Wireless computing power network resource allocation method based on digital twinning

Abstract

The invention discloses a wireless computing power network resource allocation method based on digital twinning. The method comprises the following steps: constructing a wireless computing power network architecture consisting of a physical entity layer and a digital twinning layer corresponding to the physical entity layer; the physical entity layer comprises a computing node consisting of an edge server and terminal equipment; fitting state information of the edge server and the terminal equipment through digital twinning of the digital twinning layer; according to the fitted state information and the calculation task of the wireless computing power network architecture, establishing an optimization problem with the purpose of minimizing the total time delay and the total energy consumption of the wireless computing power network architecture; and solving the optimization problem to obtain an optimal allocation strategy of the computing resources. According to the method, the entity information of the computing node is acquired in real time through digital twinning, so that available computing resources, bandwidth and other physical entity information of the computing node in the next time slot are estimated, and computing resource allocation and task migration decisions are made under the condition that the total time delay and the total energy consumption of the wireless computing power network architecture are minimized.

Description

Technical Field

[0001] The present invention relates to the field of communication technologies, and in particular to a wireless computing power network resource allocation method based on digital twins. Background Art

[0002] The sixth generation of wireless communications aims to achieve ubiquitous intelligent connectivity in the future integrated network of air, space, land and sea, with extremely low latency and enhanced global coverage, but this requires sufficient computing power support. Therefore, it is imperative to actively explore computing networks to meet ubiquitous computing needs. Wireless computing networks, due to their extensive computing nodes, can better utilize idle computing resources and become an ideal wireless computing environment. However, with the explosive growth of IoT devices, a huge amount of data has been generated, requiring a large amount of computing and communication resources. Cloud computing is a technology that provides computing resources and data storage services through the Internet, which allows users and enterprises to store, process and access data on remote servers without having to do it on local computers or personal servers. Users can access powerful computing power, storage space and various applications on demand, usually on a pay-per-use basis. This model improves resource utilization, reduces infrastructure costs, and allows rapid expansion to meet changing needs. However, cloud computing has problems such as data security, bandwidth issues and high long-term costs that are difficult to avoid. As users' demands for latency and energy consumption become more and more obvious, cloud computing cannot bring users a good experience. Mobile Edge Computing (MEC), as an extension of cloud computing, is a distributed computing paradigm that deploys cloud computing capabilities and IT service environments at the edge of cellular networks, close to the location of mobile users. Its core advantages are to reduce network congestion, lower service delivery latency, and improve application performance. MEC allows for the rapid and flexible deployment of new applications and services by running applications and processing tasks at cellular base stations or other edge nodes. MEC has a wide range of application scenarios, including smart cities, industrial IoT, media entertainment, and healthcare. However, due to the high mobility, limited resources, and time-varying nature of mobile networks, MEC faces challenges in meeting the requirements of 6G networks. The implementation of artificial intelligence algorithms in edge computing has a high consumption of computing resources for data processing and computing, while the heterogeneous capabilities of different devices have also exacerbated the resource shortage of MEC. In addition, the enhanced requirements of 6G applications make existing technologies no longer suitable for latency performance.

[0003] Therefore, how to allocate and compute tasks in wireless computing networks while minimizing latency and energy consumption becomes the main challenge in providing high-quality services in 6G communications. Summary of the invention

[0004] The present invention provides a wireless computing power network resource allocation method based on digital twins, which is used to solve the technical problem that the prior art does not fully consider the latency requirements and energy consumption of the wireless computing power network, making it difficult to efficiently allocate computing resources.

[0005] A first aspect of the present invention provides a wireless computing network resource allocation method based on digital twins, the method comprising:

[0006] Construct a wireless computing network architecture consisting of a physical entity layer and its corresponding digital twin layer; the physical entity layer includes computing nodes consisting of edge servers and terminal devices;

[0007] The digital twin of the digital twin layer is used to fit the status information of the edge server and the terminal device; the status information includes the estimated value of the available computing resources corresponding to the edge server and the terminal device, the deviation between the actual value and the estimated value of the available computing resources, the location information, the DT estimated value of the channel bandwidth, and the deviation between the DT estimated value and the actual value;

[0008] According to the fitted state information and the computing tasks of the wireless computing network architecture, an optimization problem is established with the goal of minimizing the total delay and total energy consumption of the wireless computing network architecture;

[0009] The optimization problem is solved based on the Lyapunov function to obtain the optimal allocation strategy of computing resources.

[0010] Specifically, the physical entity layer includes multiple edge servers, multiple terminal devices, and base stations, and the edge servers are deployed on the base stations; wherein the set of edge servers is represented as , the set of terminal devices is represented as , where the computing nodes It consists of terminal devices and edge servers;

[0011] The digital twin layer includes the digital twin of the edge server and the digital twin of the terminal device, and the digital twin is deployed on the edge server.

[0012] Specifically,

[0013] Edge Server Digital Twin Build as:

[0014]

[0015] Where: Represents edge server An estimate of the computing resources available, Represents edge server Deviation between the actual and estimated values ​​of available computing resources; Indicates the edge server in time slot t The position of; B represents the DT estimation value of the channel bandwidth, The deviation between the estimated DT value and the actual value of the channel bandwidth;

[0016] Terminal equipment Digital Twin Build as:

[0017]

[0018] Where: Indicates terminal equipment An estimate of the computing resources available, Indicates terminal equipment Deviation between the actual and estimated values ​​of available computing resources; Indicates that the terminal device in time slot t location; It is a terminal device The maximum transmission power.

[0019] Specifically, the step of establishing an optimization problem with the goal of minimizing the total delay and total energy consumption of the wireless computing power network architecture according to the fitted state information and the computing task of the wireless computing power network architecture includes:

[0020] According to the fitted state information and the computing tasks of the wireless computing network architecture, a transmission delay model, a computing delay model and an energy consumption model of the wireless computing network architecture are established;

[0021] Based on the computing tasks, transmission delay model, computing delay model and energy consumption model of the wireless computing power network architecture, an optimization problem is established with the goal of minimizing the total delay and total energy consumption of the wireless computing power network architecture.

[0022] Specifically, the transmission delay model includes the actual transmission delay of the computing task from the terminal device to the edge server, the actual transmission delay of the computing task from the edge server to the terminal device, and the data transmission delay between the digital twin and the edge server; the establishment process of the transmission delay model includes:

[0023] Based on the status information of edge servers and terminal devices, the estimated transmission delay of computing tasks between computing nodes is established, and the estimated deviation of computing tasks in the transmission process after the introduction of digital twins is determined. At the same time, the data transmission delay between digital twins and edge servers is established;

[0024] Based on the estimated transmission delay of the computing task between the computing nodes and the corresponding estimated deviation, establish the actual transmission delay of the computing task from the terminal device to the edge server and the actual transmission delay of the computing task from the edge server to the terminal device;

[0025] Among them, in time slot t, the actual transmission delay of the calculation task from the terminal device to the edge server is It is expressed as:

[0026]

[0027] Where: Indicates the task From the terminal device Transfer to edge server The estimated transmission delay of Indicates the task From the terminal device Transfer to edge server The deviation between the estimated transmission delay and the actual transmission delay;

[0028] At time slot t, the actual transmission delay of the task from the edge server to the terminal device is calculated It is expressed as:

[0029]

[0030] Where: Indicates the task From the edge server Transmit to terminal device The estimated transmission delay of Indicates the task From the edge server Transmit to terminal device The deviation between the estimated transmission delay and the actual transmission delay;

[0031] Data transmission delay between digital twin and edge server at time slot t It is expressed as:

[0032]

[0033] Where: w represents the delay required to transmit one unit of data per unit distance, is the size of the data to be transmitted, Represents a normal edge server To the edge server where the digital twin is deployed The shortest distance.

[0034] Specifically, the computing tasks include local computing tasks, single-node computing tasks and multi-node computing tasks; the computing delay model includes the actual computing delay of local computing, the actual computing delay of single-node computing and the actual computing delay of multi-node computing.

[0035] Specifically, the energy consumption model includes the data transmission energy consumption and the computing energy consumption of the computing task; the data transmission energy consumption includes the transmission energy consumption of the computing task between computing nodes and the data transmission energy consumption between the digital twin and the edge server;

[0036] Among them, the data transmission energy consumption between the digital twin and the edge server is It is expressed as:

[0037]

[0038] Where: Indicates the edge server where the digital twin is deployed The transmission power;

[0039] Among them, the computing energy consumption of the computing task is expressed as for:

[0040]

[0041] Where: z represents the computing task involved The number of computing nodes, Indicates that computing node i executes the task The amount of energy consumed.

[0042] Specifically, the step of introducing the Lyapunov function to solve the optimization problem and obtain the optimal allocation strategy of computing resources includes: based on the migration cost loss queue of the Lyapunov function, converting the optimization problem into a target optimization problem without long-term constraints; solving the target optimization problem to obtain the optimal allocation strategy of computing resources;

[0043] Wherein, the target optimization problem is expressed as:

[0044]

[0045] Where: Indicates time slot o is a positive control parameter used to dynamically balance the trade-off between processing delay performance and migration cost consumption; Indicates time slot The total delay of the wireless computing network architecture is Indicates time slot Migration cost loss queue under Indicates that the computing node is in the time slot The migration cost under Indicates time slot The total energy consumption of the wireless computing network architecture.

[0046] The constraints of the target optimization problem include:

[0047]

[0048] Where: C1 represents the total delay of the wireless computing network architecture Cannot exceed the latency requirement of the computational task ; C2 represents the deviation between the mapping value of the digital twin and the corresponding entity value Constraints; C3 means that users can only select Select the best computing node to provide computing services Perform calculations.

[0049] The second aspect of the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, it implements the steps of any of the wireless computing power network resource allocation methods based on digital twins as described above.

[0050] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described methods for allocating wireless computing power network resources based on digital twins.

[0051] It can be seen from the above technical solutions that the present invention has the following advantages:

[0052] The present invention provides a wireless computing power network resource allocation method based on digital twins, including: constructing a wireless computing power network architecture composed of a physical entity layer and its corresponding digital twin layer; the physical entity layer includes computing nodes composed of an edge server and a terminal device; fitting the status information of the edge server and the terminal device through the digital twin of the digital twin layer; the status information includes the estimated value of available computing resources corresponding to the edge server and the terminal device, the deviation between the actual value and the estimated value of the available computing resources, the location information, the DT estimated value of the channel bandwidth, and the deviation between the DT estimated value and the actual value; according to the fitted status information and the computing tasks of the wireless computing power network architecture, establishing an optimization problem with the goal of minimizing the total delay and total energy consumption of the wireless computing power network architecture; solving the optimization problem based on the Lyapunov function to obtain the optimal allocation strategy for computing resources.

[0053] The present invention comprehensively considers the latency and energy consumption of the wireless computing power network, and deploys digital twins in the wireless computing power network architecture to obtain the entity information of the computing nodes in real time, thereby estimating the available computing resources and physical entity information such as bandwidth of the computing nodes in the next time slot, so as to obtain better task migration decisions, and make computing resource allocation and task migration decisions while minimizing the total latency and total energy consumption of the wireless computing power network architecture, so as to make more effective use of the computing resources of terminal devices and edge servers to process increasingly large computing tasks, thereby solving the technical problem that the prior art does not fully consider the latency requirements and energy consumption of the wireless computing power network, making it difficult to efficiently allocate computing resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 paying creative labor.

[0055] Figure 1 A flowchart of a method for allocating wireless computing power network resources based on digital twins provided in an embodiment of the present invention;

[0056] Figure 2 A model diagram of a wireless computing power network architecture provided by an embodiment of the present invention;

[0057] Figure 3 A flowchart of the computing task offloading process and result return process provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0058] An embodiment of the present invention provides a wireless computing power network resource allocation method based on digital twins, which is used to solve the technical problem that the prior art does not fully consider the latency requirements and energy consumption of the wireless computing power network, making it difficult to efficiently allocate computing resources.

[0059] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0060] See also Figure 1The first aspect of the present invention provides a wireless computing network resource allocation method based on digital twins, the method comprising:

[0061] Step 101, construct a wireless computing network architecture consisting of a physical entity layer and its corresponding digital twin layer; wherein the physical entity layer includes computing nodes consisting of edge servers and terminal devices.

[0062] See also Figure 2 The wireless computing network architecture consists of a physical entity layer and its corresponding digital twin layer. The physical entity layer includes multiple edge servers, multiple terminal devices, and base stations. The edge servers are deployed on the base stations, which serve as the data transmission medium between the terminal devices and the edge servers. The set of edge servers is represented as , the set of terminal devices is represented as , where the computing nodes It consists of terminal devices and edge servers.

[0063] In the digital twin layer, the digital twin (DT) is deployed on the computing node. The digital twin obtains the information of the physical entity in real time and performs simulation to replicate the real-time status of the physical entity, provide intelligent decision-making support for the physical entity, and optimize the task allocation strategy. In this embodiment, the digital twin (DT) can be deployed on an edge server with rich computing resources, which can save the energy consumption of data transmission between DT and the computing node.

[0064] In the wireless computing power network architecture provided by the present invention, the computing power of the edge server and the terminal device is aggregated in a computing power pool, so that these computing nodes can be arbitrarily scheduled in the computing power pool, and each computing node can perform computing tasks. Then, when the computing task needs to perform task migration calculation, the computing task can be migrated from the terminal device to the edge server, or from the edge server to the terminal device. Since the terminal device and the edge server can be connected via wireless communication or wired links, the digital twins deployed in the edge server can also be connected to each other.

[0065] In this embodiment, the computing task generated in time slot t is defined as , is the size of the computing task. Computing tasks can be divided into computing tasks that are executed locally and tasks that need to be migrated to other computing nodes for computing; It is the computing resources needed to complete the computing task; It represents the computational task The delay requirement of To represent a collection of time slots.

[0066] Step 102, fitting the status information of the edge server and the terminal device through the digital twin of the digital twin layer.

[0067] It should be noted that the calculation methods of computing tasks can be divided into local computing, single-node computing and multi-node computing. In order to optimize the allocation strategy of computing tasks, the present invention uses digital twins to estimate the physical entity information such as the computing resources and channel bandwidth available in the next time slot based on the computing resources and channel bandwidth of the current time slot, so that better decisions on computing task migration can be made.

[0068] In order to perceive the dynamic distribution of computing power and channel bandwidth, the present invention integrates digital twins into the WCPN architecture, uses a lightweight digital twin model (DT model) to reflect the real-time changing computing resources, and uses the actual available computing resources and channel bandwidth in the computing nodes as training data to train the digital twin model, thereby improving the estimation accuracy of the DT model.

[0069] Then, the digital twin of computing node j is defined as follows:

[0070]

[0071] Where: represents the estimated value of the computing resources available to computing node j, represents the deviation between the actual value and the estimated value of the computing resources available to computing node j; Indicates the location of the computation node j.

[0072] In the WCPN architecture, there are two main types of computing nodes, edge servers and terminal devices. Therefore, the dynamic distribution of information such as perceived computing power and channel bandwidth consists of two types of resources: 1) edge servers; 2) terminal devices; This embodiment uses digital twins to reflect the distribution of channel bandwidth of computing power between various computing nodes.

[0073] Based on the above formula (1), for the edge server Digital Twin Build as:

[0074]

[0075] Where: Represents edge server An estimate of the computing resources available, Represents edge server Deviation between the actual and estimated values ​​of available computing resources; Indicates the edge server in time slot t The position of; B represents the DT estimation value of the channel bandwidth, The DT represents the deviation between the estimated value and the actual value of the channel bandwidth. In this architecture, the location of the edge server is usually fixed, so the edge server of this embodiment Location (t) is a fixed value.

[0076] Terminal equipment Digital Twin Build as:

[0077]

[0078] Where: Indicates terminal equipment An estimate of the computing resources available, Indicates terminal equipment Deviation between the actual and estimated values ​​of available computing resources; Indicates that the terminal device in time slot t location; It is a terminal device The maximum transmission power.

[0079] It is understandable that the terminal device is mobile, so the terminal device Location It can be expressed as:

[0080]

[0081] in, represents the position function of the terminal device, Indicates terminal equipment The moving speed, Indicates terminal equipment moving distance.

[0082] So, for the digital twin of the WCPN architecture It is expressed as:

[0083]

[0084] in, Describes the status information of the computing node in the entire architecture. It describes the estimated rewards of selecting different computing nodes. The status information includes the estimated values ​​of available computing resources corresponding to edge servers and terminal devices, the deviation between the actual value and the estimated value of available computing resources, location information, DT estimated value of channel bandwidth, and the deviation between the DT estimated value and the actual value.

[0085] The present invention estimates the computing resources, bandwidth and other information of the physical entity through digital twins to optimize the task offloading and resource allocation strategies with less delay and energy consumption. At the physical entity layer, the optimal task decision can be selected through the state information fitted by the digital twin layer.

[0086] Step 103, based on the fitted state information and the computing tasks of the wireless computing power network architecture, establish an optimization problem with the goal of minimizing the total delay and total energy consumption of the wireless computing power network architecture.

[0087] It should be noted that this step not only takes into account the computation, transmission delay and energy consumption of the computing task, but also considers the data transmission delay and transmission energy consumption between DT and the mapped edge server, and takes minimizing the total delay and total energy consumption in the above process as the goal, so as to obtain the optimal task decision.

[0088] This step specifically includes:

[0089] Sub-step 1031, based on the fitted state information and the computing tasks of the wireless computing power network architecture, establish a transmission delay model, a computing delay model and an energy consumption model of the wireless computing power network architecture.

[0090] It should be noted that when the computing task selects local computing, the computing task does not involve transmission, so the transmission delay is zero. Then in the present invention, the transmission delay model can include the actual transmission delay of the computing task from the terminal device to the edge server, the actual transmission delay of the computing task from the edge server to the terminal device, and the data transmission delay between the digital twin and the edge server.

[0091] It is understandable that when a user needs to migrate a task from one computing node to another, a migration cost will be incurred, which is defined as follows:

[0092]

[0093] in It represents the task migration decision. When the computing nodes of the tasks in time slot t and time slot t-1 are different, , otherwise 0; Indicates the migration cost from one compute node to another.

[0094] In this sub-step, based on the status information of the edge server and the terminal device, the estimated transmission delay of the computing task between the computing nodes and the estimated deviation of the computing task in the transmission process caused by the introduction of the digital twin can be established, and the data transmission delay between the digital twin and the edge server can be established at the same time; finally, based on the estimated transmission delay of the computing task between the computing nodes and the corresponding estimated deviation, the actual transmission delay of the computing task from the terminal device to the edge server and the actual transmission delay of the computing task from the edge server to the terminal device are established.

[0095] (1) The more specific process of establishing the transmission delay model is as follows:

[0096] First, the estimated transmission speed between computing nodes is established based on the state information of the fitted edge server and terminal device. When computing nodes choose to offload computing tasks, the estimated transmission speed between computing nodes can be divided into two categories: the estimated transmission speed of computing tasks from terminal devices to edge servers and the estimated transmission speed of computing tasks from edge servers to terminal devices;

[0097] In time slot t, the computing task is sent from the terminal device Transfer to edge server Estimated transfer speed for:

[0098]

[0099] In time slot t, the computation task is dispatched from the edge server Transmit to terminal device Estimated transfer speed for:

[0100]

[0101] Where: It is a terminal device Assigned to the edge server in time slot t The estimated bandwidth of It is a terminal device The transmit power at time slot t; It is an edge server Assigned to the terminal device in time slot t The estimated bandwidth of It is an edge server The transmit power at time slot t; Indicates terminal equipment With edge servers The channel gain at time slot t; is the noise power; The location of the terminal device and the location of the edge servers calculated;

[0102] in,

[0103]

[0104] in, It is a terminal device To the edge server is the path loss index between them; I is the interference from other edge servers, and the formula is expressed as:

[0105]

[0106] in, In addition to Other terminal equipment other than In addition to Other edge servers other than ; Indicates the transmit power of other terminal devices; Indicates the channel gain between other terminal devices and other edge servers.

[0107] In particular, if the computing task needs to be sent to a , then the edge server and other edge servers The wired transmission rate between is:

[0108]

[0109] Where: is the distance factor.

[0110] Next, based on formulas (7), (8) and (11), the estimated transmission delay of the computing task between computing nodes is established. From the terminal device Transfer to edge server Estimated transmission delay for:

[0111]

[0112] Where: Represents the edge server The maximum distance covered, for example, the computing task needs to be migrated to the edge server a for computing, because the transmission distance exceeds ,At this time, it is necessary to migrate the computing task to edge server b first, and then migrate from edge server b to edge server a;

[0113] Computational tasks From the terminal device Transfer to edge server The deviation between the estimated transmission delay and the actual transmission delay for:

[0114]

[0115] Then, the computational task From the terminal device Transfer to edge server The actual transmission delay for:

[0116]

[0117] Among them, the computing task From the edge server Transmit to terminal device Estimated transmission delay for:

[0118]

[0119] Computational tasks From the edge server Transmit to terminal device The deviation between the estimated transmission delay and the actual transmission delay for:

[0120]

[0121] Then, the computational task From the edge server Transmit to terminal device The actual transmission delay It is expressed as:

[0122]

[0123] In order to measure the consistency between the digital twin and the physical counterpart, the present invention introduces DT synchronization delay. Then the data transmission delay between the DT model and the mapped edge server is It is expressed as:

[0124]

[0125] Where: is the delay required to transmit one unit of data per unit distance, is the size of the data to be transmitted, is the shortest distance from the edge server to the DT edge server. The DT edge server refers to the edge server deployed with the digital twin, and the digital twin of the DT edge server can obtain the real-time information of other computing nodes.

[0126] In order to measure the accuracy of the DT model, the DT error is introduced: the deviation between the mapped value of the DT model and the actual value of the corresponding entity :

[0127]

[0128] Where: is the weight factor, is the packet loss rate, Represents edge server To DT edge server The number of hops in the shortest path.

[0129] (2) The more specific process of establishing the calculation delay model is as follows:

[0130] In the wireless computing network architecture, computing tasks are divided into three types: local computing, single-node computing, and multi-node computing. It should be noted that local computing means that computing tasks do not need to be transmitted and are directly calculated locally; single-node computing means that computing tasks are migrated to a computing node for calculation. If the computing task exceeds the maximum available computing resources of a single computing node, the computing task can be divided into multiple subtasks and migrated to multiple computing nodes for calculation.

[0131] For local computing:

[0132] In time slot , the estimated computational latency of the computational task performed locally by the terminal device It is expressed as:

[0133]

[0134] in, Indicates time slot Terminal equipment that provides computing services.

[0135] Assuming that the deviation between the terminal device and its DT can be obtained in advance, the calculated delay difference between the DT estimate and the actual value is It is expressed as:

[0136]

[0137] in, Indicates time slot The estimated available computing resources of the terminal device providing the computing service, Indicates time slot The deviation between the estimated available computing resources of the terminal device providing computing services and the actual value.

[0138] Then, the actual computational delay for local computation is It is expressed as:

[0139]

[0140] For single node calculation:

[0141] In time slot , the estimated computational latency of a computational task performed by a single computational node It is expressed as:

[0142]

[0143] in, Indicates time slot Compute nodes that provide computing services.

[0144] Assuming that the deviation between the edge server and its DT can be obtained in advance, the computational delay difference between the DT estimate and the actual value is It is expressed as:

[0145]

[0146] in, Indicates time slot The estimated available computing resources of the computing nodes providing computing services, Indicates time slot The deviation between the estimated available computing resources of the computing nodes providing computing services and the actual value.

[0147] Then, the actual computation delay for single-node computation is It is expressed as:

[0148]

[0149] For multi-node computing:

[0150] Under the required delay constraint, when the executed task exceeds the maximum computing capacity of a single computing node, the computing task needs to be offloaded to multiple computing nodes for collaborative computing. In this case, the computing task Can be divided into multiple subtasks , then in the time slot Estimated computational latency of computing tasks by multiple computing nodes It can be defined as:

[0151]

[0152] in, Represents the estimated available computing resources when executing the lth subtask.

[0153] Among them, in the time slot The difference in computational latency between the estimated computational latency and the actual value of the computational latency for computing a computational task by multiple computing nodes is:

[0154]

[0155] Then, in the time slot The actual computational latency of computing tasks by multiple computing nodes It is expressed as:

[0156]

[0157] (3) The more specific process of establishing the energy consumption model is as follows:

[0158] In the wireless computing network architecture, energy consumption includes the energy consumption of data transmission and task calculation. In the transmission stage, energy consumption is mainly reflected in the uplink task transmission and the downlink result calculation transmission. Since the size of the calculation result is much smaller than the size of the calculation task, this embodiment only considers the energy consumption of the uplink task transmission. Therefore, the transmission energy consumption between the computing nodes is:

[0159]

[0160]

[0161] Where: and They represent the energy consumption of transmitting computing tasks from terminal devices to edge servers and the energy consumption of transmitting computing tasks from edge servers to terminal devices respectively.

[0162] Therefore, the energy consumption of uplink task transmission is:

[0163]

[0164]

[0165] In addition to the uplink task transmission energy consumption, the present invention also considers the DT transmission energy consumption, that is, the data transmission energy consumption between the DT model and the mapped edge server. :

[0166]

[0167] Where: Indicates the edge server where the digital twin is deployed of transmission power.

[0168] For computing node i, execute the task Energy consumed for:

[0169]

[0170] in, is the energy consumed by the unit computing resource, is the effective switch capacitance coefficient; due to the task Can be completed by multiple computing nodes, so the computing task The energy consumption can be expressed as:

[0171]

[0172] Among them, z is the number of tasks involved in the calculation The number of computing nodes.

[0173] Sub-step 1032, based on the computing tasks, transmission delay model, computing delay model and energy consumption model of the wireless computing power network architecture, establish an optimization problem with the goal of minimizing the total delay and total energy consumption of the wireless computing power network architecture.

[0174] The optimization goal of the present invention is to minimize the total delay and total energy consumption of the above process while satisfying the delay requirements of the computing task and not exceeding the constraints of the migration cost.

[0175] Among them, based on the transmission delay model and the calculation delay model, the total delay of the wireless computing network architecture is It is expressed as:

[0176]

[0177] Where: c is a binary variable. When c=0, it means that the task is calculated locally. When c=1 and When c=1 and When , the task is offloaded to multiple computing nodes for calculation.

[0178] Based on the energy consumption model, the total energy consumption of the wireless computing network architecture It is expressed as:

[0179]

[0180] Then, with the goal of minimizing the total delay and total energy consumption mentioned above, we formulate the optimization problem P1:

[0181]

[0182] The constraints of the optimization problem P1 include:

[0183]

[0184] Among them, C1 represents the constraint of the migration cost of migrating computing tasks. It represents the upper limit of the service migration rate during the entire migration process; C2 represents the deviation between the mapping value of the digital twin and the corresponding entity value Constraints; C3 represents the total delay of the wireless computing network architecture Constraints, total delay Cannot exceed the latency requirement of the computational task ; C4 represents the computing node that provides computing services The selection constraint allows users to select only Select the best computing node to provide computing services Perform calculations.

[0185] Step 104, solving the optimization problem based on the Lyapunov function to obtain the optimal allocation strategy of computing resources.

[0186] Since the optimization problem P1 brings about additional optimization problems caused by the long-term service migration cost limit, this step adopts the dynamic loss queue optimization method of Lyapunov function for optimization and solution, and designs a dynamic virtual migration cost loss queue to make appropriate mobile unloading decisions.

[0187] The length of the migration cost loss queue can be defined as the deviation between the used migration cost and the available migration cost. time slots, the total migration cost is , then the migration cost loss queue is expressed as:

[0188]

[0189] in, represents the migration cost loss queue at time slot t, represents the deviation of the migration cost in time slot t, Indicates the upper limit of migration cost for a time slot.

[0190] According to the designed dynamic migration cost loss queue, Lyapunov optimization technology transforms the original optimization problem P1 into a target optimization problem P2 without long-term constraints. The simplified target optimization problem P2 is expressed as:

[0191]

[0192] The constraints of the target optimization problem P2 include:

[0193]

[0194] Among them, o is a positive control parameter used to dynamically balance the trade-off between processing delay performance and migration cost consumption.

[0195] The deep learning algorithm is used to solve the target optimization problem and obtain the optimal allocation strategy for computing resources. For specific allocation strategies, please refer to Figure 3 , local computing can be performed when the computing task does not need to be migrated to other nodes for computing; otherwise, the digital twin can be used to estimate the computing resources and channel bandwidth available in the next time slot based on the computing resources and channel bandwidth of the current time slot, that is, to estimate the resource information available to the physical entity in the next time slot, and guide the migration strategy of the computing task, such as single-node computing or multi-node computing, while minimizing the total delay and total energy consumption, to complete the optimization of computing resource allocation, improve the task assistance capability between computing nodes, and realize the unified scheduling and management of global resources within the wireless computing power network architecture.

[0196] The wireless computing power network resource allocation method based on digital twins provided by the present invention has the following advantages:

[0197] 1. Compared with traditional mobile edge computing, the present invention can make better use of the computing resources of terminal devices and edge servers to handle increasingly large computing tasks.

[0198] 2. The present invention deploys digital twins in the wireless computing network architecture to obtain the physical information of computing nodes in real time, thereby estimating the physical entity information such as the available computing resources and bandwidth of the computing nodes in the next time slot to obtain better task migration decisions.

[0199] 3. The actual available computing resources and channel bandwidth in the computing nodes will be used as new data to train the digital twins, thereby improving the fitting accuracy of the digital twins.

[0200] 4. In the wireless computing network where digital twins are deployed, the DT synchronization delay and the energy consumption generated by data transmission between DT and computing nodes are taken into account when minimizing the optimization target delay and energy consumption, thereby improving the real-time processing capability and optimizing resource allocation capability of the wireless computing network architecture.

[0201] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned wireless computing power network resource allocation method based on digital twins are implemented.

[0202] The present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the steps of the above-mentioned wireless computing power network resource allocation method based on digital twins.

[0203] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0204] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0205] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0206] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0207] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0208] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A wireless computing network resource allocation method based on digital twins, characterized in that: The method comprises: Construct a wireless computing network architecture consisting of a physical entity layer and its corresponding digital twin layer; the physical entity layer includes computing nodes consisting of edge servers and terminal devices; The digital twin of the digital twin layer is used to fit the status information of the edge server and the terminal device; the status information includes the estimated value of the available computing resources corresponding to the edge server and the terminal device, the deviation between the actual value and the estimated value of the available computing resources, the location information, the DT estimated value of the channel bandwidth, and the deviation between the DT estimated value and the actual value; According to the fitted state information and the computing tasks of the wireless computing network architecture, an optimization problem is established with the goal of minimizing the total delay and total energy consumption of the wireless computing network architecture; The optimization problem is solved based on the Lyapunov function to obtain the optimal allocation strategy of computing resources.

2. The wireless computing network resource allocation method according to claim 1, characterized in that: The physical entity layer includes multiple edge servers, multiple terminal devices and base stations, and the edge servers are deployed on the base stations; wherein the set of edge servers is represented as , the set of terminal devices is represented as , where the computing nodes It consists of terminal devices and edge servers; The digital twin layer includes the digital twin of the edge server and the digital twin of the terminal device, and the digital twin is deployed on the edge server.

3. The wireless computing network resource allocation method according to claim 2, characterized in that: Edge Server Digital Twin Build as: Where: Represents edge server An estimate of the computing resources available, Represents edge server Deviation between the actual and estimated values ​​of available computing resources; Indicates the edge server in time slot t The position of; B represents the DT estimation value of the channel bandwidth, The deviation between the estimated DT value and the actual value of the channel bandwidth; Terminal equipment Digital Twin Build as: Where: Indicates terminal equipment An estimate of the computing resources available, Indicates terminal equipment Deviation between the actual and estimated values ​​of available computing resources; Indicates that the terminal device in time slot t location; It is a terminal device The maximum transmission power.

4. The wireless computing network resource allocation method according to claim 3, characterized in that: The step of establishing an optimization problem with the goal of minimizing the total delay and total energy consumption of the wireless computing power network architecture according to the fitted state information and the computing task of the wireless computing power network architecture includes: According to the fitted state information and the computing tasks of the wireless computing network architecture, a transmission delay model, a computing delay model and an energy consumption model of the wireless computing network architecture are established; Based on the computing tasks, transmission delay model, computing delay model and energy consumption model of the wireless computing power network architecture, an optimization problem is established with the goal of minimizing the total delay and total energy consumption of the wireless computing power network architecture.

5. The wireless computing network resource allocation method according to claim 4, characterized in that: The transmission delay model includes the actual transmission delay of the computing task from the terminal device to the edge server, the actual transmission delay of the computing task from the edge server to the terminal device, and the data transmission delay between the digital twin and the edge server; The process of establishing the transmission delay model includes: Based on the status information of edge servers and terminal devices, the estimated transmission delay of computing tasks between computing nodes is established, and the estimated deviation of computing tasks in the transmission process after the introduction of digital twins is determined. At the same time, the data transmission delay between digital twins and edge servers is established; Based on the estimated transmission delay of the computing task between the computing nodes and the corresponding estimated deviation, establish the actual transmission delay of the computing task from the terminal device to the edge server and the actual transmission delay of the computing task from the edge server to the terminal device; Among them, in time slot t, the actual transmission delay of the calculation task from the terminal device to the edge server is It is expressed as: Where: Indicates the task From the terminal device Transfer to edge server The estimated transmission delay of Indicates the task From the terminal device Transfer to edge server The deviation between the estimated transmission delay and the actual transmission delay; At time slot t, the actual transmission delay of the task from the edge server to the terminal device is calculated It is expressed as: Where: Indicates the task From the edge server Transmit to terminal device The estimated transmission delay of Indicates the task From the edge server Transmit to terminal device The deviation between the estimated transmission delay and the actual transmission delay; Data transmission delay between digital twin and edge server at time slot t It is expressed as: Where: w represents the delay required to transmit one unit of data per unit distance, is the size of the data to be transmitted, Represents a normal edge server To the edge server where the digital twin is deployed The shortest distance.

6. The wireless computing network resource allocation method according to claim 4, characterized in that: The computing tasks include local computing tasks, single-node computing tasks and multi-node computing tasks; the computing delay model includes the actual computing delay of local computing, the actual computing delay of single-node computing and the actual computing delay of multi-node computing.

7. The wireless computing network resource allocation method according to claim 4, characterized in that: The energy consumption model includes the data transmission energy consumption and the computing energy consumption of the computing task; the data transmission energy consumption includes the transmission energy consumption of the computing task between the computing nodes and the data transmission energy consumption between the digital twin and the edge server; Among them, the data transmission energy consumption between the digital twin and the edge server is It is expressed as: Where: Indicates the edge server where the digital twin is deployed The transmission power; Among them, the computing energy consumption of the computing task is expressed as for: Where: z represents the computing task involved The number of computing nodes, Indicates that computing node i executes the task The amount of energy consumed.

8. The wireless computing network resource allocation method according to claim 2, characterized in that: The step of introducing the Lyapunov function to solve the optimization problem and obtain the optimal allocation strategy of computing resources includes: based on the migration cost loss queue of the Lyapunov function, converting the optimization problem into a target optimization problem without long-term constraints; solving the target optimization problem to obtain the optimal allocation strategy of computing resources; Wherein, the target optimization problem is expressed as: Where: Indicates time slot o is a positive control parameter used to dynamically balance the trade-off between processing delay performance and migration cost consumption; Indicates time slot The total delay of the wireless computing network architecture is Indicates time slot Migration cost loss queue under Indicates that the computing node is in the time slot The migration cost under Indicates time slot The total energy consumption of the wireless computing network architecture. The constraints of the target optimization problem include: Where: C1 represents the total delay of the wireless computing network architecture Cannot exceed the latency requirement of the computational task ; C2 represents the deviation between the mapping value of the digital twin and the corresponding entity value Constraints; C3 means that users can only select Select the best computing node to provide computing services Perform calculations.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the wireless computing network resource allocation method based on digital twins are implemented as described in any one of claims 1-8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, the steps of the wireless computing network resource allocation method based on digital twins are implemented as described in any one of claims 1 to 8.

Images