Task unloading and resource allocation method under D2D-assisted MEC

Abstract

The invention discloses a task unloading and resource allocation method under D2D-assisted MEC, and belongs to the technical field of wireless communication and computing resource management. According to the method, service equipment assisted by a terminal direct transmission communication technology is introduced, and a task processing mode is reasonably selected, so that efficient utilization of resources is realized, and resource waste is avoided; through joint optimization of calculation unloading, D2D selection, calculation resource allocation and spectrum resource allocation, the service quality can be ensured, and network congestion can be reduced, so that the user experience is improved; the two-stage iterative algorithm combines the block coordinate descent, the reconstruction linearization technology and the convex optimization method, the non-convex optimization problem is effectively solved, and the fast convergence speed and the low calculation complexity are achieved. Besides, the D2D-assisted MEC architecture and the optimization method are suitable for different types of intelligent devices and diversified application scenes, have wide applicability and flexibility, and can meet the requirements of future mobile edge computing development.

Description

Technical Field

[0001] The present invention relates to a method for task offloading and resource allocation under D2D-assisted MEC, belonging to the technical field of wireless communication and computing resource management. Background Art

[0002] With the popularity of smart devices (SD) in modern communications and computing, the demand for computing power and storage capacity is growing. However, due to hardware limitations, it is difficult for a single smart device to efficiently handle complex computing tasks. To solve this problem, Mobile Edge Computing (MEC) has become a key technology that helps smart devices handle computing tasks by deploying computing resources at the edge of the network.

[0003] The introduction of direct-to-device (D2D) technology enables the MEC architecture to communicate directly between devices and share computing resources, further improving resource utilization efficiency. D2D communication can not only provide communication services with lower latency, but also reduce the computing burden of smart devices by completing computing tasks collaboratively between terminal devices, thereby enhancing the computing power and resource utilization efficiency of the overall system.

[0004] In traditional MEC systems, resource management solutions usually rely on a single computing offloading mode and fail to fully utilize various resources in the network, resulting in low resource utilization, severe network congestion, and poor quality of service (QoS). Therefore, designing a solution that can fully utilize multiple resources and optimize offloading strategies has become an urgent problem to be solved. Summary of the invention

[0005] In order to solve the problems of low resource utilization and high execution cost in the current mobile edge computing (MEC) system, the present invention provides a task offloading and resource allocation method under D2D assisted MEC, and the technical solution is as follows:

[0006] The task offloading and resource allocation method of the present invention comprises:

[0007] Step 1: Build a D2D communication-assisted MEC system model, define the task processing mode of smart devices and the computing task offloading mode of edge servers, and divide the tasks generated by task-type smart devices into three task processing modes: local computing, D2D offloading, and edge offloading;

[0008] Step 2: Taking minimizing the task execution cost as the optimization goal, the task partitioning strategy, computing resource allocation strategy, terminal direct transmission offloading selection strategy and dynamic pricing strategy are jointly formulated;

[0009] Step 3: Use the block coordinate descent method and reconstruction-linearization technique to decompose the joint optimization problem into multiple sub-problems, approximate the non-convex problem into a convex optimization problem, and apply convex optimization theory to further optimize the task offloading ratio and computing resource allocation strategy.

[0010] Optionally, the task execution cost in step 2 is:

[0011]

[0012] in, denotes the number of computing bits allocated to local computing, D2D offloading, and edge offloading tasks, respectively, and m l , and m e represents the unit price of computing services per bit provided by task-oriented intelligent devices, service-oriented intelligent devices and edge servers, respectively, and x i,j ∈{0,1} is a binary selection variable, x i,j =1 indicates task-oriented intelligent device Service-oriented smart devices There is a D2D communication link, x i,j =0 means there is no D2D communication link between the two;

[0013] The task division strategy, computing resource allocation strategy, terminal direct offloading selection strategy and dynamic pricing strategy are jointly formulated as follows:

[0014]

[0015] Among them, L i Represents task-oriented intelligent devices The number of computational bits of the generated task, Represents SSDs The frequency of executing local task calculations, express Assigned to The calculation frequency of express The maximum frequency of Indicates that BS is assigned to The calculation frequency of Indicates the maximum frequency of BS; represents the delay of the two stages of D2D offloading and edge offloading, T i represents the maximum tolerable time threshold, E i r After the three stages of local computing, D2D offloading, and edge offloading, remaining battery energy.

[0016] Optionally, step 3 includes:

[0017] Step 31: Given a selection strategy X m , solve the following problem to obtain the local optimal unloading solution L under the current number of iterations m +1 and calculate the frequency scheme F m+1 :

[0018]

[0019] Among them, μ i,j is an auxiliary variable, C i express The number of CPU cycles required to process a unit bit task; C j express The number of CPU cycles required to process a unit bit task; Indicates the data transmission rate during the D2D offloading phase; L j Indicates the total data size of the tasks to be executed on SSDs; ν i represents auxiliary variables, Indicates the data transmission rate in edge offloading; Indicates the D2D offloading phase The transmission power; p i e Indicates the edge unloading phase Transmitting power when transmitting data to BS; E i express Initial energy state; represents the energy consumed in the local computation phase;

[0020] Step 32: The local optimal unloading solution L obtained according to step 31 m+1 and calculate the frequency scheme F m+1 , calculate the local optimal selection strategy X m+1 , the solution is as follows:

[0021]

[0022] Use the solver to solve the above equation and obtain a feasible solution;

[0023] Step 33: Iterate the steps 31 and 32 until an unloading scheme, calculation frequency and selection strategy that meet preset conditions are obtained.

[0024] Optionally, the communication process adopts orthogonal frequency division multiple access technology.

[0025] Optionally, step 3 uses a CVX solver to solve the convex optimization problem.

[0026] Optionally, the data transmission rate in the D2D offloading phase is The calculation method is:

[0027]

[0028] Where B represents the transmission bandwidth of each subchannel. and Respectively The transmit power and D2D link channel power gain are represented by , and N represents the noise power of each subchannel.

[0029] Optional, data transfer rate during edge offloading The calculation formula is:

[0030]

[0031] Where B represents the transmission bandwidth of each subchannel. and Respectively Transmitting power and cellular link channel gain when transmitting data to SSDs via D2D link, N represents the noise power of each subchannel.

[0032] The present invention provides a task offloading and resource allocation system, the system is used to implement the method as described in any one of the above, the system comprising:

[0033] The system model building module is configured to build a D2D communication-assisted MEC system model, define the task processing mode of the smart device and the computing task offloading mode of the edge server, and divide the tasks generated by the task-type smart device into three task processing modes: local computing, D2D offloading, and edge offloading;

[0034] The joint optimization formula building module is configured to take minimizing the task execution cost as the optimization goal, and jointly formulate the task partitioning strategy, computing resource allocation strategy, terminal direct transmission offloading selection strategy and dynamic pricing strategy;

[0035] The solving module is configured to decompose the joint optimization problem into multiple sub-problems by using the block coordinate descent method and reconstruction-linearization technique, approximate the non-convex problem into a convex optimization problem, and further optimize the task offloading ratio and computing resource allocation strategy by applying convex optimization theory.

[0036] The present invention provides an electronic device, characterized in that it includes a memory and a processor;

[0037] The memory is used to store computer programs;

[0038] The processor is configured to implement any of the methods described above when executing the computer program.

[0039] The present invention provides a computer-readable storage medium, characterized in that a computer program is stored on the storage medium, and when the computer program is executed by a processor, the method described in any one of the above items is implemented.

[0040] The beneficial effects of the present invention are:

[0041] The present invention minimizes the user's overhead on communication and computing services by optimizing the computing offloading and resource allocation strategy, thereby effectively reducing the overall execution cost of edge users. By introducing service-oriented devices (SSDs) assisted by terminal direct communication (D2D) technology and reasonably selecting task processing modes (local execution, D2D offloading, edge computing), efficient resource utilization is achieved and resource waste is avoided. The present invention can not only ensure the quality of service (QoS) but also reduce network congestion by jointly optimizing computing offloading, D2D selection, computing resource allocation and spectrum resource allocation, thereby improving user experience. The designed two-stage iterative algorithm combines block coordinate descent, reconstruction linearization technology (RLT) and convex optimization methods to effectively solve non-convex optimization problems, with faster convergence speed and lower computational complexity. In addition, the D2D-assisted MEC architecture and optimization method are suitable for different types of smart devices and diversified application scenarios, have wide applicability and flexibility, and can meet the needs of future mobile edge computing development. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0043] Figure 1 This is a relationship diagram between the number of TSDs and the total system cost under different unloading schemes of the present invention.

[0044] Figure 2 This is a relationship diagram between the maximum tolerable delay and the total system cost under different unloading schemes of the present invention.

[0045] Figure 3 This is a relationship diagram between task data size and total system cost under different unloading schemes of the present invention. DETAILED DESCRIPTION

[0046] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0047] Embodiment 1:

[0048] This embodiment provides a task offloading and resource allocation method, including:

[0049] Step 1: Build a D2D communication-assisted MEC system model, define the task processing mode of smart devices and the computing task offloading mode of edge servers, and divide the tasks generated by task-type smart devices into three task processing modes: local computing, D2D offloading, and edge offloading;

[0050] Step 2: Taking minimizing the task execution cost as the optimization goal, the task partitioning strategy, computing resource allocation strategy, terminal direct transmission offloading selection strategy and dynamic pricing strategy are jointly formulated;

[0051] Step 3: Use the block coordinate descent method and reconstruction-linearization technique to decompose the joint optimization problem into multiple sub-problems, approximate the non-convex problem into a convex optimization problem, and apply convex optimization theory to further optimize the task offloading ratio and computing resource allocation strategy.

[0052] Embodiment 2:

[0053] This embodiment provides a method for partial offloading of multiple tasks in an MEC environment based on terminal direct communication assistance. The system model applicable to the method is: a MEC-assisted D2D communication system, which consists of a base station (BS) and several smart devices (SDs), and the edge server is deployed on the BS. According to whether the local computing resources can complete the task on time, all SDs in the system are divided into task-oriented smart devices (TSDs) and service-oriented smart devices (SSDs), which are represented as and SSDs can establish independent D2D links with TSDs to assist Complete the task within the maximum tolerable time threshold. If some tasks cannot be completed locally on TSDs, they can be offloaded to edge servers via cellular links or to SSDs via D2D links.

[0054] Assume that TSDs generate a data tuple for a computationally intensive, latency-critical task Among them C i Indicates the number of CPU cycles required to process a unit bit task, L i Indicates the total amount of current task data, T iIndicates the maximum tolerable delay limit for executing tasks. In addition, to encourage SSDs to serve more users, TSDs adopt a dynamic charging mechanism to pay for the computing resources occupied by the unloading process, alleviating the congestion problem of limited resources under multiple users. Tasks on TSDs are divided into three parts, namely, some tasks are calculated locally, some tasks are offloaded to SSDs for calculation, and some tasks are offloaded to edge servers for calculation. and The number of task calculation bits assigned to local computing, D2D offloading, and edge offloading is represented respectively. Obviously, different task division ratios affect the battery energy and resource payment of TSDs. Generate L i tasks, and in Tasks perform local computations, The tasks are executed on SSDs via D2D links, and The task is offloaded to BS for processing.

[0055] The computation offloading and resource allocation method of this embodiment includes:

[0056] Step 1: Build a direct-to-device (D2D) assisted MEC system model to define the task processing mode of smart devices and the computing task offloading mode of edge servers.

[0057] Using the Decision Matrix Describes the D2D link association relationship, where x i,j ∈{0,1} is a binary selection variable, x i,j =1 means and The D2D communication link is successfully established, and x i,j =0 indicates that there is no D2D communication link between the two. and Respectively represent the number of task calculation bits allocated to the three stages of local computing, D2D offloading and edge offloading. Generate L i tasks, and in Tasks perform local computations, The tasks are executed on SSDs via D2D links, and The task is offloaded to the BS for processing. For example, the latency of the three stages of local computing, D2D offloading, and edge offloading is In particular, the delay of the two stages of D2D offloading and edge offloading also includes two states: transmission and calculation, marked as In time block T iIn the example, it is assumed that parameters such as channel allocation and transmit power are constant, and the transmission bandwidth and noise power of each subchannel are marked as B and N respectively.

[0058] 1) D2D offload communication: With the help of SSDs computing resources, the offloaded data packets are transferred from Directly transfer to In this process, the TSDs transmission power and the D2D link channel power gain under (i, j) pairing are defined as and According to Shannon's theorem, the data transmission rate in the D2D offloading phase is calculated Right now:

[0059]

[0060] The time and energy consumed in the D2D offloading phase are mainly affected by the D2D link transmission The TSDs task is affected. The delay and energy consumption expressions of this process are:

[0061]

[0062] 2) Edge offload communication: If the local computing cannot complete the task in time, it will be offloaded to the BS through the cellular link. Similar to the definition of D2D offloading stage, and Respectively expressed as The transmit power and cellular link channel gain when transmitting data to the BS. Therefore, the data transmission rate in edge offloading Calculated as:

[0063]

[0064] Perform edge offload processing of packets only So calculate The mathematical expressions of transmission delay and energy loss are:

[0065]

[0066]

[0067] Since BS and SSDs have certain computing and storage resources, Provide limited computing, caching and storage services, and reasonably divide resources to maximize utilization. Assume that TSDs and SSDs all use simplex communication mode and perform data processing after successfully receiving data. Depending on the computing mode, the computing model can be divided into two parts: local computing model and offload model.

[0068] 1) Local computing

[0069] Under the condition of energy consumption and resource constraints, TSDs will fully utilize its own computing resources to reduce the proportion of computing and communication resources occupied by offloaded data. The time consumed in the local computing phase is:

[0070]

[0071] Among them, f i for CPU frequency. To maximize the use of local resources, assume Then the local computing task data size is The energy consumption in the local computing phase is:

[0072]

[0073] where κ i Represents the effective switch capacitance coefficient.

[0074] 2) D2D offloading calculation

[0075] Received uninstall After receiving the data packet, it uses its remaining computing resources to process the data packet and returns the computing results to TSDs. During D2D offloading, the time it takes for SSDs to execute tasks It is expressed as:

[0076]

[0077] in, for The frequency of the calculation called. It is worth noting that SSDs may have tasks to be executed, so they need to execute local tasks first and then assist TSDs to complete the offloading. Therefore, the prerequisite for D2D offloading is that SSDs complete local tasks and complete the offloading data packet reception. represents the local computing time of SSDs, and the total delay of D2D offloading phase is:

[0078]

[0079] in,

[0080] 3) Edge offloading computing

[0081] TSDs successfully transmitted to BS After the packet, the edge server is Reasonably allocate computing resources and assist in completing task offloading. The time consumption of this process is:

[0082]

[0083] in, Assigned to BS Calculate the frequency. So far, the total edge offloading delay is:

[0084]

[0085] Step 2: Establish a joint optimization problem for multi-task partial offloading and jointly formulate the task partitioning strategy, computing resource allocation strategy, terminal direct offloading selection strategy and dynamic pricing strategy.

[0086] For local computing tasks D2D offloading tasks and edge offloading tasks definition The unit price of bit computing service provided by BS is m l , and m e In order to complete the task within the specified time, the D2D-MEC system Execution cost M i It is expressed as:

[0087]

[0088] Assumptions The initial energy state is marked as E i , after three stages of processing: local computing, D2D offloading, and edge offloading, Remaining battery energy state E i r Calculated as

[0089]

[0090] The task offloading scheme, computing resource allocation and D2D association strategy are jointly optimized to model the execution cost problem as:

[0091]

[0092] Among them, constraints C1 and C2 indicate the constraints satisfied by the task partitioning strategy, C3, C4 and C5 limit the execution of tasks within the specified time and CPU frequency range, C6 indicates the battery energy state that TSDs need to maintain, and C7 and C8 represent the D2D association strategy. Obviously, due to the discrete optimization variable x i,j The existence of the objective function and the constraints C7 and C9 is non-convex, which indicates that the objective function P1 is a non-convex problem and is difficult to solve directly using traditional convex optimization techniques.

[0093] Step 3: The block coordinate descent method and reconstruction-linearization technique (RLT) are used to decompose the joint optimization problem into multiple sub-problems. The non-convex problem is approximated as a convex optimization problem, and the convex optimization theory is applied to further optimize the task offloading ratio and computing resource allocation strategy.

[0094] use Represents the selection strategy of all SSDs, given the initial strategy X 0 is a distance-based selection strategy, X l (l=1,2,...,L) represents the selection strategy at the lth iteration. l When Resources provided to Thus, a pairing relationship (i, j) is formed between devices. At this time, there is a D2D association matrix x i,j = 1. Combining formula (10), formula (12) and formula (14), P1 can be rewritten as:

[0095]

[0096] Since there is a variable coupling relationship between C5 and C7 and Therefore, problem P2 is a non-convex problem, and the product term is transformed by using the reconstruction linearization technique (RLT). Introducing the auxiliary variable μ i,j and ν i ,make but μ i,j Substitute the RLT constraint factor product restriction into Simplifying, we get:

[0097]

[0098] Similarly, substitute Then we can get:

[0099]

[0100] In summary, problem P2.1 can be expressed as follows after RLT eliminates the quadratic term:

[0101]

[0102] Analysis shows that the transformed objective function P2.2 is a convex optimization problem, which can be solved with the help of convex optimization techniques, such as the CVX toolbox, to obtain the local optimal solution under the current number of iterations.

[0103] After resource allocation, the local optimal solution and Based on, the original optimization problem P1 is transformed into the SSDs selection subproblem. Since the decision variable x in P3 i,j For discrete variables, the objective function P3 is a non-convex function, and the relaxation variable method is needed to reduce the complexity of the algorithm. i,j , relax the restriction condition to 0≤x i,j ≤1. Then P3 is simplified to:

[0104]

[0105] After the transformation, problem P3.1 is convex and can be solved by the CVX solver. i,j To recover, the specific iterative steps of the system execution cost optimization problem are shown in Table 1.

[0106] Table 1 Task offloading and resource allocation algorithms

[0107]

[0108]

[0109] In order to make the purpose, technical solution and advantages of the present invention clearer, some classic detection algorithms will be used to compare with the method proposed in the present invention to show the performance superiority of the computing offloading and resource allocation method under the D2D-assisted MEC architecture of the present invention.

[0110] The baseline algorithms used for simulation are the traditional partial offloading scheme, the collaborative computing offloading scheme, and the relay-assisted offloading scheme. Figure 1 , 2 As shown in Figure 3, it can be seen from the simulation curve that by comparing the total costs under different numbers of intelligent devices, maximum tolerable delays and task data sizes, it is found that the computational offloading and resource allocation method of the present invention can effectively optimize the total system cost.

[0111] In summary, the simulation results verify the performance of the joint optimization method of the present invention in different scenarios, and compared with the baseline algorithm, the algorithm can greatly reduce the execution cost. Therefore, the computation offloading and resource allocation method of the present invention can effectively improve the performance of the network system while completing the low-latency task.

[0112] Some steps in the embodiments of the present invention may be implemented using software, and the corresponding software program may be stored in a readable storage medium, such as a CD or a hard disk.

[0113] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A task offloading and resource allocation method, characterized in that: The method comprises: Step 1: Build a D2D communication-assisted MEC system model, define the task processing mode of smart devices and the computing task offloading mode of edge servers, and divide the tasks generated by task-type smart devices into three task processing modes: local computing, D2D offloading, and edge offloading; Step 2: Taking minimizing the task execution cost as the optimization goal, the task partitioning strategy, computing resource allocation strategy, terminal direct transmission offloading selection strategy and dynamic pricing strategy are jointly formulated; Step 3: Use the block coordinate descent method and reconstruction-linearization technique to decompose the joint optimization problem into multiple sub-problems, approximate the non-convex problem into a convex optimization problem, and apply convex optimization theory to further optimize the task offloading ratio and computing resource allocation strategy.

2. The method according to claim 1, characterized in that The task execution cost in step 2 is: in, denotes the number of computing bits allocated to local computing, D2D offloading, and edge offloading tasks, respectively, and m l , and m e represents the unit price of computing services per bit provided by task-oriented intelligent devices, service-oriented intelligent devices and edge servers, respectively, and x i,j ∈{0,1} is a binary selection variable, x i,j =1 indicates task-oriented intelligent device Service-oriented smart devices There is a D2D communication link, x i,j =0 means there is no D2D communication link between the two; The task division strategy, computing resource allocation strategy, terminal direct offloading selection strategy and dynamic pricing strategy are jointly formulated as follows: Among them, L i Represents task-oriented intelligent devices The number of computational bits of the generated task, express The frequency of executing local task calculations, express Assigned to The calculation frequency of express The maximum frequency of Indicates that BS is assigned to The calculation frequency of Indicates the maximum frequency of the BS; k={d,e} represents the delay of the two stages of D2D offloading and edge offloading, T i Represents the maximum tolerable time threshold, After the three stages of local computing, D2D offloading, and edge offloading, remaining battery energy.

3. The method according to claim 2, characterized in that The step 3 comprises: Step 31: Given a selection strategy X m , solve the following problem to obtain the local optimal unloading solution L under the current number of iterations m+1 and calculate the frequency scheme F m+1 : Among them, μ i,j is an auxiliary variable, C i express The number of CPU cycles required to process a unit bit task; C j express The number of CPU cycles required to process a unit bit task; Indicates the data transmission rate during the D2D offloading phase; L j Indicates the total data size of the tasks to be executed on SSDs; ν i represents auxiliary variables, Indicates the data transmission rate in edge offloading; Indicates the D2D offloading phase The transmission power; Indicates the edge unloading phase Transmitting power when transmitting data to BS; E i express Initial energy state; represents the energy consumed in the local computation phase; Step 32: The local optimal unloading solution L obtained according to step 31 m+1 and calculate the frequency scheme F m+1 , calculate the local optimal selection strategy X m+1 , the solution is as follows: Use the solver to solve the above equation and obtain a feasible solution; Step 33: Iterate the steps 31 and 32 until an unloading scheme, calculation frequency and selection strategy that meet preset conditions are obtained.

4. The method according to claim 1, characterized in that The communication process adopts orthogonal frequency division multiple access technology.

5. The method according to claim 1, characterized in that The step 3 uses the CVX solver to solve the convex optimization problem.

6. The method according to claim 3, characterized in that The data transmission rate in the D2D offloading phase The calculation method is: Where B represents the transmission bandwidth of each subchannel. and Respectively The transmit power and D2D link channel power gain are represented by , and N represents the noise power of each subchannel.

7. The method according to claim 3, characterized in that Data transfer rate in edge offloading The calculation formula is: Where B represents the transmission bandwidth of each subchannel. and Respectively Transmitting power and cellular link channel gain when transmitting data to SSDs via D2D link, N represents the noise power of each subchannel.

8. A task offloading and resource allocation system, characterized in that: The system is used to implement the method according to any one of claims 1 to 7, and the system includes: The system model building module is configured to build a D2D communication-assisted MEC system model, define the task processing mode of the smart device and the computing task offloading mode of the edge server, and divide the tasks generated by the task-type smart device into three task processing modes: local computing, D2D offloading, and edge offloading; The joint optimization formula building module is configured to take minimizing the task execution cost as the optimization goal, and jointly formulate the task partitioning strategy, computing resource allocation strategy, terminal direct transmission offloading selection strategy and dynamic pricing strategy; The solving module is configured to decompose the joint optimization problem into multiple sub-problems by using the block coordinate descent method and reconstruction-linearization technique, approximate the non-convex problem into a convex optimization problem, and further optimize the task offloading ratio and computing resource allocation strategy by applying convex optimization theory.

9. An electronic device, characterized in that: including memory and processor; The memory is used to store computer programs; The processor is configured to implement the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images