tasks by using D2D communication, and the D2D communication is controlled by a cellular network to access a wireless channel by adopting an orthogonal frequency division multiple access mode; obtaining the user i belongs to D based on free space propagation path loss and Rayleigh fading t And user j ∈ D r Achievable data rate r for D2D communication between them ij 。
3. The multi-user computing offload resource optimization decision method based on D2D communication of claim 1, wherein: in the step 2), the meters generated in the task transmission and task execution stages of computation and unloading among usersThe calculation cost comprises time cost and energy cost; in the task transmission stage, the total calculation cost of the user i calculation task including time cost and energy cost is
Wherein, the first and the second end of the pipe are connected with each other,
is the input data B of the task to be calculated by the user i i The transmission time consumed for transmission to user j via D2D communication,
is the power P of user i i Will input data B i The energy consumed by the transmission to user j via D2D communication,
weights representing time overhead and energy overhead, respectively; in the task execution stage, the user i calculates the total calculation cost of the task including time cost and energy cost into
Wherein the content of the first and second substances,
is the computation time consumed by user j to compute the received computation task,
is the computational energy consumed by user j to compute the received computational task.
4. The multi-user computing offload resource optimization decision-making method based on D2D communication of claim 1, wherein: in the step 3), in order to minimize the time delay in the whole task unloading process and minimize the energy consumption of the user, the user experience is improved, so as to minimize the total calculation overhead in the task unloading process of all the users in the whole system, that is, the total time overhead and energy overhead in the task transmission and task execution stages are taken as objective functions, and a linear global optimization problem is established.
5. The multi-user computing offload resource optimization decision method based on D2D communication of claim 1, wherein: in the step 4), in order to solve the established optimization problem with low complexity, the user set D of the unloading request is collected t And a set of users D of the computing offload service r The two sets are regarded as two sets participating in matching, the unloading process of the computing task can be regarded as bilateral matching of one user task matching one service user, and the final matching result is the optimization decision of multi-user computing unloading; the method comprises the following steps of firstly establishing a preference sequence of a bilateral user according to different preferences of the user, and establishing the preference sequence by a requesting user and a service user according to an optimization target by taking the size of the calculated total cost as a sequencing basis: the smaller the total overhead, the higher the user's preference and the higher the ranking.
6. The multi-user computing offload resource optimization decision method based on D2D communication of claim 1, wherein: in the step 5), after the preference sequence is established, a stable matching algorithm is used for bilateral matching, the requesting user initiates a matching request to the highest-preference service user ranked first according to the preference sequence, the corresponding service user selects acceptance or rejection according to the preference sequence of the requesting user, the algorithm is iterated, the unmatched requesting user continues to issue a new matching/task unloading request to the highest-preference service user in the current preference sequence until the matching with the service user is successful or no service user can initiate the request in the preference sequence, the algorithm is terminated when no new matching request exists, and a final matching result theta is output, namely a decision for optimizing the unloading resources by multi-user D2D computing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110762459.5A CN113518330B (en) | 2021-07-06 | 2021-07-06 | Multi-user computing unloading resource optimization decision method based on D2D communication |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110762459.5A CN113518330B (en) | 2021-07-06 | 2021-07-06 | Multi-user computing unloading resource optimization decision method based on D2D communication |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113518330A CN113518330A (en) | 2021-10-19 |
| CN113518330B true CN113518330B (en) | 2022-11-11 |
Family
ID=78066878
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110762459.5A Active CN113518330B (en) | 2021-07-06 | 2021-07-06 | Multi-user computing unloading resource optimization decision method based on D2D communication |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113518330B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109788069A (en) * | 2019-02-27 | 2019-05-21 | 电子科技大学 | Calculating discharging method based on mobile edge calculations in Internet of Things |
| CN110401936A (en) * | 2019-07-24 | 2019-11-01 | 哈尔滨工程大学 | A kind of task unloading and resource allocation methods based on D2D communication |
| CN111124666A (en) * | 2019-11-25 | 2020-05-08 | 哈尔滨工业大学 | Efficient and safe multi-user multi-task unloading method in mobile Internet of things |
| CN112000481A (en) * | 2020-08-25 | 2020-11-27 | 东北大学秦皇岛分校 | Task unloading method for maximizing computing capacity of D2D-MEC system |
| CN112969163A (en) * | 2021-02-22 | 2021-06-15 | 安徽大学 | Cellular network computing resource allocation method based on self-adaptive task unloading |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3293927B1 (en) * | 2016-09-09 | 2020-10-21 | Nokia Solutions and Networks Oy | Efficient and dynamic support of mobile low latency services |
| CN107995660B (en) * | 2017-12-18 | 2021-08-17 | 重庆邮电大学 | Joint task scheduling and resource allocation method supporting D2D-edge server unloading |
| EP3826368A1 (en) * | 2019-11-19 | 2021-05-26 | Commissariat à l'énergie atomique et aux énergies alternatives | Energy efficient discontinuous mobile edge computing with quality of service guarantees |
| CN111193615B (en) * | 2019-12-13 | 2023-02-28 | 南京理工大学 | Edge computing node selection method in mobile edge computing network |
| CN111132077B (en) * | 2020-02-25 | 2021-07-20 | 华南理工大学 | Multi-access edge computing task unloading method based on D2D in Internet of vehicles environment |
| CN111556576B (en) * | 2020-05-06 | 2023-04-07 | 南京邮电大学 | Time delay optimization method based on D2D _ MEC system |
-
2021
- 2021-07-06 CN CN202110762459.5A patent/CN113518330B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109788069A (en) * | 2019-02-27 | 2019-05-21 | 电子科技大学 | Calculating discharging method based on mobile edge calculations in Internet of Things |
| CN110401936A (en) * | 2019-07-24 | 2019-11-01 | 哈尔滨工程大学 | A kind of task unloading and resource allocation methods based on D2D communication |
| CN111124666A (en) * | 2019-11-25 | 2020-05-08 | 哈尔滨工业大学 | Efficient and safe multi-user multi-task unloading method in mobile Internet of things |
| CN112000481A (en) * | 2020-08-25 | 2020-11-27 | 东北大学秦皇岛分校 | Task unloading method for maximizing computing capacity of D2D-MEC system |
| CN112969163A (en) * | 2021-02-22 | 2021-06-15 | 安徽大学 | Cellular network computing resource allocation method based on self-adaptive task unloading |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113518330A (en) | 2021-10-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Yang et al. | DEBTS: Delay energy balanced task scheduling in homogeneous fog networks | |
| Zhang et al. | Dynamic task offloading and resource allocation for mobile-edge computing in dense cloud RAN | |
| Diao et al. | Joint computing resource, power, and channel allocations for D2D-assisted and NOMA-based mobile edge computing | |
| CN109684075B (en) | Method for unloading computing tasks based on edge computing and cloud computing cooperation | |
| CN107708135B (en) | Resource allocation method suitable for mobile edge computing scene | |
| CN108809695B (en) | Distributed uplink unloading strategy facing mobile edge calculation | |
| CN111586696B (en) | Resource allocation and unloading decision method based on multi-agent architecture reinforcement learning | |
| Chai et al. | Task execution cost minimization-based joint computation offloading and resource allocation for cellular D2D MEC systems | |
| Chen et al. | Qoe-aware decentralized task offloading and resource allocation for end-edge-cloud systems: A game-theoretical approach | |
| CN109151864B (en) | Migration decision and resource optimal allocation method for mobile edge computing ultra-dense network | |
| WO2023024219A1 (en) | Joint optimization method and system for delay and spectrum occupancy in cloud-edge collaborative network | |
| Zhao et al. | Task proactive caching based computation offloading and resource allocation in mobile-edge computing systems | |
| Sun et al. | Energy-efficient multimedia task assignment and computing offloading for mobile edge computing networks | |
| Li et al. | A delay-aware caching algorithm for wireless D2D caching networks | |
| Zhang et al. | DMRA: A decentralized resource allocation scheme for multi-SP mobile edge computing | |
| Wang et al. | Task allocation mechanism of power internet of things based on cooperative edge computing | |
| Li et al. | Deep reinforcement learning-based resource allocation and seamless handover in multi-access edge computing based on SDN | |
| Meng et al. | Hierarchical evolutionary game based dynamic cloudlet selection and bandwidth allocation for mobile cloud computing environment | |
| Sun et al. | A joint learning and game-theoretic approach to multi-dimensional resource management in fog radio access networks | |
| He et al. | Strategy for task offloading of multi-user and multi-server based on cost optimization in mobile edge computing environment | |
| Di Pietro et al. | An optimal low-complexity policy for cache-aided computation offloading | |
| Cong et al. | CoopEdge: cost-effective server deployment for cooperative multi-access edge computing | |
| CN116828534B (en) | Intensive network large-scale terminal access and resource allocation method based on reinforcement learning | |
| Wang et al. | Offloading schemes in mobile edge computing with an assisted mechanism | |
| CN113518330B (en) | Multi-user computing unloading resource optimization decision method based on D2D communication |
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 |