An energy-efficient edge computing offloading decision and resource allocation method

Abstract

The present invention discloses a high - energy - efficiency edge - computing offloading decision - making and resource - allocation method, including the steps of: establishing an edge - computing network; initializing the offloading decisions of N wireless nodes based on binary and obtaining the corresponding maximum energy - efficiency values and computing - resource allocation results; generating N candidate offloading decisions according to the offloading decision Ml; obtaining the maximum energy - efficiency values and computing - resource allocation results of each candidate offloading decision; obtaining the optimal candidate offloading decision Ml(j*); determining whether the maximum energy - efficiency value θl(j*) corresponding to Ml(j*) is greater than the maximum energy - efficiency value θl corresponding to Ml. If so, taking Ml(j*) and θl(j*) as Ml + 1 and its corresponding maximum energy - efficiency value respectively, and returning to execute the next iteration. Otherwise, taking Ml and the corresponding computing - resource allocation result as the final decision - making allocation result. This method ensures the non - splitability of computing tasks and avoids communication interference between nodes, can quickly solve the optimal decision - making allocation result, thereby improving the energy efficiency of wireless nodes.

Description

Technical Field

[0001] The present invention belongs to the field of edge computing technology, and specifically relates to an energy-efficient edge computing offloading decision-making and resource allocation method. Background Art

[0002] With the rapid development of the Internet of Things, more and more real-time interactive applications have emerged, such as autonomous driving and smart warehousing. However, these applications often involve a lot of computing and require low latency to meet real-time requirements. However, many wireless devices in the Internet of Things have limited computing resources and working power, so there is an unacceptable delay when processing such tasks. Mobile Edge Computing (MEC) technology allows local users to offload part or all of the computing load to edge computing servers, thereby greatly enhancing the task processing capabilities of wireless nodes.

[0003] Most edge computing network research works in the prior art adopt a partial offloading mode, that is, it is assumed that the computing tasks of wireless devices can be split at will, and part of the tasks are offloaded to the edge server for calculation, and part of the tasks are left for local calculation. However, in practical applications, many tasks are often inseparable, and the task offloading communication between nodes often interferes with each other. In addition, because edge servers often have a stable energy source, such as power grid power supply. Wireless nodes often have no reliable energy source and mostly use batteries with limited power, so the energy consumption problem cannot be ignored. Summary of the invention

[0004] The purpose of the present invention is to propose an energy-efficient edge computing offloading decision-making and resource allocation method to address the above-mentioned problems. This method ensures the indivisibility of computing tasks while avoiding mutual interference in task offloading communications between nodes. It can quickly solve the optimal offloading decision and computing resource allocation results, thereby improving the energy efficiency of wireless nodes.

[0005] To achieve the above object, the technical solution adopted by the present invention is:

[0006] The present invention proposes an energy-efficient edge computing offloading decision and resource allocation method, comprising the following steps:

[0007] S1. Establish an edge computing network, which includes an edge server and N wireless nodes;

[0008] S2. Initialize the offloading decision M of N wireless nodes 0 =[m1,m2,…,m i ,…,m N ], and obtain M 0 The maximum energy efficiency value and the corresponding computing resource allocation result, where m i ∈{0,1}, if m i=1, indicating that the i-th wireless node chooses to completely offload the computing task to the edge server, which is recorded as the first type of wireless node. i =0, indicating that the i-th wireless node selects completely local calculation and is recorded as the second type of wireless node. The calculation resource allocation result includes the unloaded transmission power of the first type of wireless node and the CPU calculation frequency of the second type of wireless node;

[0009] S3, according to the unloading decision M of the first iteration l Generate N candidate offloading decisions {M l (1),M l (2),…,M l (j),…,M l (N)}, where represents the binary sum, i.e., transforming the decision of the jth wireless node in the lth iteration, l = 0, 1, 2, ...;

[0010] S4. Obtain the maximum energy efficiency value of each of N candidate offloading decisions and the corresponding computing resource allocation results;

[0011] S5. Take the candidate unloading decision corresponding to the maximum value among the N maximum energy efficiency values ​​as the best candidate unloading decision M. l (j * ),in, θ l (j) is the maximum energy efficiency value of the jth candidate unloading decision in the lth iteration;

[0012] S6. Determine M l (j * ) corresponds to the maximum energy efficiency value θ l (j * ) is greater than M l The corresponding maximum energy efficiency value θ l If so, change M l (j * ) as the next iteration's unloading decision M l+1 , and θ l (j * ) as the next iteration's uninstall decision M l+1 The corresponding maximum energy efficiency value is returned to step S3 to perform the next iteration. Otherwise, M l And the corresponding computing resource allocation results are taken as the final decision allocation results.

[0013] Preferably, each first-class wireless node completely offloads the computing task to the edge server through frequency division multiplexing, and the bandwidth is W / O, where O is the number of first-class wireless nodes and W is the total bandwidth.

[0014] Preferably, the maximum energy efficiency value and the corresponding computing resource allocation result are obtained in the following process:

[0015] S31. Initialize the energy efficiency value according to the current unloading decision or the current candidate unloading decision

[0016] S32, based on the energy efficiency value of the tth iteration Calculate the unloaded transmission power of the first type of wireless nodes and the CPU calculation frequency of the second type of wireless nodes, where:

[0017] The unloaded transmission power of the first type of wireless nodes is as follows:

[0018]

[0019] In the formula, is the unloaded transmission power of the oth wireless node, W is the total bandwidth, is the number of first-class wireless nodes under the current offloading decision or the current candidate offloading decision, h o is the channel gain from the oth wireless node to the edge server, n0 is the power spectrum density of Gaussian white noise, t=0,1,2,…;

[0020] The CPU calculation frequency of the second type of wireless node is as follows:

[0021]

[0022] In the formula, is the CPU calculation frequency of the kth wireless node, κ is the calculation energy efficiency coefficient of the wireless node, and φ is the number of CPU cycles required for the wireless node to locally calculate a one-bit task;

[0023] S33, preset convergence accuracy ξ, and the CPU calculation frequency of the first type of wireless node and the unloaded transmission power of the second type of wireless node are both equal to 0, determine whether it satisfies If yes, get the energy efficiency value for the next iteration Return to step S32 to perform the next iteration. If not, the unloading transmission power of the first type of wireless nodes and the CPU computing frequency of the second type of wireless nodes in this iteration are used as the computing resource allocation result of the corresponding unloading decision, and the energy efficiency value of this iteration is As the maximum energy efficiency value of the corresponding unloading decision, where:

[0024]

[0025] Compared with the prior art, the present invention has the following beneficial effects:

[0026] The method takes the energy efficiency of wireless nodes as the optimization goal and adopts the binary offloading mode of nodes, that is, it is considered that the computing tasks of wireless devices can only be completely calculated locally or completely offloaded to the edge server for computing, and uses the orthogonality of FDMA communication technology to divide the total network bandwidth equally for task offloading, thereby ensuring the indivisibility of computing tasks and avoiding the mutual interference of task offloading communications between nodes, and optimizing the offloading decision and computing resource allocation scheme of wireless nodes. Compared with the existing technology, this method does not only consider the energy consumption of wireless nodes, but also considers the energy efficiency of wireless nodes, that is, the number of task bits that can be processed by consuming one joule of energy, and can quickly solve the optimal offloading decision and computing resource allocation results (including binary offloading decisions, CPU computing frequency of wireless nodes that are completely calculated locally, and offloading transmission power of wireless nodes that are completely task offloaded), thereby improving the energy efficiency of wireless nodes and solving a complex mixed integer programming problem. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 A flowchart of the energy-efficient edge computing offloading decision and resource allocation method of the present invention;

[0028] Figure 2 It is a schematic diagram of the structure of the edge computing network of the present invention.

[0029] Explanation of the accompanying drawings: 1. Edge server; 2. Base station. DETAILED DESCRIPTION

[0030] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0031] It should be noted that, unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application.

[0032] like Figure 1-2 As shown, a highly energy-efficient edge computing offloading decision and resource allocation method includes the following steps:

[0033] S1. Establish an edge computing network, which includes an edge server and N wireless nodes;

[0034] S2. Initialize the offloading decision M of N wireless nodes 0 =[m1,m2,…,mi ,…,m N ], and obtain M 0 The maximum energy efficiency value and the corresponding computing resource allocation result, where m i ∈{0,1}, if m i =1, indicating that the i-th wireless node chooses to completely offload the computing task to the edge server, which is recorded as the first type of wireless node. i =0, indicating that the i-th wireless node selects completely local calculation and is recorded as the second type of wireless node. The calculation resource allocation result includes the unloaded transmission power of the first type of wireless node and the CPU calculation frequency of the second type of wireless node;

[0035] S3, according to the unloading decision M of the first iteration l Generate N candidate offloading decisions {M l (1),M l (2),…,M l (j),…,M l (N)}, where represents the binary sum, i.e., transforming the decision of the jth wireless node in the lth iteration, l = 0, 1, 2, ...;

[0036] S4. Obtain the maximum energy efficiency value of each of N candidate offloading decisions and the corresponding computing resource allocation results;

[0037] S5. Take the candidate unloading decision corresponding to the maximum value among the N maximum energy efficiency values ​​as the best candidate unloading decision M. l (j * ),in, θ l (j) is the maximum energy efficiency value of the jth candidate unloading decision in the lth iteration;

[0038] S6. Determine M l (j * ) corresponds to the maximum energy efficiency value θ l (j * ) is greater than M l The corresponding maximum energy efficiency value θ l If so, change M l (j * ) as the next iteration's uninstall decision M l+1 , and θ l (j * ) as the next iteration's uninstall decision M l+1 The corresponding maximum energy efficiency value is returned to step S3 to perform the next iteration. Otherwise, M l And the corresponding computing resource allocation results are taken as the final decision allocation results.

[0039] In one embodiment, each first-class wireless node completely offloads computing tasks to the edge server through frequency division multiplexing, and the bandwidth is W / O, where O is the number of first-class wireless nodes and W is the total bandwidth.

[0040] In one embodiment, the maximum energy efficiency value and the corresponding computing resource allocation result are obtained in the following process:

[0041] S31. Initialize the energy efficiency value according to the current unloading decision or the current candidate unloading decision

[0042] S32, based on the energy efficiency value of the tth iteration Calculate the unloaded transmission power of the first type of wireless nodes and the CPU calculation frequency of the second type of wireless nodes, where:

[0043] The unloaded transmission power of the first type of wireless nodes is as follows:

[0044]

[0045] In the formula, is the unloaded transmission power of the oth wireless node, W is the total bandwidth, is the number of first-class wireless nodes under the current offloading decision or the current candidate offloading decision, h o is the channel gain from the oth wireless node to the edge server, n0 is the power spectrum density of Gaussian white noise, t=0,1,2,…;

[0046] The CPU calculation frequency of the second type of wireless node is as follows:

[0047]

[0048] In the formula, is the CPU calculation frequency of the kth wireless node, κ is the calculation energy efficiency coefficient of the wireless node, and φ is the number of CPU cycles required for the wireless node to locally calculate a one-bit task;

[0049] S33, preset convergence accuracy ξ, and the CPU calculation frequency of the first type of wireless node and the unloaded transmission power of the second type of wireless node are both equal to 0, determine whether it satisfies If yes, get the energy efficiency value for the next iteration Return to step S32 to perform the next iteration. If not, the unloading transmission power of the first type of wireless nodes and the CPU computing frequency of the second type of wireless nodes in this iteration are used as the computing resource allocation result of the corresponding unloading decision, and the energy efficiency value of this iteration is As the maximum energy efficiency value of the corresponding unloading decision, where:

[0050]

[0051] Specifically, Figure 2 As shown, the edge server 1 can be equipped to the base station 2. In this embodiment, there are 5 wireless nodes with computing needs. Each wireless node needs to make an offloading decision, that is, choose to completely offload the computing task to the edge server or completely calculate locally. The process of choosing a completely offloaded wireless node to complete the task is mainly divided into three parts: wireless node transmission task, edge server computing task and edge server backtransmission of task results, which is a conventional operation well known to people in this field.

[0052] Initialize the offloading decision M of 5 wireless nodes 0 =[1,0,1,0,1], where the 1st, 3rd, and 5th wireless nodes choose full task offloading (i.e., the first type of wireless nodes). These 3 wireless nodes use frequency division multiplexing (FDMA) to perform task offloading communication and equally divide the total bandwidth W of the channel in the edge computing network. That is, the 1st, 3rd, and 5th wireless nodes that choose full task offloading evenly share W / 3 of the offloading transmission bandwidth, and the 2nd and 4th wireless nodes choose full local computing (i.e., the second type of wireless nodes). At the same time, determine the offloading decision M 0 The maximum energy efficiency value and the computing resource allocation result corresponding to the maximum energy efficiency value, the computing resource allocation result includes the CPU computing frequency of the wireless node for complete local computing and the unloading transmission power of the wireless node for complete task offloading;

[0053] According to the current uninstall decision M 0 Generate 5 candidate offloading decisions {M 0 (1),M 0 (2),M 0 (3),M 0 (4),M 0 (5)}, where represents binary sum ( or ), that is, changing the decision of the jth wireless node in the 0th iteration, such as M 0 (1) = [0,0,1,0,1].

[0054] For the five candidate offloading decisions, the respective maximum energy efficiency values ​​and the computing resource allocation results corresponding to the maximum energy efficiency values ​​are calculated respectively. The computing resource allocation results are that for wireless nodes with completely local computing, the CPU computing frequency is calculated, and for wireless nodes with completely task offloading, the offloading transmission power is calculated.

[0055] For example, the candidate unloading decision M 0 (1)=[0,0,1,0,1] as an example, other candidate offloading decisions M 0 (2) M 0 (3) M0 (4) M 0 (5) and the uninstall decision M 0 Similarly, the maximum energy efficiency value and the corresponding computing resource allocation results are determined as follows:

[0056] 1) In the candidate offloading decision M 0 (1) Initialize the energy efficiency value

[0057] 2) Based on the current energy efficiency value If the tth iteration, the energy efficiency value is The following formula is used to calculate the unloaded transmission power of the wireless nodes (the 3rd and 5th wireless nodes) with complete task unloading:

[0058]

[0059] in, For the candidate offloading decision M 0 (1) Select the number of wireless nodes for full task offloading.

[0060] The CPU calculation frequency of the wireless nodes (1st, 2nd, and 4th wireless nodes) that perform completely local calculations is calculated using the following formula:

[0061]

[0062] It is easy to understand that the unloaded transmission power of a wireless node with complete local computing is 0, and the CPU frequency of a wireless node with complete task offloading is 0.

[0063] 3) If If the convergence precision ξ is greater than the preset one, and the convergence precision ξ can be determined according to actual needs, the energy efficiency value can be obtained according to the following formula: Return to step 2) and continue iterating:

[0064]

[0065] like If the convergence precision is less than the preset convergence precision ξ, the CPU computing frequency of the current fully local computing wireless node and the unloading transmission power of the fully task offloading wireless node are used as M 0 (1) The corresponding computing resource allocation result ends the iteration and uses the current energy efficiency value As M 0 (1) The corresponding maximum energy efficiency value θ 0 (1) Output.

[0066] Compare the maximum energy efficiency values ​​of all candidate offloading decisions, and select the candidate offloading decision corresponding to the maximum value of the maximum energy efficiency value as the best candidate offloading decision M 0 (j* ),in, If M 0 (j * ) corresponds to the maximum energy efficiency value θ 0 (j * ) is greater than M 0 The corresponding maximum energy efficiency value θ 0 , then M 0 (j * ) is assigned to the uninstall decision M 1 , return to execute the next iteration, otherwise output the current uninstall decision M 0 and its corresponding computing resource allocation results as the final decision allocation result.

[0067] The method takes the energy efficiency of wireless nodes as the optimization goal and adopts the binary offloading mode of nodes, that is, it is considered that the computing tasks of wireless devices can only be completely calculated locally or completely offloaded to the edge server for computing, and uses the orthogonality of FDMA communication technology to divide the total network bandwidth equally for task offloading, thereby ensuring the indivisibility of computing tasks and avoiding the mutual interference of task offloading communications between nodes, and optimizing the offloading decision and computing resource allocation scheme of wireless nodes. Compared with the existing technology, this method does not only consider the energy consumption of wireless nodes, but also considers the energy efficiency of wireless nodes, that is, the number of task bits that can be processed by consuming one joule of energy, and can quickly solve the optimal offloading decision and computing resource allocation results (including binary offloading decisions, CPU computing frequency of wireless nodes that are completely calculated locally, and offloading transmission power of wireless nodes that are completely task offloaded), thereby improving the energy efficiency of wireless nodes and solving a complex mixed integer programming problem.

[0068] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0069] The above-described embodiments only express the more specific and detailed embodiments described in this application, but they cannot be understood as limiting the scope of the patent application. It should be pointed out that for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of this application, which all belong to the protection scope of this application. Therefore, the protection scope of the patent application shall be based on the attached claims.

Claims

1. A highly energy-efficient edge computing offloading decision and resource allocation method, characterized by: The energy-efficient edge computing offloading decision and resource allocation method comprises the following steps: S1. Establish an edge computing network, wherein the edge computing network includes an edge server and N wireless nodes; S2. Initialize the offloading decision M of N wireless nodes 0 =[m1,m2,…,m i ,…,m N ], and obtain M 0 The maximum energy efficiency value and the corresponding computing resource allocation result, where m i ∈{0,1}, if m i =1, indicating that the i-th wireless node chooses to completely offload the computing task to the edge server, which is recorded as the first type of wireless node. i =0, indicating that the i-th wireless node selects completely local calculation, and is recorded as the second type of wireless node. The calculation resource allocation result includes the unloaded transmission power of the first type of wireless node and the CPU calculation frequency of the second type of wireless node; S3, according to the unloading decision M of the first iteration l Generate N candidate offloading decisions {M l (1),M l (2),…,M l (j),…,M l (N)}, where represents the binary sum, i.e., transforming the decision of the jth wireless node in the lth iteration, l = 0, 1, 2, ...; S4. Obtain the maximum energy efficiency value of each of N candidate offloading decisions and the corresponding computing resource allocation results; S5. Take the candidate unloading decision corresponding to the maximum value among the N maximum energy efficiency values ​​as the best candidate unloading decision M. l (j * ),in, θ l (j) is the maximum energy efficiency value of the jth candidate unloading decision in the lth iteration; S6. Determine M l (j * ) corresponds to the maximum energy efficiency value θ l (j * ) is greater than M l The corresponding maximum energy efficiency value θ l If so, change M l (j * ) as the next iteration's unloading decision M l+1 , and θ l (j * ) as the next iteration's unloading decision M l+1 The corresponding maximum energy efficiency value is returned to step S3 to perform the next iteration. Otherwise, M l And the corresponding computing resource allocation results are taken as the final decision allocation results.

2. The energy-efficient edge computing offloading decision and resource allocation method according to claim 1, characterized in that: Each of the first-class wireless nodes completely offloads the computing tasks to the edge server through frequency division multiplexing, and the bandwidth is W / O, where O is the number of the first-class wireless nodes and W is the total bandwidth.

3. The energy-efficient edge computing offloading decision and resource allocation method according to claim 2, characterized in that: The maximum energy efficiency value and the corresponding computing resource allocation result are obtained as follows: S31. Initialize the energy efficiency value according to the current unloading decision or the current candidate unloading decision S32, based on the energy efficiency value of the tth iteration Calculate the unloaded transmission power of the first type of wireless nodes and the CPU calculation frequency of the second type of wireless nodes, where: The unloaded transmission power of the first type of wireless nodes is as follows: In the formula, is the unloaded transmission power of the oth wireless node, W is the total bandwidth, is the number of first-class wireless nodes under the current offloading decision or the current candidate offloading decision, h o is the channel gain from the oth wireless node to the edge server, n0 is the power spectrum density of Gaussian white noise, t=0,1,2,…; The CPU calculation frequency of the second type of wireless node is as follows: In the formula, is the CPU calculation frequency of the kth wireless node, κ is the calculation energy efficiency coefficient of the wireless node, and φ is the number of CPU cycles required for the wireless node to locally calculate a one-bit task; S33, preset convergence accuracy ξ, and the CPU calculation frequency of the first type of wireless node and the unloaded transmission power of the second type of wireless node are both equal to 0, determine whether it satisfies If yes, get the energy efficiency value for the next iteration Return to step S32 to perform the next iteration. If not, the unloading transmission power of the first type of wireless nodes and the CPU computing frequency of the second type of wireless nodes in this iteration are used as the computing resource allocation result of the corresponding unloading decision, and the energy efficiency value of this iteration is As the maximum energy efficiency value of the corresponding unloading decision, where:

Images