Energy-saving unloading method, device and equipment for meta-computing task and storage medium

Abstract

The invention provides an energy-saving unloading method and device for meta-computing tasks, equipment and a storage medium. The method comprises the following steps: adding white noise into a task signal to form a sending signal; capacity is calculated according to signals received by the main channel and the eavesdropping channel, and then the secrecy rate and the traversal value and the lower limit of the secrecy rate are determined; calculating power consumption and delay between a plurality of meta-calculation users and the node through traversing the secrecy rate; constructing a bipartite graph to determine an unloading path; and calculating the optimal power of each path according to the lower limit of the traversal secrecy rate. According to the method, task signals and white noise are combined, the secrecy rate is calculated through channel information, power consumption, delay and unloading paths are determined according to the secrecy rate, the optimal power is calculated, and safe, efficient and energy-saving unloading of calculation tasks is achieved. The whole process aims at optimizing the unloading strategy of the meta-computing task, improving the system efficiency and reducing the energy consumption.

Description

Technical Field

[0001] The present application belongs to the field of computing task offloading, and in particular, relates to an energy-saving offloading method, device, equipment and storage medium for meta-computing tasks. Background Art

[0002] With the rapid development of information technology, metacomputing, as a distributed computing model, provides users with efficient and flexible computing services by integrating and utilizing a large amount of computing resources. However, in a metacomputing environment, the offloading and execution of computing tasks not only need to consider computing efficiency and resource utilization, but also need to pay attention to energy consumption and security. Especially in wireless communication environments, how to ensure the secure transmission and offloading of computing tasks while reducing energy consumption has become an urgent problem to be solved.

[0003] In traditional meta-computing task offloading methods, users usually directly send computing tasks to meta-computing nodes for processing. Although this method is simple and direct, it ignores the security and energy consumption issues during the transmission process. Especially in wireless communication environments, due to the openness and broadcast nature of signal transmission, computing tasks may be eavesdropped and attacked during transmission, resulting in data leakage. At the same time, energy consumption during transmission is also an issue that cannot be ignored. Excessive energy consumption will not only increase user costs, but also have a negative impact on the environment. Summary of the invention

[0004] The purpose of this application is to overcome the defects existing in the above-mentioned prior art and to provide an energy-saving offloading method, device, equipment and storage medium for meta-computing tasks.

[0005] The present application provides an energy-saving offloading method for a meta-computing task, comprising:

[0006] The sending signal is obtained by adding white noise to the task signal of the meta-computing user sending the computing task;

[0007] Calculating the main channel capacity and the wiretap channel capacity according to the transmission signals received by the main channel and the wiretap channel respectively;

[0008] Calculating the confidentiality rate of the transmitted signal according to the main channel capacity and the wiretap channel capacity, and determining the ergodic confidentiality rate and the ergodic confidentiality rate lower limit according to the confidentiality rate;

[0009] Calculate the power consumption and delay between the plurality of said meta-computing users and the plurality of meta-computing nodes under the maximum radio frequency power by using the ergodic confidentiality rate;

[0010] constructing a bipartite graph of the meta-computing user and the meta-computing node according to the power consumption and the delay;

[0011] Determining, according to the bipartite graph, an offloading path for the meta-computing user to offload the computing task to the meta-computing node;

[0012] Calculating the optimal power of each of the offloading paths according to the traversal confidentiality rate lower limit;

[0013] The computing task of the meta-computing user is offloaded to the meta-computing node according to the offloading path and the optimal power.

[0014] Optionally, the formulas for the main channel capacity and the wiretap channel capacity are respectively:

[0015]

[0016] Where I is the number of metacomputing users, J is the number of metacomputing nodes, and H i,j represents the channel matrix between the ith meta-computing user and the jth meta-computing node, H i,e represents the channel matrix between the i-th element-calculating user and the eavesdropper,

[0017] C i,j and C i,e They represent the main channel capacity and the eavesdropping channel capacity respectively, P i represents the transmit power of the sender, where Represents the meta-computation node O j The noise power at the receiving end, B represents the transmission bandwidth, log2(.) represents the logarithm operation with base 2, det(.) represents the matrix determinant, represents the Rayleigh quotient, and the admissible detection vector of the eavesdropper is defined as the solution to the Rayleigh quotient maximization problem, where represents the noise power at the eavesdropper’s receiving end, is the identity matrix, H1 and g i,e is the equivalent channel matrix, N i represents the number of antennas of the user for the calculation, represent The maximum eigenvalue of , (.) H is the conjugate transpose operation.

[0018] Optionally, the confidentiality rate of the transmitted signal is calculated according to the main channel capacity and the wiretap channel capacity, and the expression is as follows:

[0019] R i,h =[C i,j -C i,e ] +

[0020] Where [x] + =max(x,0).

[0021] Optionally, constructing a bipartite graph of the meta-computing user and the meta-computing node according to the power consumption and the delay includes:

[0022] The meta-computing nodes whose computing power is lower than the preset threshold are deleted, and the connection weight between each meta-computing user and the meta-computing node is calculated based on power consumption and latency.

[0023] Optionally, before offloading the computing task of the meta-computing user to the meta-computing node, the method further includes:

[0024] The computing tasks are segmented and encoded.

[0025] Optionally, calculating the optimal power of each of the offloading paths according to the traversal confidentiality rate lower limit includes:

[0026] The minimum power consumption point that meets the delay constraint is found within a given RF power range through numerical methods.

[0027] Optionally, the construction of the white noise includes:

[0028] Calculating a singular value decomposition of the task signal null space;

[0029] The white noise is obtained by selecting columns corresponding to zero singular values ​​as a generator matrix of the white noise.

[0030] The present application also provides an energy-saving unloading device for a meta-computing task, comprising:

[0031] A signal module, which adds white noise to the task signal of the meta-computing user sending the computing task to form a sending signal;

[0032] A capacity module, calculating the main channel capacity and the wiretap channel capacity according to the transmission signals received by the main channel and the wiretap channel;

[0033] A confidentiality rate module, which calculates the confidentiality rate of the transmitted signal according to the capacity of the main channel and the capacity of the wiretap channel, and determines the ergodic confidentiality rate and the lower limit of the ergodic confidentiality rate according to the confidentiality rate;

[0034] A calculation module, which calculates the power consumption and delay between the plurality of said meta-computing users and the plurality of meta-computing nodes under the maximum radio frequency power by using the ergodic confidentiality rate;

[0035] A graph module, constructing a bipartite graph of the meta-computing user and the meta-computing node according to the power consumption and the delay;

[0036] A path module, which determines, according to the bipartite graph, an offloading path for the meta-computing user to offload the computing task to the meta-computing node;

[0037] A power module, calculating the optimal power of each of the unloading paths according to the lower limit of the ergodic confidentiality rate;

[0038] An execution module is configured to offload the computing task of the meta-computing user to the meta-computing node according to the offloading path and the optimal power.

[0039] Optionally, the formulas for the main channel capacity and the wiretap channel capacity are respectively:

[0040]

[0041] Where I is the number of metacomputing users, J is the number of metacomputing nodes, and H i,j represents the channel matrix between the ith meta-computing user and the jth meta-computing node, H i,e represents the channel matrix between the i-th element-calculating user and the eavesdropper,

[0042] C i,j and C i,e They represent the main channel capacity and the eavesdropping channel capacity respectively, P i represents the transmit power of the sender, where Represents the meta-computation node O j The noise power at the receiving end, B represents the transmission bandwidth, log2(.) represents the logarithm operation with base 2, det(.) represents the matrix determinant, represents the Rayleigh quotient, and the admissible detection vector of the eavesdropper is defined as the solution to the Rayleigh quotient maximization problem, where represents the noise power at the eavesdropper’s receiving end, is the identity matrix, H1 and g i,e is the equivalent channel matrix, N i represents the number of antennas of the user for the calculation, represent The maximum eigenvalue of , (.) H is the conjugate transpose operation.

[0043] Optionally, the confidentiality rate module calculates the confidentiality rate of the transmitted signal according to the main channel capacity and the wiretap channel capacity, and the expression is as follows:

[0044] R i,j =[C i,j -C i,e ] +

[0045] Where [x] + =max(x,0).

[0046] Optionally, the graph module constructs a bipartite graph of the meta-computing user and the meta-computing node according to the power consumption and the delay, including:

[0047] The meta-computing nodes whose computing power is lower than the preset threshold are deleted, and the connection weight between each meta-computing user and the meta-computing node is calculated based on power consumption and latency.

[0048] Optionally, before the execution module offloads the computing task of the meta-computing user to the meta-computing node, the execution module further includes:

[0049] The computing tasks are segmented and encoded.

[0050] Optionally, the power module calculates the optimal power of each of the unloading paths according to the traversal confidentiality rate lower limit, including:

[0051] The minimum power consumption point that meets the delay constraint is found within a given RF power range through numerical methods.

[0052] Optionally, the construction of the white noise includes:

[0053] Calculating a singular value decomposition of the task signal null space;

[0054] The white noise is obtained by selecting columns corresponding to zero singular values ​​as a generator matrix of the white noise.

[0055] The present application also provides an energy-saving offloading device for a meta-computing task, comprising:

[0056] Memory;

[0057] A processor, for storing a computer executable program of the above-mentioned method for energy-saving offloading of a meta-computing task, executing: obtaining a sending signal by adding white noise to a task signal of a meta-computing user sending a computing task; calculating the main channel capacity and the eavesdropping channel capacity according to the sending signal received by the main channel and the eavesdropping channel respectively; calculating the confidentiality rate of the sending signal according to the main channel capacity and the eavesdropping channel capacity, and determining the ergodic confidentiality rate and the ergodic confidentiality rate lower limit according to the confidentiality rate; calculating the power consumption and delay between multiple meta-computing users and multiple meta-computing nodes under maximum radio frequency power according to the ergodic confidentiality rate; constructing a bipartite graph of the meta-computing users and the meta-computing nodes according to the power consumption and the delay; determining an offloading path for the meta-computing user to offload the computing task to the meta-computing node according to the bipartite graph; calculating the optimal power of each of the offloading paths according to the ergodic confidentiality rate lower limit; and offloading the computing task of the meta-computing user to the meta-computing node according to the offloading path and the optimal power.

[0058] The present application also provides a storage medium, comprising: storing a computer executable program, wherein the computer executable program is used to be called by a processor to execute the steps of the energy-saving unloading method of the above-mentioned meta-computing task.

[0059] The beneficial effects of this application are:

[0060] The present application provides an energy-saving unloading method for a meta-computing task, comprising: obtaining a transmission signal by adding white noise to a task signal of a meta-computing user sending a computing task; calculating the main channel capacity and the eavesdropping channel capacity according to the transmission signal received by the main channel and the eavesdropping channel respectively; calculating the confidentiality rate of the transmission signal according to the main channel capacity and the eavesdropping channel capacity, and determining the ergodic confidentiality rate and the ergodic confidentiality rate lower limit by the confidentiality rate; calculating the power consumption and delay between multiple meta-computing users and multiple meta-computing nodes under maximum radio frequency power by the ergodic confidentiality rate; constructing a bipartite graph of the meta-computing user and the meta-computing node according to the power consumption and the delay; determining the unloading path for the meta-computing user to unload the computing task to the meta-computing node according to the bipartite graph; calculating the optimal power of each unloading path according to the ergodic confidentiality rate lower limit; and unloading the computing task of the meta-computing user to the meta-computing node according to the unloading path and the optimal power. The present application forms a transmission signal by combining a task signal and white noise, and calculates the confidentiality rate by using information of the main channel and the eavesdropping channel. Then, the traversal confidentiality rate and the traversal confidentiality rate lower limit are determined according to the confidentiality rate, and the power consumption and delay between multiple meta-computing users and multiple meta-computing nodes are calculated accordingly. Finally, by constructing a bipartite graph, determining the offloading path, and calculating the optimal power, the safe, efficient, and energy-saving offloading of computing tasks is achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 It is a schematic diagram of the energy-saving offloading process of the meta-computing task in this application;

[0062] Figure 2 This is a schematic diagram of the energy-saving and secure meta-computing architecture in this application;

[0063] Figure 3 It is a schematic diagram of the bipartite graph algorithm (KM algorithm) in this application. DETAILED DESCRIPTION

[0064] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that various forms of implementing the present disclosure should not be limited by the embodiments set forth herein. On the contrary, the embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0065] Please refer to Figure 1 As shown, the present application provides an energy-saving offloading method for a meta-computing task, comprising:

[0066] S101. Obtain a sending signal by adding white noise to a task signal of a meta-computing user sending a computing task.

[0067] Metacomputing is a technology that uses distributed computing resources in a network to enhance processing capabilities and improve overall system performance. In the Energy Internet of Things, because a large network of terminal nodes such as sensors, smart meters, and distributed energy are involved, a large amount of data is generated and processed, but the terminal nodes have limited computing power. By offloading computationally intensive tasks to more powerful nodes in the network (i.e., metacomputing nodes MC N), the Energy Internet of Things system can reduce latency, reduce energy consumption of terminal nodes, and improve processing efficiency.

[0068] In the energy-saving and secure meta-computing method of the energy Internet of Things based on the physical layer security solution, the computing tasks of the meta-computing users need to be executed, but are limited by their own computing capabilities, so the task data is uploaded to the meta-computing nodes. After receiving the tasks, the meta-computing nodes perform the calculations and feed back the results to the meta-computing users. This process is called computation offloading.

[0069] like Figure 2 As shown in the figure, a secure meta-computing system includes I energy IoT nodes, namely, meta-computing users (MCUs), represented by {U1, U2, ..., U I}. A set of J meta-computing nodes (MCNs), denoted as {O1, O2, ..., O J}, providing computing resources for MCU. j With M j A receiving antenna.

[0070] In the present application, a passive N e Antenna eavesdropper (eavesdropper) can eavesdrop on information in all frequency spectrums.

[0071] I of the MCUs unload their computing tasks, represented by {Z1, Z2, ..., Z I}, and sent to MCN through wireless channel. After receiving the task, MCN executes the task and feeds back the result to MCU.

[0072] Based on the above-mentioned meta-computing system settings, if it is assumed that all channels obey Rayleigh fading (because the signal propagates through multiple paths and the field strength at the receiving point comes from different propagation paths, the delay time of each path is different, and the superposition of component waves in each direction produces standing wave field strength, thus forming fast signal fading.), then U i The eavesdropping channel to the eavesdropper is defined as Will U i to j The communication link is defined as

[0073] The instantaneous CSI (channel state information, i.e. H i,j ) can be obtained through channel estimation, and the eavesdropper’s instantaneous CSI H i,e Assuming that the MCU uses orthogonal spectrum to communicate with the MCN, there is no multi-user interference.

[0074] For safety reasons, the MCU's U i Transmit confidential content to MCN's O j At the same time, U i Generate an artificial noise (AN) signal to confuse the eavesdropper. In this case, O j The signal received by the eavesdropper can be expressed as:

[0075] y i,j =α i,j H i,j (b i x i +G i v i )+n i,j

[0076] y e,k =α i,e H i,e (b i x i +G i v i )+n i,e

[0077] where α i,j For U i and O j The path loss between e,i For U i and the path loss between the eavesdropper, x i For U i to j The information-carrying signal between them satisfies E(|x i | 2 )=P k / N i .

[0078] It should be noted that P k is normalized by the bandwidth B. i YesN i ×1 beamforming vector. n i,j and n i,e Obedience and Additive White Gaussian Noise (AWGN) vector. AN signal space G i It is located in Hi,j N in the null space of i ×(N i -1) matrix. AN signal v i Is an obedience The complex Gaussian AN signal.

[0079] S102: Calculate the main channel capacity and the wiretap channel capacity according to the transmission signals received by the main channel and the wiretap channel respectively.

[0080] Specifically, b i and G i The design is as follows:

[0081] First, U i Use b i As the information-carrying beamforming vector, diversity gain of the MIMO system is achieved. Specifically, U i implement The eigenvalue decomposition of is:

[0082]

[0083] U1 contains eigenvectors of , Λ is a diagonal matrix containing All eigenvalues ​​of is the maximum eigenvalue, and its corresponding eigenvector is b i .

[0084] In addition, O j Use b i Collect information-bearing signals from different antennas, i.e. It should be noted that b i With O j is synchronous, or O j Through the above U i implement The eigenvalue decomposition method is used independently by H i,j Calculate b i .

[0085] U i is the AN signal v i Generate N i ×(N i -1) Matrix G i , it is H i,j The null space, that is, null(H i,j ).

[0086] null(H i,j ) function can calculate H i,jThe singular value decomposition (SVD) of [U2,S,V H ]=SVD(H i,j ). The columns in V corresponding to 0 singular values ​​form a set of standard orthogonal basis vectors, i.e., the null space. i,j G i v i =0, the AN signal can be eliminated during the signal propagation process of the wireless channel.

[0087] By b i Processing i,j Afterwards, O j The processed signal at is expressed as:

[0088]

[0089] Among them, n i,j Is has In this case, the formulas for the main channel capacity and the eavesdropping channel capacity are:

[0090]

[0091]

[0092] There are I meta-computing users and J meta-computing nodes. i,j Represents the ith meta-computing user U i With the jth meta-computation node O j The channel matrix, H i,e Represents U i and the eavesdropper’s channel matrix,

[0093] C i,j and C i,e They represent the main channel capacity and the eavesdropping channel capacity respectively, P i represents the transmit power of the sender, where Represents the meta-computation node O j The noise power at the receiving end, B represents the transmission bandwidth. log2(.) represents the logarithm operation with base 2. det(.) represents the matrix determinant. The admissible detection vector of the eavesdropper is defined as the solution to the Rayleigh quotient maximization problem. represents the Rayleigh quotient, where represents the noise power at the eavesdropper’s receiving end, is the identity matrix, H1 and g i,e is the equivalent channel matrix. N i Indicates the number of antennas used by the user. represent The largest eigenvalue of . (.)H is the conjugate transpose operation.

[0094] S103: Calculate the confidentiality rate of the transmitted signal according to the main channel capacity and the wiretap channel capacity, and determine the ergodic confidentiality rate and the ergodic confidentiality rate lower limit according to the confidentiality rate.

[0095] U i With the receiver R i,j The confidentiality rate between is expressed as:

[0096] R i,j =[C i,j -C i,e ] +

[0097] Where [x] + =max(x,0).

[0098] The confidentiality rate can be viewed as an inherent property of a given instantaneous CSI implementation.

[0099] The ergodic confidentiality rate is the average confidentiality rate of all CSIs. The expression of the ergodic confidentiality rate can be used to derive the security performance index of computing resource allocation.

[0100] Specifically, equation R i,j =[C i,j -C i,e ] + The equality holds if and only if {C i,j -C i,e} is always non-negative in all channel states. However, due to the lack of H i,e , so it is impossible to determine whether the instantaneous confidentiality rate is a non-negative value, so it is necessary to derive the lower limit of the true ergodic confidentiality rate as the ergodic confidentiality rate.

[0101] Due to C i,j Greater than C i,e The probability is very high, in this case R i,j The ergodic confidentiality rate for:

[0102]

[0103] in:

[0104]

[0105] m=max(N i ,M j ) and n=min(N i ,M j ), max is the maximum value between the two, and min is the minimum value between the two.

[0106] The elements in the ath row and bth column of the (n×n) real matrix Ω(x) are defined as:

[0107] [Ω(x)] a,b =Γ(m-n+a+b-1,x)

[0108] where a, b = 1, …, n, and Γ(·, ·) is the upper incomplete gamma function, defined as:

[0109]

[0110] H1=H i，e , G i

[0111] H2=[g i,e ,H1]

[0112]

[0113] v = min(a, b), μ = max(a, b), for any E τ (z) is an exponential integral of order τ, defined as:

[0114]

[0115] where Re(z) is the real part of z.

[0116] Although except There is no integral expression outside of a few special functions, but it is a complex equation that incurs a lot of computational overhead.

[0117] The present application provides high SNR and large transmit antenna area The closed-form expression for the lower bound is a common metric in security because it represents the worst case, that is, when P i and N i When the value of is large, The lower limit It can be expressed as:

[0118]

[0119] S104. Calculate the power consumption and delay between the plurality of meta-computing users and the plurality of meta-computing nodes under the maximum radio frequency power through the ergodic confidentiality rate.

[0120] Meta-computing tasks are uploaded to meta-computing node O j , O j After receiving the task data, perform the meta-computation user U iThe task of the task has a calculation period of c j D i / f j , where c j (CCN / bit) is 0 j The number of CPU cycles per bit processed (CPU cycle number, CCN) indicates the computing efficiency of the CPU chip, f j (CCN / s) is O j CCN processed per second.

[0121] O j The energy consumption per second is (joules / second), where η j O j The computing energy efficiency coefficient of the CPU chip in the upload can be expressed as D i / (R i,j B), where R i,j (bit / s) is U i and O j is the confidentiality rate of the wireless channel, and B is the bandwidth.

[0122] With the help of O j , Z i The energy consumption can be expressed as:

[0123]

[0124] Among them, P0 and P i Represent static power consumption and RF power consumption respectively.

[0125] Furthermore, in addition to energy consumption issues, meta-computing also focuses on computing latency issues.

[0126] Z i In O j The auxiliary delay model formula can be expressed as:

[0127]

[0128] Where max(x,y) is the maximum value of x and y. and are the calculation delay and transmission delay respectively.

[0129] Taking latency into account, the computation offloading problem is formulated as an energy consumption minimization problem:

[0130]

[0131] The constraint T i,j ≤∈ requires that the delay time of all tasks should be less than the threshold ∈. i,j={0 or 1} is the i-th row and j-th column of the matrix W, representing the binary constraint of the matching factor. i,j =1 means U i The task is carried out by O j Processing. Constrained by It can be seen that each MCU can only be assigned to one MCN, and each MCN can only be assigned to one MCU. j ≤P max The RF power constraint of the MCU.

[0132] It can be seen that in the optimization, the RF power consumption P i , and W are coupling variables, and P1 is a non-convex mixed integer problem. To solve this problem, the present application transforms P1 into two sub-problems: computing resource allocation and power allocation.

[0133] Since the coherence time is much shorter than the computation offloading duration, the confidentiality rate R i,j is time-varying. In this case, the ergodic secrecy rate is used As a measure of allocating computing resources, use To calculate E i,j Instead of R i,j .

[0134] Then, use E i,j To optimize the computing resource allocation matrix W. In the power process, use Find the best power.

[0135] Computing resource allocation:

[0136] MCNs are sorted by computing power, and then the weaker MCNs are removed during the computing resource allocation process. For a given maximum RF power P max , calculate the energy consumption E' of each pair of MCU and MCN i,j and delay T' i,j , as shown below,

[0137]

[0138] S105. Construct a bipartite graph of the meta-computing user and the meta-computing node according to the power consumption and the delay.

[0139] like Figure 3 As shown, a bipartite graph algorithm is generated to record all MCU-MCN combinations to satisfy the constraint T i,j ≤∈, T' of any MCU-MCN pair i,j >∈ should not be recorded in the bipartite graph. Therefore, if T' i,j >∈, then E' i,jSet to ∞.

[0140] S106. Determine, according to the bipartite graph, an offloading path for the meta-computing user to offload the computing task to the meta-computing node.

[0141] This application sets an I×I matching matrix W, whose (i,j)th element represents an edge, for example W i,j =1 means vertices i and j are connected. For another edge W ii,jj , we have i≠ii and j≠jj to satisfy the constraint that any two edges do not share vertices. With W, the computing resource allocation problem can be equivalently transformed into a two-dimensional matching problem as follows:

[0142]

[0143] Obviously, W i,j is a binary variable. When O j Assigned to U i When it is 1, otherwise it is 0. i,j = {0 or 1} indicates that each MCU can only be assigned to one MCN, and each MCN can only be assigned to one MCU. Optimization is to find the optimal W to minimize P2 is a convex function, and the optimal W can be solved by the bipartite graph algorithm.

[0144] S107: Calculate the optimal power of each of the offloading paths according to the traversal confidentiality rate lower limit.

[0145] Increasing RF power can reduce the overall latency of computation offloading, but this will result in a lot of energy consumption. If the RF power usage is reduced too much, it may cause the latency constraint to be not met. Therefore, optimizing RF power is a key issue.

[0146] The calculation is too complicated, resulting in high computational overhead of the power optimization algorithm. Therefore, this application adopts an approximate expression, namely Optimize for power and then use The exact expression is verified.

[0147] For a given W, we can Replace with Convert P1 to P3 as follows,

[0148]

[0149] P i ≤P max .

[0150] In fact, Formula P3 focuses on power optimization for a single user.

[0151] In order to extend it to multi-user scenarios, this application uses a bipartite graph algorithm to assign multiple MCUs (mobile computing units) to MCNs (mobile computing nodes). i Its optimal MCN O has been assigned by the bipartite graph algorithm j Based on this allocation, power optimization is performed individually for each user, resulting in the best power allocation for all MCUs.

[0152] This application uses the following theorem to determine the feasible region of RF power in P3.

[0153] Specifically, when and P min ≤P max When U i The optimum power exist:

[0154]

[0155] Therefore, the algorithm aims to determine the optimal RF power allocation for multiple MCUs in a wireless network. The goal is to minimize the total power consumption while meeting the latency constraints for data transmission and computation.

[0156] Verification: The feasibility of the solution is first checked by computing the delay constraints, and then numerical methods are used to verify the feasibility of the solution in [P min ,P max ] found in Finally, it is verified whether the solution meets the delay constraint. If the delay constraint is not met, the total RF power is used to transmit the message so that the computation task can be completed as soon as possible. In this algorithm, Matlab CVX can be used as a convex optimization tool with golden section search.

[0157] S108. Offload the computing task of the meta-computing user to the meta-computing node according to the offloading path and the optimal power.

[0158] The offloading path refers to the matching relationship between the meta-computing user (MCU) and the meta-computing node (MCN), that is, which MCU's computing tasks should be assigned to which MCN for processing. This matching relationship is determined by the bipartite graph algorithm mentioned above. The bipartite graph algorithm can find the most suitable MCN for each MCU based on multiple factors such as computing resources, computing power, and delay constraints.

[0159] Optimal power refers to the radio frequency (RF) power allocation scheme that can minimize total power consumption while meeting the latency constraints of data transmission and computing. This optimal power is obtained through a series of complex calculations and optimization processes, which ensures that the meta-computing tasks can be completed with the lowest energy consumption during offloading.

[0160] After the offloading path and optimal power are determined, the computing tasks of the meta-computing users can be offloaded to the designated meta-computing nodes for processing according to these parameters. This process achieves the effective utilization of computing resources and minimizes energy consumption, and is an important part of the secure computing offloading solution in the energy Internet of Things.

[0161] The present application also provides an energy-saving unloading device for a meta-computing task, comprising:

[0162] The signal module obtains the sending signal by adding white noise to the task signal of the meta-computing user sending the computing task;

[0163] A capacity module, calculating the main channel capacity and the wiretap channel capacity according to the transmission signals received by the main channel and the wiretap channel respectively;

[0164] A confidentiality rate module, which calculates the confidentiality rate of the transmitted signal according to the capacity of the main channel and the capacity of the wiretap channel, and determines the ergodic confidentiality rate and the lower limit of the ergodic confidentiality rate according to the confidentiality rate;

[0165] A calculation module, which calculates the power consumption and delay between the plurality of said meta-computing users and the plurality of meta-computing nodes under the maximum radio frequency power by using the ergodic confidentiality rate;

[0166] A graph module, constructing a bipartite graph of the meta-computing user and the meta-computing node according to the power consumption and the delay;

[0167] A path module, which determines, according to the bipartite graph, an offloading path for the meta-computing user to offload the computing task to the meta-computing node;

[0168] A power module, calculating the optimal power of each of the unloading paths according to the ergodic confidentiality rate lower limit;

[0169] An execution module is configured to offload the computing task of the meta-computing user to the meta-computing node according to the offloading path and the optimal power.

[0170] Furthermore, the formulas for the main channel capacity and the wiretap channel capacity are:

[0171]

[0172] There are I meta-computing users and J meta-computing nodes. i,j Represents the ith meta-computing user U i With the jth meta-computation node O j The channel matrix, H i,e Represents U i and the eavesdropper’s channel matrix,

[0173] C i,j and C i,e They represent the main channel capacity and the eavesdropping channel capacity respectively, P i represents the transmit power of the sender, where Represents the meta-computation node O j The noise power at the receiving end, B represents the transmission bandwidth. log2(.) represents the logarithm operation with base 2. det(.) represents the matrix determinant. The admissible detection vector of the eavesdropper is defined as the solution to the Rayleigh quotient maximization problem. represents the Rayleigh quotient, where represents the noise power at the eavesdropper’s receiving end, is the identity matrix, H1 and g i,e is the equivalent channel matrix. N i Indicates the number of antennas used by the user. represent The largest eigenvalue of . (.) H is the conjugate transpose operation.

[0174] Furthermore, the confidentiality rate module calculates the confidentiality rate of the transmitted signal according to the main channel capacity and the wiretap channel capacity, and the expression is as follows:

[0175] R i,j =[C i,j -C i,e ] +

[0176] Where [x] + =max(x,0).

[0177] Further, the graph module constructs a bipartite graph of the meta-computing user and the meta-computing node according to the power consumption and the delay, including:

[0178] The meta-computing nodes whose computing power is lower than the preset threshold are deleted, and the connection weight between each meta-computing user and the meta-computing node is calculated based on power consumption and latency.

[0179] Furthermore, before the execution module offloads the computing task of the meta-computing user to the meta-computing node, the execution module further includes:

[0180] The computing tasks are segmented and encoded.

[0181] Furthermore, the power module calculates the optimal power of each of the unloading paths according to the lower limit of the traversal confidentiality rate, including:

[0182] The minimum power consumption point that meets the delay constraint is found within a given RF power range through numerical methods.

[0183] Furthermore, the construction of the white noise includes:

[0184] Calculating a singular value decomposition of the task signal null space;

[0185] The white noise is obtained by selecting columns corresponding to zero singular values ​​as a generator matrix of the white noise.

[0186] The present application also provides an energy-saving offloading device for a meta-computing task, comprising:

[0187] Memory;

[0188] A processor, for storing a computer executable program of the above-mentioned method for energy-saving offloading of a meta-computing task, executing: obtaining a sending signal by adding white noise to a task signal of a meta-computing user sending a computing task; calculating the main channel capacity and the eavesdropping channel capacity according to the sending signal received by the main channel and the eavesdropping channel respectively; calculating the confidentiality rate of the sending signal according to the main channel capacity and the eavesdropping channel capacity, and determining the ergodic confidentiality rate and the ergodic confidentiality rate lower limit according to the confidentiality rate; calculating the power consumption and delay between multiple meta-computing users and multiple meta-computing nodes under maximum radio frequency power according to the ergodic confidentiality rate; constructing a bipartite graph of the meta-computing users and the meta-computing nodes according to the power consumption and the delay; determining an offloading path for the meta-computing user to offload the computing task to the meta-computing node according to the bipartite graph; calculating the optimal power of each of the offloading paths according to the ergodic confidentiality rate lower limit; and offloading the computing task of the meta-computing user to the meta-computing node according to the offloading path and the optimal power.

[0189] The present application also provides a storage medium, comprising: storing a computer executable program, wherein the computer executable program is used to be called by a processor to execute the steps of the energy-saving unloading method of the above-mentioned meta-computing task.

Claims

1. A method for energy-saving offloading of meta-computing tasks, characterized in that: include: The sending signal is obtained by adding white noise to the task signal of the meta-computing user sending the computing task; Calculating the main channel capacity and the wiretap channel capacity according to the transmission signals received by the main channel and the wiretap channel respectively; Calculating the confidentiality rate of the transmitted signal according to the main channel capacity and the wiretap channel capacity, and determining the ergodic confidentiality rate and the ergodic confidentiality rate lower limit according to the confidentiality rate; Calculate the power consumption and delay between the plurality of said meta-computing users and the plurality of meta-computing nodes under the maximum radio frequency power by using the ergodic confidentiality rate; constructing a bipartite graph of the meta-computing user and the meta-computing node according to the power consumption and the delay; Determining, according to the bipartite graph, an offloading path for the meta-computing user to offload the computing task to the meta-computing node; Calculating the optimal power of each of the offloading paths according to the traversal confidentiality rate lower limit; The computing task of the meta-computing user is offloaded to the meta-computing node according to the offloading path and the optimal power.

2. The energy-saving offloading method of a meta-computing task according to claim 1, characterized in that: The formulas for the main channel capacity and the eavesdropping channel capacity are: Where I is the number of metacomputing users, J is the number of metacomputing nodes, and H i,j represents the channel matrix between the ith meta-computing user and the jth meta-computing node, H i,e represents the channel matrix between the i-th element-calculating user and the eavesdropper, C i,j and C i,e They represent the main channel capacity and the eavesdropping channel capacity respectively, P i represents the transmit power of the sender, where Represents the meta-computation node O j The noise power at the receiving end, B represents the transmission bandwidth, log2(.) represents the logarithm operation with base 2, det(.) represents the matrix determinant, represents the Rayleigh quotient, and the admissible detection vector of the eavesdropper is defined as the solution to the Rayleigh quotient maximization problem, where represents the noise power at the eavesdropper’s receiving end, is the identity matrix, H1 and g i,e is the equivalent channel matrix, N i represents the number of antennas of the user for the calculation, represent The maximum eigenvalue of , (.) H is the conjugate transpose operation.

3. The energy-saving offloading method of a meta-computing task according to claim 1, characterized in that: The confidentiality rate of the transmitted signal is calculated according to the main channel capacity and the wiretap channel capacity, and the expression is as follows: R i,j =[C i,j -C i,e ] + Where [x] + =max(x,0).

4. The energy-saving offloading method of a meta-computing task according to claim 1, characterized in that: Constructing a bipartite graph of the meta-computing user and the meta-computing node according to the power consumption and the delay, including: The meta-computing nodes whose computing power is lower than the preset threshold are deleted, and the connection weight between each meta-computing user and the meta-computing node is calculated based on power consumption and delay.

5. The energy-saving offloading method of a meta-computing task according to claim 1, characterized in that: Before offloading the computing task of the meta-computing user to the meta-computing node, the method further includes: The computing tasks are segmented and encoded.

6. The energy-saving offloading method of a meta-computing task according to claim 1, characterized in that: According to the traversal confidentiality rate lower limit, the optimal power of each of the offloading paths is calculated, including: The minimum power consumption point that meets the delay constraint is found within a given RF power range through numerical methods.

7. The energy-saving offloading method of a meta-computing task according to claim 1, characterized in that: The construction of the white noise includes: Calculating a singular value decomposition of a null space in the task signal; The white noise is obtained by selecting columns corresponding to zero singular values ​​as a generator matrix of the white noise.

8. An energy-saving unloading device for meta-computing tasks, characterized in that: include: The signal module obtains the sending signal by adding white noise to the task signal of the meta-computing user sending the computing task; A capacity module, calculating the main channel capacity and the wiretap channel capacity according to the transmission signals received by the main channel and the wiretap channel respectively; A confidentiality rate module, which calculates the confidentiality rate of the transmitted signal according to the capacity of the main channel and the capacity of the wiretap channel, and determines the ergodic confidentiality rate and the lower limit of the ergodic confidentiality rate according to the confidentiality rate; A calculation module, which calculates the power consumption and delay between the plurality of said meta-computing users and the plurality of meta-computing nodes under the maximum radio frequency power by using the ergodic confidentiality rate; A graph module, constructing a bipartite graph of the meta-computing user and the meta-computing node according to the power consumption and the delay; A path module, which determines, according to the bipartite graph, an offloading path for the meta-computing user to offload the computing task to the meta-computing node; A power module, calculating the optimal power of each of the unloading paths according to the lower limit of the ergodic confidentiality rate; An execution module is configured to offload the computing task of the meta-computing user to the meta-computing node according to the offloading path and the optimal power.

9. According to the energy-saving unloading device for meta-computing tasks of claim 8, the formulas of the main channel capacity and the eavesdropping channel capacity are respectively: in, I is the number of metacomputing users, J is the number of metacomputing nodes, H i,j represents the channel matrix between the ith meta-computing user and the jth meta-computing node, H i,e represents the channel matrix between the i-th element-calculating user and the eavesdropper, C i,j and C i,e They represent the main channel capacity and the eavesdropping channel capacity respectively, P i represents the transmit power of the sender, where Represents the meta-computation node O j The noise power at the receiving end, B represents the transmission bandwidth, log2(.) represents the logarithm operation with base 2, det(.) represents the matrix determinant, represents the Rayleigh quotient, and the admissible detection vector of the eavesdropper is defined as the solution to the Rayleigh quotient maximization problem, where represents the noise power at the eavesdropper’s receiving end, is the identity matrix, H1 and g i,e is the equivalent channel matrix, N i represents the number of antennas of the user for the calculation, represent The maximum eigenvalue of , (.) H is the conjugate transpose operation.

10. The energy-saving unloading device for meta-computing tasks according to claim 8, characterized in that: The confidentiality rate module calculates the confidentiality rate of the transmitted signal according to the main channel capacity and the wiretap channel capacity, and the expression is as follows: R i,j =[C i,j -C i,e ] + Where [x] + =max(x,0).

11. The energy-saving unloading device for meta-computing tasks according to claim 8, characterized in that: The graph module constructs a bipartite graph of the meta-computing user and the meta-computing node according to the power consumption and the delay, including: The meta-computing nodes whose computing power is lower than the preset threshold are deleted, and the connection weight between each meta-computing user and the meta-computing node is calculated based on power consumption and latency.

12. The energy-saving unloading device for meta-computing tasks according to claim 8, characterized in that: Before the execution module offloads the computing task of the meta-computing user to the meta-computing node, the execution module further includes: The computing tasks are segmented and encoded.

13. The energy-saving unloading device for meta-computing tasks according to claim 8, characterized in that: The power module calculates the optimal power of each of the unloading paths according to the traversal confidentiality rate lower limit, including: The minimum power consumption point that meets the delay constraint is found within a given RF power range through numerical methods.

14. The energy-saving unloading device for meta-computing tasks according to claim 8, characterized in that: The construction of the white noise includes: Calculating a singular value decomposition of the task signal null space; The white noise is obtained by selecting columns corresponding to zero singular values ​​as a generator matrix of the white noise.

15. An energy-saving offloading device for meta-computing tasks, characterized in that: include: Memory; A processor, used to store a computer executable program of a method for energy-saving offloading of a meta-computing task according to any one of claims 1 to 7, and to execute: adding a sending signal formed by white noise to a task signal of a meta-computing user sending a computing task; calculating the main channel capacity and the wiretap channel capacity according to the sending signal received by the main channel and the wiretap channel; Calculating the confidentiality rate of the transmitted signal according to the main channel capacity and the wiretap channel capacity, and determining the ergodic confidentiality rate and the ergodic confidentiality rate lower limit according to the confidentiality rate; Calculating the power consumption and delay between the plurality of the meta-computing users and the plurality of the meta-computing nodes under the maximum radio frequency power by using the traversal confidentiality rate; constructing a bipartite graph of the meta-computing users and the meta-computing nodes according to the power consumption and the delay; Determining, according to the bipartite graph, an offloading path for the meta-computing user to offload the computing task to the meta-computing node; Calculating the optimal power of each of the offloading paths according to the traversal confidentiality rate lower limit; The computing task of the meta-computing user is offloaded to the meta-computing node according to the offloading path and the optimal power.

16. A storage medium, characterized in that: include: A computer executable program is stored, and the computer executable program is used to be called by a processor to execute the steps of the energy-saving unloading method of a meta-computing task as described in any one of claims 1 to 7.

Images