Double-layer task unloading optimization method and system in marine MEC network

Abstract

The invention discloses a double-layer task unloading optimization method and system in a marine MEC network, and belongs to the technical field of communication. The invention discloses a double-layer task offloading optimization method in a marine MEC network. The method comprises the following steps: constructing a marine MEC network double-layer offloading model comprising a USN, an AUV, an SN and a UAV; on the basis of a marine MEC network double-layer unloading model, a joint optimization problem is constructed with the purpose of minimizing the total completion time delay of USN data processing, and optimization variables are an unloading strategy of a USN, the position of an AUV and an unloading strategy of an SN; a joint optimization problem is decomposed into three optimization sub-problems of an unloading strategy of a USN, a position of an AUV and an unloading strategy of an SN, a gradient descent method, a particle swarm algorithm and CVX are respectively used to obtain solutions of the optimization sub-problems, and a solution of the joint optimization problem is obtained through alternate iteration. According to the method, the total time delay of task completion can be remarkably reduced, and better experience is brought to a user.

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and in particular relates to a method and system for optimizing double-layer task offloading in a marine MEC network. Background Art

[0002] As an extension of terrestrial networks, ocean networks connect different types of ocean sensing devices, collect ocean data, and then upload this data to ocean observation systems or cloud platforms for use in various applications and services. With the continued development of marine resources, new applications such as seabed exploration, disaster warning, and ship navigation have posed new challenges to the real-time performance and computing efficiency of marine communications. By introducing Mobile Edge Computing (MEC) technology into the marine environment and sinking communication and computing power to the edge of the marine network, a new approach is being proposed to address these issues. By offloading computing tasks from sensor devices to nearby edge devices, such as drones, unmanned surface vessels, and unmanned submersibles, marine edge computing can effectively reduce latency in data processing and analysis, improving the overall performance and responsiveness of marine communication networks.

[0003] Existing research on edge computing in marine networks has largely focused on optimizing single scenarios, such as autonomous underwater vehicle (AUV) path planning or unmanned aerial vehicle (UAV) trajectory design, lacking systematic research on multi-level task offloading strategies. To improve the computing efficiency of marine networks in hybrid underwater acoustic and radio frequency transmission environments, it is necessary to study the task allocation and resource scheduling mechanisms of underwater sensor nodes (USNs) to minimize the total latency of sensor data processing. Summary of the Invention

[0004] In view of the deficiencies in the prior art, the purpose of the present invention is to provide a method and system for optimizing dual-layer task offloading in marine MEC networks, which solves the problems in the prior art.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] A method for optimizing dual-layer task offloading in a marine MEC network includes the following steps:

[0007] Construct a dual-layer offloading model for marine MEC networks including USN, AUV, SN, and UAV;

[0008] Based on the dual-layer offloading model of the marine MEC network, a joint optimization problem is constructed with the goal of minimizing the total delay of USN data processing. The optimization variables are the offloading strategy of the USN, the position of the AUV, and the offloading strategy of the SN.

[0009] The joint optimization problem is decomposed into three optimization sub-problems: the unloading strategy of USN, the position of AUV and the unloading strategy of SN. The solutions of each optimization sub-problem are obtained by gradient descent method, particle swarm algorithm and CVX respectively, and the solution of the joint optimization problem is obtained through alternating iteration.

[0010] Furthermore, the USN offloads its tasks to the AUV and SN respectively in the NOMA manner. The AUV and SN receive the data uploaded by the USN through underwater acoustic signals and process the computing tasks. The SN continues to offload part of its data to the UAV in the FDMA manner. The UAV receives the data from the SN through radio frequency signals and processes it.

[0011] Furthermore, the total delay for completing USN data processing is t m Expressed as:

[0012]

[0013] in, is the transmission delay, To handle delay, for USN m Data transmission to SN n The delay, Calculate the latency locally. SN n Transmission delay of uploading data to UAV, is the computational delay of UAV; C m for USN m The total number of CPUs required for the task, f auv For AUV i Computational resources, R m for USN m Total number of bits in the task, β m for USN m To SN n Uninstall rate, r m,i for USN m Upload data to AUV i The transmission rate, r m,n for USN m Upload data to SN n The transmission rate, f sn SN n Computational resources, α m SN nUnloading rate to UAV; r n,uav SN n Transmission rate between UAV; f uav It is the computing resource of UAV.

[0014] Furthermore, the joint optimization problem is:

[0015]

[0016] stC1:0≤β m ≤1,m∈M

[0017] C2:0≤α m ≤1,m∈M

[0018] C3:0≤p m ≤p max ,m∈M

[0019] C4: 0≤x i ,y i ≤100

[0020] Among them, C1-C4 are the constraints of the optimization problem P1, I represents the AUV cluster, N represents the SN cluster, and x i For AUV i x coordinate, y i For AUV i The y coordinate of p max is the maximum transmission power of USN, p m for USN m The transmission power of , i represents the i-th AUV, m represents the m-th USN, n represents the n-th SN; M represents the underwater sensor cluster.

[0021] Furthermore, the USN unloading strategy sub-problem is expressed as:

[0022]

[0023] The SN offloading strategy sub-problem is expressed as:

[0024]

[0025] The AUV position optimization subproblem is expressed as:

[0026]

[0027] in, is a set of positive integers.

[0028] Furthermore, the steps of solving the USN unloading strategy sub-problem using the gradient descent method include:

[0029] S311, use smooth approximation to process the max function contained in the objective function of problem P3, the inner function InnerSmooth after the approximation of the P3 objective function m for:

[0030]

[0031] Among them, τ is the control parameter.

[0032] Approximate outer function OuterSmooth m for:

[0033]

[0034] Combining the smoothing results of the inner and outer layers, the complete smoothing objective function is:

[0035]

[0036] S312, calculate the gradient ▽f(β (t) ), outer approximation function OuterSmooth m The partial derivative of is:

[0037]

[0038] in:

[0039]

[0040] Inner approximation function InnerSmooth m Partial derivatives of :

[0041]

[0042] in:

[0043]

[0044] Therefore, the objective function The partial derivative of is:

[0045]

[0046] S313, update the variable β according to the step size formula of the gradient descent method: (t+1) =β (t) -η·▽f(β (t) );

[0047] S314, after each update, m To project:

[0048] S315, substitute the current β (t+1) , calculate the smoothing objective function value f (t+1) :

[0049] S316, check convergence conditions: |f (t+1) -f (t) |<ò, ò is the convergence threshold, and the final value is returned if the convergence condition is met. As the optimization result, otherwise continue to iterate and update.

[0050] A dual-layer task offloading optimization system in a marine MEC network, comprising:

[0051] Model construction module: Construct a two-layer offloading model of the marine MEC network including USN, AUV, SN and UAV;

[0052] Optimization problem construction module: Based on the dual-layer offloading model of the marine MEC network, with the goal of minimizing the total delay of USN data processing, a joint optimization problem is constructed. The optimization variables are the USN offloading strategy, the AUV position, and the SN offloading strategy.

[0053] In addition, the optimization problem decomposition and solution module: the joint optimization problem is decomposed into three optimization sub-problems: the unloading strategy of USN, the position of AUV, and the unloading strategy of SN. The gradient descent method, particle swarm algorithm and CVX are used to obtain the solution of each optimization sub-problem respectively, and then the solution of the joint optimization problem is obtained through alternating iteration.

[0054] A computer storage medium stores a readable program, which, when executed by a processor, can execute the above-mentioned two-layer task offloading optimization method in an ocean MEC network.

[0055] An electronic device comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus;

[0056] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned dual-layer task offloading optimization method in the marine MEC network.

[0057] A computer program product includes computer instructions, which instruct a computing device to perform operations corresponding to the above-mentioned two-layer task offloading optimization method in an ocean MEC network.

[0058] Beneficial effects of the present invention:

[0059] 1. This paper proposes a joint optimization method for AUV position, USN unloading strategy, and Sea Surface Sink Node (SN) unloading strategy. The proposed method first discretizes the AUV position and uses a particle swarm optimization algorithm to find the optimal AUV position. Next, the max function of the USN unloading strategy subproblem is converted into a differentiable LSE function, and the optimal unloading strategy is obtained using gradient descent. A projection method is also used to ensure that the optimal solution falls within the feasible region. Finally, the SN unloading strategy subproblem is proven to be convex, and the optimal solution is obtained using the CVX tool.

[0060] 2. This paper combines the characteristics of ocean noise with the free space path loss model to construct a two-layer task offloading model that integrates underwater acoustic and radio frequency transmission, and analyzes the offloading process and communication delay of the two-layer task offloading model.

[0061] 3. This invention uses UAVs, SNs, and AUVs as edge servers to jointly process USN computing tasks. It also dynamically adjusts the offloading rate from USNs to SNs / AUVs and the secondary offloading rate from SNs to UAVs to minimize USN task completion latency. Simulation results show that this invention can effectively reduce the total USN task latency and improve the computing efficiency of marine networks in a hybrid underwater acoustic-RF transmission environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0063] Figure 1 Schematic diagram of the application model corresponding to the dual-layer task offloading optimization method in the marine MEC network of the present invention;

[0064] Figure 2 It is a simulation schematic diagram comparing the performance of the algorithm of the present invention with other algorithms;

[0065] Figure 3 Schematic diagram of the performance simulation of the dual-layer task offloading optimization method in the marine MEC network of the present invention under different task complexities;

[0066] Figure 4 This is a performance simulation diagram of the dual-layer task offloading optimization method in the marine MEC network of the present invention under different UAV computing capabilities. DETAILED DESCRIPTION

[0067] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments 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 making any creative efforts shall fall within the scope of protection of the present invention.

[0068] Example 1

[0069] Aiming at the dual-layer task offloading scenario of marine network, the present invention establishes a computational offloading model of marine network by hybrid NOMA and FDMA, and analyzes the total delay of USN from data transmission to processing completion. The particle swarm optimization algorithm, CVX tool and gradient descent method are used to optimize the position of AUV, the offloading rate of SN and the offloading rate of USN respectively.

[0070] A method for optimizing dual-layer task offloading in a marine MEC network includes the following steps:

[0071] S1, builds a two-layer offloading model for the marine MEC network, including: multiple underwater sensors (USNs), unmanned underwater vehicles (AUVs), surface nodes (SNs), and unmanned aerial vehicles (UAVs);

[0072] In the dual-layer task offloading scenario of the ocean network, the underwater sensor (USN) offloads its tasks to the unmanned underwater vehicle (AUV) and the surface node (SN) respectively using the NOMA method. The AUV and SN receive the data uploaded by the USN through underwater acoustic signals and process the computing tasks. The SN further offloads part of its data to the unmanned aerial vehicle (UAV) using the FDMA method. The UAV receives the data from the SN through radio frequency signals and processes it.

[0073] Specifically, if Figure 1 As shown in the figure, in a dual-layer task offloading scenario for a marine network, M underwater sensors are randomly scattered on the seafloor, each at an equal vertical distance from the sea surface. The underwater sensor cluster is denoted as M = {1,…m,…M}. Each underwater sensor network (USN) possesses a certain level of computing power. A single underwater vehicle (USV) is deployed on the surface as an edge computing node, assisting the smart sensors in processing computing tasks. The underwater sensor nodes offload data to the USV via an underwater acoustic channel. Frequency division multiple access (FDMA) communication is used between the underwater sensor cluster M and the USV to ensure orderly and efficient data transmission.

[0074] The offloading strategy of the underwater sensor cluster M is denoted as β = {β1,β2,…,β m ,…,β M}, where β m ∈{0,1},β m= 0 means that the mth underwater sensor performs local calculation, β m =1 means that the mth underwater sensor offloads the task to the USV for calculation. The data processing task of the mth user is recorded as (C m ,R m ,κ,T max ), where C m is the mth underwater sensor (USN m )The total number of CPUs required for the task, R m for USN m The total number of bits of the task, κ is the task complexity, C m and R m and κ satisfy the relationship C m =κR m , T max is the maximum delay that can be tolerated to complete the task. Assume that each time the unloading time slot is calculated, the USV has all the unloading tasks and position information of the underwater sensor cluster M known.

[0075] In this embodiment, the dual-layer offloading model of the marine MEC network includes two transmission stages, namely, the underwater acoustic transmission segment between the USN and the AUV / SN in the first stage and the radio frequency transmission segment between the SN and the UAV in the second stage.

[0076] In the first stage, without loss of generality, the AUV i and SN n Analysis of underwater sensor group M within the working range. USN m Upload their workloads to the AUVs separately in the NOMA manner i and SN n , USN m With AUV i The data between them is transmitted through the underwater acoustic channel. The underwater acoustic communication model is expressed as:

[0077]

[0078] in, for USN m With AUV i The parameter γ is the expansion factor, the parameter θ is the center frequency of the underwater acoustic signal, and L(θ) is the absorption coefficient.

[0079] Likewise, the USN m With SN n The data between them is also transmitted through the underwater acoustic channel. The underwater acoustic communication model is expressed as:

[0080]

[0081] in, for USN m With SN n The Euclidean distance between .

[0082] The absorption coefficient L(θ) is obtained according to Thorp's empirical formula:

[0083]

[0084] Unlike traditional white noise, underwater acoustic signals are generally affected by ocean noise during transmission. Ocean noise usually has four main components: turbulence W1(θ), unmanned ship W2(θ), wind and waves W3(θ), and thermal noise W4(θ), which are expressed as follows:

[0085] 10log 10 W1(θ)=17-30log 10 θ

[0086] 10log 10 W2(θ)=40+20(a-0.5)+26log 10 θ-60log 10 (θ+0.03)

[0087] 10log 10 W3(θ)=50+7.5b 0.5 +20log 10 θ-40log 10 (θ+0.4)

[0088] 10log 10 W4(θ)=-15+20log 10 θ

[0089] Among them, a represents the activity coefficient of the unmanned ship, and b represents the wind speed, in m / s.

[0090] The noise power density W(θ) can be expressed as:

[0091] W(θ)=W1(θ)+W2(θ)+W3(θ)+W4(θ)

[0092] USN m and AUVs i The underwater acoustic channel gain g m,i It can be expressed as:

[0093]

[0094] Among them B m,i for USN m Upload data to AUV i The channel width;

[0095] Due to the adoption of NOMA technology, USNs within the same AUV coverage area share the same channel, so USNs within the same AUV coverage area will interfere with each other, causing the AUV to i USN signals within the coverage area are sorted by received power:

[0096] p1g 1,i >…>p m g m,i >…p j g j,i …>p M g M,i

[0097] Among them, p m for USN m transmission power.

[0098] When USN m Towards AUV i When sending data, USN m Upload data to AUV i The transmission rate r m,i It can be expressed as:

[0099]

[0100] Among them, B m,i for USN m Upload data to AUV i The channel width, p m for USN m Transmit power of uploaded data, σ 2 represents the additive white Gaussian noise power at the underwater sensor.

[0101] Similarly, USNs within the same SN coverage area will interfere with each other, n The signals sent by USNs within the coverage area are sorted by the received power:

[0102] p1g 1,n >…>p m g m,n >…p k g k,n …>p M g M,n

[0103] When USN m To SN n When sending data, USN m Upload data to SN n The transmission rate r m,n It can be expressed as:

[0104]

[0105] Among them, B m,n for USN m Upload data to SN n channel bandwidth.

[0106] In the second phase;

[0107] SN continues to unload part of its data to UAV in FDMA mode. For the radio frequency transmission stage on the water, since UAV n Always visible, so UAV and SN n The communication link between them can be reasonably modeled as a line-of-sight (LoS) link, then the channel gain H n It can be described by the free space path loss model:

[0108]

[0109] Among them, d n,uav It's SN n The parameter χ is the channel power gain per unit distance (i.e., 1 m), and the parameter δ is the path loss exponent.

[0110] SN n The transmission rate between the UAV and the n,uav It can be expressed as:

[0111]

[0112] Among them, B n SN n The channel width, p n SN n The transmission power, N represents SN n The additive white Gaussian noise power at .

[0113] S2, based on the dual-layer offloading model of the marine MEC network, aims to minimize the total delay of USN data processing and constructs a joint optimization problem. The optimization variables are the offloading strategy of the USN, the position of the AUV, and the offloading strategy of the SN.

[0114] The total delay in completing USN data processing is t m Unloading to AUV i Processing delay t m,i and uninstall to SN n By SN n Latency of co-processing with UAV t m,n It consists of two parts:

[0115] t m =max{tm,i ,t m,n}

[0116] 1) Delay t m,i It consists of two parts: transmission delay and processing delay Transmission delay and processing delay Respectively expressed as:

[0117]

[0118] Among them, C m for USN m The total number of CPUs required for the task, f auv For AUV i Computational resources, R m for USN m Total number of bits in the task, β m for USN m Uninstall rate, r m,i for USN m Upload data to AUV i The transmission rate, B m,i for USN m Upload data to AUV i The channel width, g m,i for USN m and AUVs i The underwater acoustic channel gain between j,i for USN j and AUVs i The underwater acoustic channel gain between 2 represents the additive white Gaussian noise power at the underwater sensor.

[0119] 2) USN m Uninstall to SN n By SN n Latency of co-processing with UAV t m,n Contains two parts: USN m Data transmission to SN n Delay and SN n Data processing latency Right now:

[0120]

[0121] USN m Data transmission to SN n Delay Expressed as:

[0122]

[0123] Among them, r m,n for USN m Upload data to SN n The transmission rate, B m,n for USN m Upload data to SN n The channel bandwidth, g m,n for USN m With SN n The channel gain between k,n for USN k With SN n The channel gain between

[0124] SN n Data processing latency Including local computing latency Processing delay with continued offloading to UAV Right now:

[0125]

[0126] SN n Local computing latency Expressed as:

[0127]

[0128] Among them, f sn SN n Computational resources, α m SN n (About USN m mission portion) to the UAV.

[0129] SN n Continue to offload to UAV processing delay By SN n Transmission delay to UAV Computational delay with UAV It consists of two parts, among which, and Respectively expressed as

[0130]

[0131] Among them, r n,uav SN n The transmission rate between the UAV and the n SN n The channel width, p n SN n The transmission power, N represents SN nThe additive white Gaussian noise power at uav Computational resources for UAVs;

[0132] In summary, the total delay for USN data processing is t m It can be expressed as:

[0133]

[0134] The joint optimization problem of USN total delay is:

[0135]

[0136] stC1:0≤β m ≤1,m∈M

[0137] C2:0≤α m ≤1,m∈M

[0138] C3:0≤p m ≤p max ,m∈M

[0139] C4: 0≤x i ,y i ≤100

[0140] Among them, C1-C4 are the constraints of the optimization problem P1, I represents the AUV cluster, N represents the SN cluster, and x i For AUV i x coordinate, y i For AUV i The y coordinate of p max The upper limit of the sensor transmission power; p m for USN m The transmission power of M is the underwater sensor cluster.

[0141] S3, decomposes the joint optimization problem into three optimization sub-problems: the unloading strategy of USN, the position of AUV, and the unloading strategy of SN. The gradient descent method, particle swarm algorithm, and CVX are used to obtain the solutions of each optimization sub-problem respectively, and then the solution of the joint optimization problem is obtained through alternating iterations.

[0142] 1) The USN unloading strategy sub-problem is expressed as:

[0143]

[0144] stC1:0≤β m ≤1,m∈M

[0145] The gradient descent method is used to solve the optimization sub-problems of the optimization algorithm of the USN unloading strategy, including:

[0146] S311, use smooth approximation to process the max function contained in the objective function of problem P3, the inner function InnerSmooth after the approximation of the P3 objective function m for:

[0147]

[0148] Outer approximation function OuterSmooth m for:

[0149]

[0150] Combining the smoothing results of the inner and outer layers, the complete smoothing objective function is:

[0151]

[0152] S312, calculate the gradient ▽f(β (t) ), outer approximation function OuterSmooth m The partial derivative of is:

[0153]

[0154] in,

[0155]

[0156]

[0157] Inner approximation function InnerSmooth m Partial derivatives of :

[0158]

[0159] in,

[0160]

[0161] Therefore, the objective function The partial derivative of is:

[0162]

[0163] S313, update the variable according to the step size formula of the gradient descent method, β (t+1) =β (t) -η·▽f(β (t) );

[0164] S314, due to β m Constrained 0≤β m ≤1, β needs to be adjusted after each update mProject it to ensure that it falls within the feasible region.

[0165] S315, substitute the current β (t+1) , calculate the smoothing objective function value f (t+1) ,

[0166] S316, check the convergence conditions (i.e. |f (t+1) -f (t) |<ò, ò is the convergence threshold), and if the convergence condition is met, the final As the optimization result, otherwise continue to iterate and update.

[0167] 2) The SN offloading strategy sub-problem is expressed as:

[0168]

[0169] stC2:0≤α m ≤1,m∈M

[0170] The convex optimization CVX toolbox is used to solve the SN offloading strategy subproblem, including:

[0171] S321. Input the current USN unloading strategy and AUV position, and prove that the SN unloading strategy optimization sub-problem is a convex optimization problem;

[0172] S322, use the CVX toolbox provided by MATLAB to solve the SN unloading strategy optimization sub-problem.

[0173] 3) The AUV position optimization sub-problem is expressed as:

[0174]

[0175] in, represents the set of positive integers;

[0176] The particle swarm algorithm is used to solve the AUV position optimization sub-problem, including:

[0177] S331, construct the optimization objective function of P5 as the fitness function of the particle swarm algorithm, and use the penalty function method to incorporate constraint C4 into the objective function of problem P5. The fitness function of the particle is constructed as follows:

[0178]

[0179] S332, input the initial particle swarm position information and calculate the initial particle swarm fitness;

[0180] S333, in each iteration, each particle At speed Vl t+1 Update its new position based on The particle's velocity and position are updated as:

[0181]

[0182] Where ω is the inertia weight; c1 and c2 are learning factors, usually called social confidence or cognitive confidence respectively; r1 and r2 are random numbers between 0 and 1; is the historical optimal position of particle n, gbest t is the historical optimal position of the population;

[0183] S334, after updating the position, it needs to be checked. If the X l1 or X l2 If the value exceeds the range of [0,100], the exceeding part will be truncated to the boundary value (0 if less than 0, 100 if greater than 100), and the coordinate value will be rounded to a positive integer;

[0184] S335: Substitute the updated and constrained particle position into the objective function to calculate the new fitness value. Update the individual optimal position and global optimal position of the particle.

[0185] S336, when the number of iterations reaches the set maximum number of iterations, the iteration is stopped. The global optimal position at this time is the approximate optimal AUV position found by the particle swarm algorithm, and the corresponding global optimal fitness value is the approximate optimal objective function value; if it is not reached, the iteration is continued.

[0186] Based on similar inventive concepts, an embodiment of the present invention also provides a computer storage medium storing a readable program. When the program is executed by a processor, it can execute the above-mentioned dual-layer task offloading optimization method in an ocean MEC network.

[0187] Based on similar inventive concepts, an embodiment of the present invention provides an electronic device, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus;

[0188] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the above-mentioned dual-layer task offloading optimization method in the marine MEC network.

[0189] Based on similar inventive concepts, an embodiment of the present invention also provides a computer program product, including computer instructions, which instruct a computing device to perform operations corresponding to the above-mentioned dual-layer task offloading optimization method in an ocean MEC network.

[0190] Example 2

[0191] In this embodiment, the dual-layer task offloading optimization method in the marine MEC network proposed in Example 1 (the optimization method of the present invention) is simulated and verified;

[0192] Figure 2 The performance comparison chart of the method of the present invention and other methods under different sensor numbers is given, among which method 1 is the SN unloading rate α m =0.5, other parameters are the same as those in the optimization method of the present invention; USN unloading rate β in method 2 m =0.5, and the other parameters are the same as those in the optimization method of the present invention; in method 3, the AUV position is random, and the other parameters are the same as those in the optimization method of the present invention; Figure 2 It can be seen from the figure that the performance of the present invention is the best compared with other methods under all the numbers of underwater sensors.

[0193] Figure 3 The influence of task complexity on the total energy consumption of underwater sensors is given. Figure 3 As can be seen from the figure, as the task complexity κ increases, the performance index value also increases. This indicates that the parameter κ has a significant impact on system performance. A larger κ increases the system load, resulting in an increase in the total energy consumption of the underwater sensor. At the same time, when the complexity is low and the number of USNs increases, the system is better able to cope with the load pressure brought by the increase in the number of USNs, and the performance is relatively stable. At the higher complexity levels of κ = 2000 and κ = 3000, the system performance is more sensitive to the increase in the number of USNs, and further optimization may be required to cope with high-load scenarios.

[0194] Figure 4 Given the AUV computing capability f auv The impact on the total delay of underwater sensor mission completion is as follows: Figure 4 It can be seen that under the same number of USNs, f auv The larger it is, the shorter the total delay in completing the task. This is because the more computing resources the AUV has, the more computing resources are allocated to each underwater sensor task, and the shorter the task processing delay unloaded to the AUV. At the same time, the computing power of the AUV is enhanced, and it can undertake more computing tasks, which reduces the computing pressure of the SN and UAV, and the task processing delay unloaded to the SN also becomes shorter.

[0195] Example 3

[0196] Based on the dual-layer task offloading optimization method in a marine MEC network proposed in Example 1, this embodiment proposes a dual-layer task offloading optimization system in a marine MEC network, specifically including:

[0197] Model construction module: Construct a two-layer offloading model of the marine MEC network including USN, AUV, SN and UAV;

[0198] Optimization problem construction module: Based on the dual-layer offloading model of the marine MEC network, with the goal of minimizing the total delay of USN data processing, a joint optimization problem is constructed. The optimization variables are the USN offloading strategy, the AUV position, and the SN offloading strategy.

[0199] In addition, the optimization problem decomposition and solution module: the joint optimization problem is decomposed into three optimization sub-problems: the unloading strategy of USN, the position of AUV, and the unloading strategy of SN. The gradient descent method, particle swarm algorithm and CVX are used to obtain the solution of each optimization sub-problem respectively, and then the solution of the joint optimization problem is obtained through alternating iteration.

[0200] The method of the present invention can be implemented in hardware, firmware, or as software or computer code that can be stored in a recording medium (such as a CDROM, RAM, floppy disk, hard disk or magneto-optical disk), or as computer code that is originally stored in a remote recording medium or a non-temporary machine-readable medium downloaded over a network and will be stored in a local recording medium, so that the method described herein can be stored in such software processing on a recording medium using a general-purpose computer, a special-purpose processor or programmable or special-purpose hardware (such as an ASIC or FPGA). It will be understood that a computer, a processor, a microprocessor controller or programmable hardware includes a storage component (e.g., RAM, ROM, flash memory, etc.) that can store or receive software or computer code, and when the software or computer code is accessed and executed by a computer, a processor or hardware, the method described herein is implemented. In addition, when a general-purpose computer accesses the code for implementing the method shown here, the execution of the code converts the general-purpose computer into a special-purpose computer for executing the method shown here.

[0201] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention, and such changes and modifications fall within the scope of the invention as claimed.

Claims

1. A dual-layer task offloading optimization method in marine MEC networks, characterized in that: The following steps are involved: Construct a dual-layer offloading model for marine MEC networks including USN, AUV, SN, and UAV; Based on the dual-layer offloading model of the marine MEC network, a joint optimization problem is constructed with the goal of minimizing the total delay of USN data processing. The optimization variables are the offloading strategy of the USN, the position of the AUV, and the offloading strategy of the SN. The joint optimization problem is decomposed into three optimization sub-problems: the unloading strategy of USN, the position of AUV and the unloading strategy of SN. The solutions of each optimization sub-problem are obtained by gradient descent method, particle swarm algorithm and CVX respectively, and the solution of the joint optimization problem is obtained through alternating iteration.

2. A dual-layer task offloading optimization method in a marine MEC network according to claim 1, characterized in that: The USN offloads its tasks to the AUV and SN respectively in the NOMA mode. The AUV and SN receive the data uploaded by the USN through underwater acoustic signals and process the computing tasks. The SN continues to offload part of its data to the UAV in the FDMA mode. The UAV receives the data from the SN through radio frequency signals and processes it.

3. A dual-layer task offloading optimization method in a marine MEC network according to claim 1, characterized in that: The total delay in completing USN data processing is t m Expressed as: in, is the transmission delay, To handle delay, for USN m Data transmission to SN n The delay, Calculate the latency locally. SN n Transmission delay of uploading data to UAV, is the computational delay of UAV; C m for USN m The total number of CPUs required for the task, f auv For AUV i Computational resources, R m for USN m Total number of bits in the task, β m for USN m To SN n Uninstall rate, r m,i for USN m Upload data to AUV i The transmission rate, r m,n for USN m Upload data to SN n The transmission rate, f sn SN n Computational resources, α m SN n Unloading rate to UAV; r n,uav SN n Transmission rate between UAV; f uav It is the computing resource of UAV.

4. A dual-layer task offloading optimization method in a marine MEC network according to claim 3, characterized in that: The joint optimization problem is: Among them, C1-C4 are the constraints of the optimization problem P1, I represents the AUV cluster, N represents the SN cluster, and x i For AUV i x coordinate, y i For AUV i The y coordinate of p max is the maximum transmission power of USN, p m for USN m The transmission power of , i represents the i-th AUV, m represents the m-th USN, n represents the n-th SN; M represents the underwater sensor cluster.

5. A dual-layer task offloading optimization method in a marine MEC network according to claim 4, characterized in that: The USN offloading strategy sub-problem is expressed as: The SN offloading strategy sub-problem is expressed as: The AUV position optimization subproblem is expressed as: in, is a set of positive integers.

6. A dual-layer task offloading optimization method in a marine MEC network according to claim 5, characterized in that: The steps of solving the USN unloading strategy sub-problem using the gradient descent method include: S311, use smooth approximation to process the max function contained in the objective function of problem P3, the inner function InnerSmooth after the approximation of the P3 objective function m for: Among them, τ is the control parameter. Approximate outer function OuterSmooth m for: Combining the smoothing results of the inner and outer layers, the complete smoothing objective function is: S312, calculate gradient Outer approximation function OuterSmooth m The partial derivative of is: in: Inner approximation function InnerSmooth m Partial derivatives of : in: Therefore, the objective function The partial derivative of is: S313, update the variables according to the step size formula of the gradient descent method: S314, after each update, m To project: S315, substitute the current β (t+1) , calculate the smoothing objective function value f (t+1) : S316, check convergence conditions: |f (t+1) -f (t) |<ò, ò is the convergence threshold, and the final value is returned if the convergence condition is met. As the optimization result, otherwise continue to iterate and update.

7. A dual-layer task offloading optimization system in marine MEC networks, characterized by: include: Model construction module: Construct a two-layer offloading model of the marine MEC network including USN, AUV, SN and UAV; Optimization problem construction module: Based on the dual-layer offloading model of the marine MEC network, with the goal of minimizing the total delay of USN data processing, a joint optimization problem is constructed. The optimization variables are the USN offloading strategy, the AUV position, and the SN offloading strategy. In addition, the optimization problem decomposition and solution module: the joint optimization problem is decomposed into three optimization sub-problems: the unloading strategy of USN, the position of AUV, and the unloading strategy of SN. The gradient descent method, particle swarm algorithm and CVX are used to obtain the solution of each optimization sub-problem respectively, and then the solution of the joint optimization problem is obtained through alternating iteration.

8. A computer storage medium storing a readable program, characterized in that: When the program is executed by the processor, it can execute the dual-layer task offloading optimization method in the marine MEC network described in any one of claims 1 to 6.

9. An electronic device, characterized in that: include: A processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the dual-layer task offloading optimization method in the marine MEC network according to any one of claims 1 to 6.

10. A computer program product comprising computer instructions, characterized in that The computer instructions instruct the computing device to perform operations corresponding to the dual-layer task offloading optimization method in the marine MEC network as described in any one of claims 1-6.

Images