Big-task-oriented D2D (Device-to-Device) routing method for calculation fusion in damaged cellular network

Abstract

The invention relates to the technical field of damaged cellular networks, in particular to a large-task-oriented calculation fusion D2D routing method in a damaged cellular network, which comprises the following steps of: firstly, acquiring a network topology of the damaged cellular network, task request information of a terminal and attribute information of the terminal; then, taking the minimum time delay as an optimization target, and constructing a transfer calculation joint optimization model of path selection and calculated amount distribution; discrete variables and continuous variables of the joint optimization problem are decoupled and decomposed into two sub-problems of path selection and calculation amount distribution, and the solving complexity is reduced; and finally, an HPSO-KOA algorithm is adopted to search a D2D path and optimize the calculation amount of the terminal, and an optimal transmission path and optimal calculation amount distribution are determined according to a solving result, so that the large task processing time delay under the condition of limited communication is effectively reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of damaged cellular networks, and in particular to a transmission-computation-fused D2D routing method for large tasks in damaged cellular networks. Background Art

[0002] With the rapid development of technologies such as the Internet of Things, artificial intelligence, and big data, many compute-intensive and data-intensive applications have emerged, such as video rendering, image processing, AI model training, and data preprocessing. These tasks typically involve large amounts of data and high computational complexity, placing higher demands on computing and communication resources. Due to the remote location of cloud servers, offloading tasks to edge servers for processing is widely used. However, this centralized processing places significant computing and communication pressure on MEC servers.

[0003] Especially during natural disasters or severe weather conditions, the quality of cellular links within a cell can degrade significantly, leaving only a few terminals able to establish a tenuous connection with the nearest base station. This further exacerbates communication pressure and access bottlenecks on edge servers. In these situations, for large-scale computing tasks that need to be uploaded to remote decision-making centers via the base station, directly offloading the raw data can result in long transmission times or even failure to transmit successfully. Summary of the Invention

[0004] The purpose of the present invention is to provide a D2D routing method for large-scale tasks in a damaged cellular network, aiming to utilize inter-device communication to assist edge computing and effectively solve the problem of large-scale task processing under communication-constrained conditions.

[0005] To achieve the above objectives, the present invention provides a D2D routing method for large-scale tasks in a damaged cellular network with integrated transmission and computation, comprising the following steps:

[0006] Step 1: Obtain the network topology of the damaged cellular network, the task request information of the terminal, and the attribute information of the terminal;

[0007] Step 2: With latency minimization as the optimization goal, a joint optimization model for path selection and computational allocation is constructed.

[0008] Step 3: Decouple the discrete variables and continuous variables of the joint optimization problem and decompose it into two sub-problems: path selection and computational allocation;

[0009] Step 4: Use the HPSO-KOA algorithm to search for D2D paths and optimize the terminal's computational workload. Determine the optimal transmission path and optimal computational workload allocation based on the solution results.

[0010] Optionally, the network topology of the damaged cellular network obtained in step 1 is used to clarify available connection relationships between terminals and between terminals and base stations;

[0011] The task request information of the terminal includes the size of the corresponding task, the required computing resources and the maximum allowed delay;

[0012] The terminal attribute information includes the terminal's transmission rate, transmission power, computing resources, and available energy.

[0013] Optionally, in step 2, the path selection and the terminal's computing workload distribution are used as optimization variables, and the objective function is determined with the goal of minimizing the total delay in completing the task;

[0014] The expression of the objective function is:

[0015]

[0016] Where T is the total delay to complete the task, which is the sum of the transmission delay and computation delay of all links on the selected path. Q represents the path quality of the selected path, which is evaluated by the number of hops and computing resources of the path. χ represents the computational load of the terminals on the path. P represents the path set from the source terminal to the destination terminal. r (s,d) represents any path from the source terminal to the destination terminal, Indicates terminal v k Transfer tasks to terminal v k+1 The delay, Indicates terminal v k The amount of data transferred, Indicates terminal v k To terminal v k+1 The transmission rate, Indicates terminal v k The latency of the computation task, x k Indicates terminal v k The amount of calculation, I is the computational complexity of the task, For terminal v k The calculation frequency of .

[0017] Optionally, in step 2, the constraints of path selection, terminal computing resources, available energy of the terminal, and total delay for completing the task are used as constraints;

[0018] The constraint expression is:

[0019]

[0020] Where C1 and C2 are used to prevent the path from being too long and the terminal from taking too long to calculate. C3 indicates that the delay in completing the task transmission and calculation cannot exceed the maximum tolerable delay of the task. C4 indicates that every terminal in the routing process participates in the calculation. C5 indicates that the energy consumed by the terminal cannot exceed the energy available to the terminal, preventing the task from being interrupted during the terminal transmission or calculation process.

[0021] Optionally, the execution process of step 3 is specifically to first use the hybrid particle swarm algorithm to solve the path selection subproblem, and then use the Kepler algorithm to solve the computational allocation subproblem based on the first subproblem, so as to obtain the optimal solution to the entire optimization problem.

[0022] The present invention provides a transmission-computation-integrated D2D routing method for large tasks in a damaged cellular network. The method first obtains the network topology of the damaged cellular network, the task request information of the terminal, and the attribute information of the terminal. Then, with minimizing the delay as the optimization goal, a transmission-computation joint optimization model for path selection and computation allocation is constructed. The discrete variables and continuous variables of the joint optimization problem are decoupled and decomposed into two sub-problems of path selection and computation allocation to reduce the complexity of the solution. Finally, the HPSO-KOA algorithm is used to search for D2D paths and optimize the computation of the terminal. The optimal transmission path and optimal computation allocation are determined based on the solution results, effectively reducing the processing delay of large tasks under communication-restricted conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] 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, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0024] Figure 1 This is a schematic flow chart of the steps of a transmission-computation-fused D2D routing method for large tasks in a damaged cellular network of the present invention.

[0025] Figure 2 2 is a schematic diagram of the system model structure of a damaged cellular network in a specific embodiment of the present invention.

[0026] Figure 3 It is a flowchart of using the HPSO-KOA algorithm to search for D2D paths and optimize terminal computing power in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0027] The following describes embodiments of the present invention in detail, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and are not to be construed as limiting the present invention.

[0028] The present invention provides a D2D routing method for large-scale tasks in a damaged cellular network with integrated transmission and computation, comprising the following steps:

[0029] S1: Obtain the network topology of the damaged cellular network, the task request information of the terminal, and the attribute information of the terminal;

[0030] S2: Taking latency minimization as the optimization goal, a joint optimization model for path selection and computational allocation is constructed;

[0031] S3: Decouple the discrete and continuous variables of the joint optimization problem and decompose it into two sub-problems: path selection and computational allocation.

[0032] S4: Use the HPSO-KOA algorithm to search for D2D paths and optimize the terminal's computational workload, and determine the optimal transmission path and optimal computational workload allocation based on the solution results.

[0033] Please refer to the execution process Figure 1 , the following is further explained in conjunction with specific embodiments and execution steps:

[0034] Step S1: obtaining the network topology of the damaged cellular network, the task request information of the terminal, and the attribute information of the terminal;

[0035] The terminals are randomly distributed within a circle with a radius of 400 meters. Most terminals cannot establish a direct connection with the base station and can only perform multi-hop transmission through D2D communication. Only a few terminals can communicate with the base station nearby. The maximum communication distance between any two terminals is 100 meters. The network topology is generated on this basis. In this embodiment, the task is a large task that requires remote decision-making. Only one terminal is considered to have a request task. The task request information is expressed as θ = {D, I, T max}, D is the size of the task, I is the computational complexity of the task, which means the number of cycles required to calculate 1 bit of data, T max The maximum delay allowed to complete this task; the terminal attribute information includes the terminal's transmission rate R, transmission power p T , computing resources and available energy E avl .

[0036] Step S2: Taking the minimum latency as the optimization goal, a joint optimization model for path selection and computational allocation is constructed;

[0037] In a damaged cellular network, where cell links are poor, offloading large data tasks directly to the base station can result in high transmission latency or even transmission failure. This paper provides a D2D routing method for large-scale tasks that integrates transmission and computation, suitable for use in damaged cellular networks. This method first searches for an optimal path based on network topology and terminal resources that can both complete computations and reliably reach the base station. It then dynamically adjusts the computational load of each terminal based on the computing power of each terminal along the optimal path.

[0038] refer to Figure 2 The terminals in the embodiment of the present invention are divided into two categories: aggregation terminals and relay terminals. The aggregation terminal refers to the terminal that can communicate with the base station nearby. The other terminals are relayed to the aggregation terminal through multi-hop D2D, and then the aggregation terminal communicates with the base station. The set of aggregation terminals is represented by M = {1,2,3,…m}, and the set of relay terminals is represented by N = {1,2,3,…n}. In this embodiment, the task performs multi-terminal collaborative calculation on the multi-hop D2D transmission path, and defines the path set P from the source terminal to the destination terminal = {p1(s,d),…p r (s,d),…p t (s,d)}, p r (s,d)∈P is any path from the source terminal s to the destination terminal d. are the number of hops and computing resources of any path, H th 、F th are the number of hops in the path and the computing resource threshold respectively.

[0039] The selected path must satisfy the following constraints: First, among all paths from the source terminal to the destination terminal, paths with too many hops must be excluded to prevent excessive transmission delays for the task: Secondly, eliminate paths with too few computing resources to prevent excessive task computation delays: Q is defined as path quality, which is evaluated by the number of hops and computing resources of the path. The fewer hops and the more computing resources a path has, the better the path quality and the lower the latency. Path quality can be expressed as: Among them, μ1 and μ2 are the weights of hop count and computing power resources respectively, μ1+μ2=1, and The normalized value of the path hop count and computing power resources. Assuming that the endpoint of the path can complete the task, the path with the best quality is selected as the optimal path.

[0040] The task is calculated during the routing process. Each time it passes a terminal, the amount of data for the task will decrease. The amount of calculation for each terminal is different. If the optimal path selected has h terminals, then the task is at the kth terminal v on the path. k The amount of data transferred out Indicates that xl is the lth terminal v on the path l The amount of calculation, and each terminal in the routing process participates in the calculation, can be used express.

[0041] The present invention uses path selection and terminal computing workload allocation as optimization variables, takes minimizing the total task completion delay as the goal, and determines an objective function; uses the path selection, the terminal computing resources and the terminal's available energy, and the limit of the total task completion delay as constraints; and constructs a joint optimization model for path selection and computing workload allocation based on the optimization variables, objective function, and constraints.

[0042] The expression of the objective function is:

[0043]

[0044] Where T is the total delay to complete the task, which is the sum of the transmission delay and computation delay of all links on the selected path. Q represents the path quality of the selected path, and χ represents the computational load of the terminal on the path. Indicates terminal v k Transfer tasks to terminal v k+1 The delay, Indicates terminal v k To terminal v k+1 The transmission rate, Indicates terminal v k The delay of the calculation task. The constraint condition expression is:

[0045]

[0046] In the formula, C1 and C2 are used to prevent paths from being too long and terminal computations from taking too long. C3 indicates that the latency required to complete task transmission and computation cannot exceed the maximum tolerable latency of the task. C4 indicates that every terminal in the routing process participates in computation. C5 indicates that the energy consumed by a terminal cannot exceed its available energy, preventing task interruptions during transmission or computation at the terminal.

[0047] Step S3: Decouple the discrete variables and continuous variables of the joint optimization problem and decompose it into two sub-problems: path selection and computational allocation;

[0048] Path selection is a discrete problem, while computational allocation is a continuous problem. Therefore, the proposed optimization problem is a mixed-integer nonlinear optimization problem, which is relatively complex to solve. To simplify the problem, we decouple the discrete and continuous variables in the optimization problem, breaking it down into two subproblems: path selection and computational allocation. We first use a hybrid particle swarm optimization algorithm to solve the path selection subproblem. Based on the solution to the first subproblem, we then use the Kepler algorithm to solve the computational allocation subproblem, ultimately obtaining the optimal solution for the entire optimization problem.

[0049] Step S4: Use the HPSO-KOA algorithm to search for D2D paths and optimize the computing workload of the terminal, and determine the optimal transmission path and optimal computing workload allocation based on the solution results.

[0050] Figure 3 The flowchart of the embodiment of the present invention for searching D2D paths and optimizing terminal computing amount by using the HPSO-KOA algorithm includes the following steps: using the HPSO algorithm to solve the path selection subproblem, first randomly initialize particles, each particle represents a candidate path, and define the position of the initial particle i as: X in =[x i1 ,x i2 ,...,x in ], where x in represents the terminal number, and n is the dimension of the particle position. The velocity of the initial particle i is expressed as: V in =[v i1 ,v i2 ,...,v in ]; calculate the fitness value of the particle, update the individual optimum and the group optimum, and the fitness value is expressed as the path quality: Select the crossover mutation operator to update the position of particles, generate new particles, and obtain the optimal solution through continuous iterative optimization.

[0051] After obtaining the optimal path, the KOA algorithm is used to optimize the computational effort of each terminal on the path. First, the position, orbital deflection rate, orbital period, and velocity of the planet are initialized. In the computational effort allocation subproblem, a computational effort allocation scheme for the optimal path is defined as a planet, and the position of the planet is expressed as Where j represents the dimension of the planet's position, which is determined by the number of terminals on the optimal path. The orbital deflection rate of the planet can be expressed as e i =rand [0,1] ,i=1,...,N, the orbital period of the planet is expressed as T i =|r|,i=1,...,N, where r is a number randomly generated according to normal distribution, N is the number of planets, and the speed of the planet is defined as V i(t); calculate the fitness value of the planet, and the planet with the smallest fitness value is regarded as the sun. Our goal is to use the KOA algorithm to solve the optimal computation allocation on the optimal path, so as to minimize the task completion time. Therefore, the fitness function can be expressed as: To update the position of a planet, as it orbits the sun, it will continuously move closer to and further away from the sun. When a planet is close to the sun, KOA focuses on a refined search in a local area and updates the position of the planet close to the sun using the following formula.

[0052]

[0053] in, represents the position of the planet at time t+1, represents a random vector of 0 or 1, represents the optimal sun position at time t, and represents two solutions randomly selected from all planets. h is an adaptive factor that controls the distance between the sun and the planet. When a planet moves away from the sun, the KOA algorithm focuses on exploring the vast space and uses a formula to update the position of the planet farther from the sun.

[0054]

[0055] Among them, F represents the sign of the planet search direction, represents a random vector of 0 or 1, F gi (t) represents the gravitational force between the planet and the sun, expressed as: Where u(t) is a function that controls the search accuracy and decreases exponentially with time (t). and is the normalized value of the mass of the sun and planets, determined by the fitness value. The KOA algorithm uses this periodic transformation to combine the advantages of both global and local searches, improving both efficiency and accuracy in solving problems.

[0056] The above disclosure is merely one or more preferred embodiments of the present invention, and certainly cannot be used to limit the scope of the present invention. A person skilled in the art can understand that all or part of the processes of the above embodiments and equivalent changes made in accordance with the claims of the present invention still fall within the scope of the invention.

Claims

1. A D2D routing method for large-scale tasks in a damaged cellular network with integrated transmission and computation, characterized in that: The following steps are involved: Step 1: Obtain the network topology of the damaged cellular network, the task request information of the terminal, and the attribute information of the terminal; Step 2: With latency minimization as the optimization goal, a joint optimization model for path selection and computational allocation is constructed. Step 3: Decouple the discrete variables and continuous variables of the joint optimization problem and decompose it into two sub-problems: path selection and computational allocation; Step 4: Use the HPSO-KOA algorithm to search for D2D paths and optimize the terminal's computational workload. Determine the optimal transmission path and optimal computational workload allocation based on the solution results.

2. The D2D routing method for large-scale tasks in a damaged cellular network according to claim 1, wherein: The network topology of the damaged cellular network obtained in step 1 is used to clarify the available connection relationships between terminals and between terminals and base stations; The task request information of the terminal includes the size of the corresponding task, the required computing resources and the maximum allowed delay; The terminal attribute information includes the terminal's transmission rate, transmission power, computing resources, and available energy.

3. The D2D routing method for large-scale tasks in a damaged cellular network according to claim 1, wherein: In step 2, the path selection and the terminal's computational workload allocation are used as optimization variables, and the objective function is determined with the goal of minimizing the total delay in completing the task. The expression of the objective function is: Where T is the total delay to complete the task, which is the sum of the transmission delay and computation delay of all links on the selected path. Q represents the path quality of the selected path, which is evaluated by the number of hops and computing resources of the path. χ represents the computational load of the terminals on the path. P represents the path set from the source terminal to the destination terminal. r (s,d) represents any path from the source terminal to the destination terminal, Indicates terminal v k Transfer tasks to terminal v k+1 The delay, Indicates terminal v k The amount of data transferred, Indicates terminal v k To terminal v k+1 The transmission rate, Indicates terminal v k The latency of the computation task, x k Indicates terminal v k The amount of calculation, I is the computational complexity of the task, For terminal v k The calculation frequency of .

4. The D2D routing method for large-scale tasks in a damaged cellular network according to claim 1, wherein: In step 2, the constraints are path selection, terminal computing resources, terminal available energy, and the total delay for completing the task; The constraint expression is: Where C1 and C2 are used to prevent the path from being too long and the terminal from taking too long to calculate. C3 indicates that the delay in completing the task transmission and calculation cannot exceed the maximum tolerable delay of the task. C4 indicates that every terminal in the routing process participates in the calculation. C5 indicates that the energy consumed by the terminal cannot exceed the energy available to the terminal, preventing the task from being interrupted during the terminal transmission or calculation process.

5. The D2D routing method for large-scale tasks in a damaged cellular network according to claim 1, wherein: The execution process of step 3 is to first use the hybrid particle swarm algorithm to solve the path selection subproblem, and then use the Kepler algorithm to solve the computational allocation subproblem based on the first subproblem, so as to obtain the optimal solution of the entire optimization problem.

Images