A SFC deployment method for UAV networks based on multifunctional task offloading

Abstract

The present invention seeks protection for a drone network SFC deployment method based on multi-functional task offloading, which belongs to the field of communication technology. In view of the problem that uneven distribution of users in low-altitude edge networks may cause network load imbalance and low resource utilization, and lead to a decline in task offloading quality and user service experience, a drone network SFC deployment method based on multi-functional task offloading is proposed. This method uses the idea of ​​differential evolution to establish a drone network SFC deployment optimization model based on constraints such as the location distribution of ground users, offloading task requirements, and airborne VNF types. By jointly iteratively optimizing the location distribution of drone swarms and SFC deployment, the resource utilization of the drone network is maximized, thereby effectively improving the task offloading completion rate and reducing the system unit task offloading delay.

Description

Technical Field

[0001] The present invention belongs to the field of communication technology, and in particular relates to a SFC deployment method for an unmanned aerial vehicle network based on multi-functional task offloading. Background Art

[0002] By leveraging drones' flexible maneuverability and superior line-of-sight communication capabilities, deploying and configuring low-altitude drone edge networks can migrate cloud server resources to the network edge closer to users, providing ground users with compute-intensive and latency-sensitive task offload services. Network Function Virtualization (NFV) technology, by decoupling network functions from the underlying hardware, enables fast and flexible virtual network function (VNF) mapping and service function chaining (SFC) orchestration, providing flexible and efficient network function support for a variety of applications.

[0003] Existing research on drone network task offloading mostly focuses on single-function and single-drone task offloading, without considering the collaborative offloading of multi-function tasks across multiple drones. Furthermore, there is relatively little research on drone network VNF mapping and SFC deployment, and related solutions do not consider task offloading services and requirements. In real-world scenarios, the uneven distribution of ground user service density can cause an imbalance in drone server load, leading to low network resource utilization and a decreased user service experience. To address this issue, the present invention proposes a drone network SFC deployment method based on multi-function task offloading. By jointly optimizing the 3D position distribution of the drone swarm and SFC deployment decisions, it can effectively improve task offloading quality and user service experience. Summary of the Invention

[0004] The present invention aims to solve the above problems of the prior art. It proposes a SFC deployment method for UAV networks based on multi-functional task offloading. The technical solution of the present invention is as follows:

[0005] A method for deploying SFC in a UAV network based on multifunctional task offloading includes the following steps:

[0006] 101. Use differential evolution algorithm to build optimization model. Let the number of drones be K, the number of individuals in the population be M, M≤K!, the mutation scaling factor be ε, the crossover probability be λ, the maximum number of iterations be G, let the iteration count variable g = 0, the population individual count variable m = 0, and use k-means algorithm to generate the 3D position vector set {D g,m | m≤M};

[0007] 102. If g≤G, go to step 103; otherwise, jump to step 107;

[0008] 103. Let m = m + 1. If m ≤ M, according to the current position vector D of the drone group g,m , obtain the node mapping decision A through the UAV network service function chain SFC deployment algorithm g,m , Link Mapping Decision B g,m and resource allocation decision C g,m , calculate the objective function value Γ(A g,m ,B g,m ,C g,m ,D g,m ), jump to step 104, otherwise, go to step 106;

[0009] 104. D g,m Perform mutation and crossover operations to obtain the position vector of the drone group after mutation and crossover Obtain the node mapping decision of the mth mutation crossover individual in the gth generation through the UAV network SFC deployment algorithm Link mapping decision and resource allocation decisions Calculate the objective function value

[0010] 105. If make Jump to step 103, otherwise, jump to step 103;

[0011] 106. From the M individuals of the g-th generation population {D g,m | m≤M} select the solution to obtain the optimal individual (A g ,B g ,C g ,D g )=argmin{Γ(A g,m ,B g,m ,C g,m ,D g,m )| m∈[1,M]}, where A g , B g , C g , D g are the optimal node mapping decision, link mapping decision, resource allocation decision, and drone swarm position vector in the g-th generation population, respectively. Let g = g + 1, m = 0, and jump to step 102;

[0012] 107. The solution of selecting the best individual from a total of G generations of populations (A *,B * ,C * ,D * )=argmin{Γ(A g ,B g ,C g ,D g )| g∈[1,G]}, where A * , B * , C * , D * They are the globally optimal node mapping decision, link mapping decision, resource allocation decision, and drone swarm position vector, respectively. End.

[0013] Furthermore, in step 101, the k-means algorithm is used to generate the 3D position vector set {D g,m | m≤M The methods of} specifically include:

[0014] Let the 3D position vector of drone j be D g,m (j) = (x j ,y j ,z j ), the drone group position vector D g,m ={D g,m (j)| j∈J}, where (x j ,y j ), z j Denote the horizontal coordinate and vertical coordinate of drone j respectively, J is the drone set, the population individual count variable m≤M, the population iteration number g≤G, let g=0, according to the location distribution of ground users, use the k-means algorithm to find K cluster centers, and randomly deploy K drones carrying different VNF ​​types to the K cluster centers to generate the M individual drone swarm 3D position vector set {D g,m | m≤M}, where the number of individuals in the population M≤K!.

[0015] Furthermore, the UAV network SFC deployment algorithm in step 103 and step 104 includes the following steps:

[0016] 1) Establish a temporary set I′=I, where I is the ground user set, and the task tolerance delay T of each user i is i max The value of , sort the elements in the set I′ in ascending order;

[0017] 2) If Take the first user i from the set I′, and put the virtual network functions VNF required by user i into the set S in sequence according to the task offloading request of user i i , jump to step 3), otherwise, jump to step 9);

[0018] 3) Add the drones in the drone set J that meet the access conditions of user i to the access drone set if Jump to step 4), otherwise, jump to step 2);

[0019] 4) If There is no UAV that meets the first VNF ​​type requirement of SFC, jump to step 5), otherwise, Remove all drones that do not meet the first VNF ​​type requirement of SFC and jump to step 8);

[0020] 5) According to the communication distance between user i and drone j, the set Arrange the elements in ascending order, take out the first drone j in J′, and let user i’s access drone j a =j;

[0021] 6) According to the access drone j a and the UAV network status, construct the mapping decision of the SFC required by user i {A i ,B i}, if successful, jump to step 7), otherwise, jump to step 2);

[0022] 7) According to the mapping decision {A i ,B i} and the UAV network status, construct the resource allocation decision of the SFC required by user i {C i}, jump to step 2);

[0023] 8) According to the collection The available computing resources c of each UAV j in j The value of the set The elements in are arranged in descending order, and the ones with the same c are sorted according to the communication distance between user i and drone j. j Arrange the elements of the value in ascending order and take out The first drone j in the list, allowing user i to access drone j a =j, jump to step 6);

[0024] 9) For the available computing resources c in set J j >0, each drone j will take turns to add the minimum unit computing resources to each VNF instance on drone j and update the resource allocation decision {C i}, until cj =0;

[0025] 10) Output SFC deployment decision {A, B, C} = {A i ,B i ,C i | i∈I}, the algorithm ends.

[0026] Furthermore, in step 6), the mapping decision {A i ,B i} method is as follows:

[0027] 11) Initialize temporary variable j t =j a , temporary set J′=J, Count variables n=0, l=0;

[0028] 12) Let n = n + 1, and find the VNF type that meets the SFC nth requirement and has available computing resources c in J′. j UAVs with >c0 are added to the set J″, where c0 represents the basic computing resources required for VNF instantiation, and the node mapping cost τ of each UAV in the set J″ is updated. j ;

[0029] 13) If According to τ j , sort the elements in the set J″ in ascending order, take out the first drone j in J″ as the nth VNF mapping node and add it to the node mapping decision A i , the drone j t The minimum cost path to UAV j Join link mapping decision B as the lth virtual link i , let j t = j, remove j from the set J′ and jump to step 14), otherwise, jump to step 15);

[0030] 14) If n<|S i |, let l=l+1, Jump to step 12), otherwise, output the mapping decision of user i {A i ,B i}, jump to step 15);

[0031] 15) The algorithm ends.

[0032] Furthermore, the node mapping cost τ of the drone in step 12) j The calculation method is shown in formula (1):

[0033]

[0034] In formula (1), 0≤α≤1, represents the normalized available computing resources of UAV j, as shown in formula (2), Indicates the path The normalized distance is shown in formula (3).

[0035] Furthermore, in step 7), the resource allocation decision {C i The method comprises the following steps:

[0036] 21) Decision A based on node mapping i , add the mapping node to the temporary set J′, and allocate the basic computing resources c0 required for VNF instantiation to each drone in J′, let c j =c j -c0 Decision B based on link mapping i , the bandwidth b required by the virtual link i Add resource allocation decision C i , and obtain the total transmission delay of user i's offloading path in, is the cumulative transmission delay of task data through each physical link on the offloading path, and the counting variable n is set to 0;

[0037] 22) Let the temporary set J″=J′. For each UAV j in the set J′, calculate the task computing resources that UAV j needs to allocate to user i according to the task offloading request made by user i. And add resource allocation decision C i ,make

[0038] 23) If the task calculates the delay Among them, T i max =Tolerance delay of task of user i, jump to step 24), otherwise, output resource allocation decision C i , jump to step 25);

[0039] 24) c j = 0 are removed from the set J″ if According to c j The value of , sort the elements in the set J″ in descending order, and for the first drone j in J″, add a unit of computing resource c1 to the VNF instance of user i and update the resource allocation decision C i , let c j =c j -c1, jump to step 23), otherwise, jump to step 25);

[0040] 25) The algorithm ends.

[0041] Furthermore, in step 22), the task computing resources that the drone j needs to allocate to the user i are calculated. The method is shown in formula (4):

[0042]

[0043] In formula (4), u i represents the amount of task data for user i, w i represents the task complexity of user i.

[0044] Furthermore, the task calculation delay in step 23) is The calculation method is shown in formula (5):

[0045]

[0046] Furthermore, the definition of the objective function in step 103, step 104, step 105, step 106 and step 107 is as shown in formula (6):

[0047]

[0048] In formula (6), (A, B, C, D) is expressed as (A g,m ,B g,m ,C g,m ,D g,m )or or (A g ,B g ,C g ,D g ) or (A * ,B * ,C * ,D * ), where 0≤β≤1, ψ represents the system task completion rate, Indicates the normalized system unit task latency.

[0049] Furthermore, in step 104, the position vector D of the drone group is g,m The methods for performing mutation and crossover operations are as follows:

[0050] 1) Mutation operation: According to formula (7), by g,m The position vector D of UAV j g,m (j) Perform mutation processing to obtain the mutation position vector Thus, variant individuals are obtained

[0051]

[0052] In formula (7), ξ1, ξ2, ξ3 are randomly generated integers in the range [1, M], where ξ1≠ξ2≠ξ3≠m;

[0053] 2) Crossover operation: According to formula (8), by g,m The position vector D of UAV j g,m (j) Perform cross processing to obtain the cross position vector Thus, we obtain crossover individuals

[0054]

[0055] In formula (8), rand(0,1) is a random number generated in the range of (0,1).

[0056] The advantages and beneficial effects of the present invention are as follows:

[0057] This invention discloses a method for deploying SFCs in drone networks based on multi-functional task offloading. Existing research on drone network task offloading mostly focuses on single-function and single-drone task offloading, but fails to consider the collaborative offloading of multi-functional tasks across multiple drones. Furthermore, existing research on VNF mapping and SFC deployment in drone networks is limited, and related solutions fail to consider task offloading services and requirements. In real-world scenarios, the uneven distribution of ground user traffic density can lead to unbalanced load on drone servers, resulting in low network resource utilization and a degraded user service experience. To address this issue, the present invention utilizes differential evolution to establish a drone network SFC deployment optimization model based on constraints such as the location distribution of ground users, offload task requirements, and onboard VNF types. Based on the different VNF ​​types required for offloaded tasks, drone node location distribution is planned and corresponding SFC mapping decisions are made. Resource allocation decisions are then made based on the state of onboard computing resources and task latency requirements. Through iterative optimization of drone swarm location distribution and SFC deployment, resource utilization in the drone network is maximized, effectively improving the task offload completion rate and reducing the system-wide unit task offload latency. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 The present invention provides a preferred embodiment of a UAV network SFC deployment method based on multi-functional task offloading. DETAILED DESCRIPTION

[0059] The following will describe the technical solutions in the embodiments of the present invention in detail with reference to the accompanying drawings. The described embodiments are only a part of the embodiments of the present invention.

[0060] The technical solution of the present invention to solve the above technical problems is:

[0061] The concepts and models involved in the present invention are as follows:

[0062] 1. System model:

[0063] This paper assumes that a low-altitude edge network consists of several drones with a certain level of computing power. Each drone is equipped with a limited number and type of VNFs, and drones can choose to communicate directly with each other in line-of-sight mode or through drone relay forwarding based on the effective communication range. Based on user distribution and task offloading requirements, the low-altitude edge network can organize and orchestrate collaborative drones with corresponding VNFs to build the required SFCs and provide task offloading services. VNFs on the same SFC cannot be mapped to the same drone node.

[0064] 2. Other symbols involved in the present invention are explained as follows:

[0065] ψ: System task completion rate

[0066] Normalized system unit task offloading latency

[0067] A i : Node mapping decision of user i

[0068] B i : Link mapping decision of user i

[0069] C i : Resource allocation decision of user i

[0070] τ j : Node mapping cost of drone j

[0071] c j : Available computing resources of UAV j

[0072] Computing resources allocated by drone j to user i

[0073] c0: Computing resources required for VNF instantiation

[0074] b i : Task bandwidth requirement of user i

[0075] S i : The VNF set required by user i

[0076] u i : The amount of task data for user i

[0077] w i : Task complexity of user i

[0078] T i max : User i's task tolerance delay

[0079] The technical solution of the present invention is described as follows:

[0080] 1. Use the k-means algorithm to generate the 3D position vector set {D g,m | m≤M} method

[0081] Let the 3D position vector of drone j be D g,m (j) = (x j ,y j ,z j ), the drone group position vector D g,m ={D g,m (j)| j∈J}, where (x j ,y j ) and z j Denote the horizontal coordinate and vertical coordinate of drone j respectively, J is the drone set, the population individual count variable m≤M, the population iteration number g≤G, let g=0, according to the location distribution of ground users, use the k-means algorithm to find K cluster centers, and randomly deploy K drones carrying different VNF ​​types to the K cluster centers to generate the M individual drone swarm 3D position vector set {D g,m | m≤M}, where the number of individuals in the population M≤K!;

[0082] 2. Node mapping cost τ of drone j Calculation method

[0083]

[0084] In formula (1), 0≤α≤1, represents the normalized available computing resources of UAV j, as shown in formula (2), Indicates the path The normalized distance of is shown in formula (3);

[0085] 3. Computing resources allocated by drone j to user i Calculation method

[0086]

[0087] In formula (4), u i represents the amount of task data for user i, w i represents the task complexity of user i; 4. Task calculation delay Calculation method

[0088]

[0089] 5. Definition of objective function

[0090]

[0091] In formula (6), (A, B, C, D) is expressed as (A g,m ,B g,m ,C g,m ,D g,m )or or (A g ,B g ,C g ,D g ) or (A * ,B * ,C * ,D * ), where 0≤β≤1, ψ represents the system task completion rate, represents the normalized system unit task delay;

[0092] 6. Position vector D of the drone group g,m Methods for performing mutation and crossover operations

[0093] 1) Mutation operation: According to formula (7), by g,m The position vector D of UAV j g,m (j) Perform mutation processing to obtain the mutation position vector Thus, variant individuals are obtained

[0094]

[0095] In formula (7), ξ1, ξ2, ξ3 are randomly generated integers in the range [1, M], where ξ1≠ξ2≠ξ3≠m;

[0096] 2) Crossover operation: According to formula (8), by g,m The position vector D of UAV j g,m (j) Perform cross processing to obtain the cross position vector Thus, we obtain crossover individuals

[0097]

[0098] In formula (8), rand(0,1) is a random number generated in the range (0,1);

[0099] 7. UAV Network SFC Deployment Algorithm

[0100] 1) Establish a temporary set I′=I, where I is the ground user set, and the task tolerance delay T of each user i is i max The value of , sort the elements in the set I′ in ascending order;

[0101] 2) If Take the first user i from the set I′, and put the virtual network functions VNF required by user i into the set S in sequence according to the task offloading request of user i i , jump to step 3), otherwise, jump to step 9);

[0102] 3) Add the drones in the drone set J that meet the access conditions of user i to the access drone set if Jump to step 4), otherwise, jump to step 2);

[0103] 4) If There is no UAV that meets the first VNF ​​type requirement of SFC, jump to step 5), otherwise, Remove all drones that do not meet the first VNF ​​type requirement of SFC and jump to step 8);

[0104] 5) According to the communication distance between user i and drone j, the set Arrange the elements in ascending order, take out the first drone j in J′, and let user i’s access drone j a =j;

[0105] 6) According to the access drone j a and the UAV network status, call sub-algorithm 1 to obtain the mapping decision of the SFC required by user i {A i ,B i}, if successful, jump to step 7), otherwise, jump to step 2);

[0106] 7) According to the mapping decision {A i ,B i} and the UAV network status, call sub-algorithm 2 to obtain the resource allocation decision of the SFC required by user i {C i}, jump to step 2);

[0107] 8) According to the collection The available computing resources c of each UAV j in j The value of the set The elements in are arranged in descending order, and the ones with the same c are sorted according to the communication distance between user i and drone j. j Arrange the elements of the value in ascending order and take out The first drone j in the list, allowing user i to access drone j a =j, jump to step 6);

[0108] 9) For the available computing resources c in set J j >0, each drone j takes turns to add the minimum unit computing resources to each VNF instance on drone j and update the resource allocation decision {C i}, until c j =0;

[0109] 10) Output SFC deployment decision {A, B, C} = {A i ,B i ,C i | i∈I}, the algorithm ends.

[0110] 8. Sub-algorithm 1: Obtaining SFC mapping decision

[0111] 11) Initialize temporary variable j t =j a , temporary set J′=J, Count variables n=0, l=0;

[0112] 12) Let n = n + 1, and find the VNF type that meets the SFC nth requirement and has available computing resources c in J′. j UAVs with >c0 are added to the set J″, where c0 represents the basic computing resources required for VNF instantiation, and the node mapping cost τ of each UAV in the set J″ is updated. j ;

[0113] 13) If According to τ j , sort the elements in the set J″ in ascending order, take out the first drone j in J″ as the nth VNF mapping node and add it to the node mapping decision A i , the drone j t The minimum cost path to UAV j Join link mapping decision B as the lth virtual link i , let j t = j, remove j from the set J′ and jump to step 14), otherwise, jump to step 15);

[0114] 14) If n<|S i |, let l=l+1, Jump to step 12), otherwise, output the mapping decision of user i {A i ,B i}, jump to step 15);

[0115] 15) The algorithm ends.

[0116] 9. Sub-algorithm 2: Obtaining SFC resource allocation decisions

[0117] 21) Decision A based on node mapping i , add the mapping node to the temporary set J′, and allocate the basic computing resources c0 required for VNF instantiation to each drone in J′, let c j =c j -c0 Decision B based on link mapping i , the bandwidth b required by the virtual link i Add resource allocation decision C i , and obtain the total transmission delay of user i's offloading path in, is the cumulative transmission delay of task data through each physical link on the offloading path, and the counting variable n is set to 0;

[0118] 22) Let the temporary set J″=J′. For each UAV j in the set J′, calculate the task computing resources that UAV j needs to allocate to user i according to the task offloading request made by user i. And add resource allocation decision C i ,make

[0119] 23) If the task calculates the delay Among them, T i max =Tolerance delay of task of user i, jump to step 24), otherwise, output resource allocation decision C i , jump to step 25);

[0120] 24) c j = 0 are removed from the set J″ if According to c j The value of , sort the elements in the set J″ in descending order, and for the first drone j in J″, add a unit of computing resource c1 to the VNF instance of user i and update the resource allocation decision C i , let c j =c j -c1, jump to step 23), otherwise, jump to step 25);

[0121] 25) The algorithm ends.

[0122] A method for deploying SFC in a UAV network based on multifunctional task offloading is specifically implemented by the following steps:

[0123] Step 1: Use differential evolution algorithm to build optimization model. Assume that the number of drones is K, the number of individuals in the population is M≤K!, the mutation scaling factor is ε, the crossover probability is λ, the maximum number of iterations is G, the iteration count variable g=0, the population individual count variable m=0, and use k-means algorithm to generate the 3D position vector set {D g,m | m≤M};

[0124] Step 2: If g≤G, go to step 3; otherwise, go to step 7.

[0125] Step 3: Let m = m + 1. If m ≤ M, according to the current drone group position vector D g,m , call the UAV network service function chain SFC deployment algorithm to obtain the node mapping decision A g,m , Link Mapping Decision B g,m and resource allocation decision C g,m , calculate the objective function value Γ(A g,m ,B g,m ,C g,m ,D g,m ), jump to step 4, otherwise, go to step 6;

[0126] Step 4: D g,m Perform mutation and crossover operations to obtain the position vector of the drone group after mutation and crossover Call the UAV network SFC deployment algorithm to obtain the node mapping decision of the mth mutation crossover individual in the gth generation Link mapping decision and resource allocation decisions Calculate the objective function value

[0127] Step 5: If make Jump to step 3, otherwise, jump to step 3;

[0128] Step 6: From the M individuals of the g-th generation population {D g,m | m≤M} select the solution to obtain the optimal individual (A g ,B g ,C g ,D g )=argmin{Γ(A g,m ,B g,m ,C g,m ,D g,m )| m∈[1,M]}, where A g , B g , C g , Dg are the optimal node mapping decision, link mapping decision, resource allocation decision, and drone swarm position vector in the g-th generation population, respectively. Let g = g + 1, m = 0, and jump to step 2.

[0129] Step 7: Select the optimal individual solution (A * ,B * ,C * ,D * )=argmin{Γ(A g ,B g ,C g ,D g )| g∈[1,G]}, where A * , B * , C * , D * These are the globally optimal node mapping decision, link mapping decision, resource allocation decision, and drone swarm position vector, and the algorithm ends.

[0130] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions.

[0131] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0132] The above embodiments should be understood as merely illustrating the present invention and not as limiting the scope of protection of the present invention. After reading the contents of the present invention, technicians may make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A UAV network SFC deployment method based on multifunctional task offloading, characterized in that: The following steps are involved:

101. Use differential evolution algorithm to build optimization model. Let the number of drones be K, the number of individuals in the population be M, M≤K!, the mutation scaling factor be ε, the crossover probability be λ, the maximum number of iterations be G, let the iteration count variable g = 0, the population individual count variable m = 0, and use k-means algorithm to generate the 3D position vector set {D g,m | m≤M }; 102. If g≤G, go to step 103; otherwise, jump to step 107; 103. Let m = m + 1. If m ≤ M, according to the current position vector D of the drone group g,m , obtain the node mapping decision A through the UAV network service function chain SFC deployment algorithm g,m , Link Mapping Decision B g,m and resource allocation decision C g,m , calculate the objective function value Γ(A g,m ,B g,m ,C g,m ,D g,m ), jump to step 104, otherwise, go to step 106; 104. D g,m Perform mutation and crossover operations to obtain the position vector of the drone group after mutation and crossover Obtain the node mapping decision of the mth mutation crossover individual in the gth generation through the UAV network SFC deployment algorithm Link mapping decision and resource allocation decisions Calculate the objective function value 105. If make Jump to step 103, otherwise, jump to step 103; 106. From the M individuals of the g-th generation population {D g,m | m≤M } select the solution to obtain the optimal individual (A g ,B g ,C g ,D g )=argmin{Γ(A g,m ,B g,m ,C g,m ,D g,m )| m∈ [ 1,M] }, where A g , B g , C g , D g are the optimal node mapping decision, link mapping decision, resource allocation decision, and drone swarm position vector in the g-th generation population, respectively. Let g = g + 1, m = 0, and jump to step 102; 107. The solution of selecting the best individual from a total of G generations of populations (A * ,B * ,C * ,D * )=argmin{Γ(A g ,B g ,C g ,D g )| g∈ [ 1,G] }, where A * , B * , C * , D * They are the globally optimal node mapping decision, link mapping decision, resource allocation decision, and drone swarm position vector, respectively. End; The UAV network SFC deployment algorithm in step 103 and step 104 includes the following steps: 1) Establish a temporary set I′=I, where I is the ground user set, and the task tolerance delay T of each user i is i max The value of , sort the elements in the set I′ in ascending order; 2) If Take the first user i from the set I′, and put the virtual network functions VNF required by user i into the set S in sequence according to the task offloading request of user i i , jump to step 3), otherwise, jump to step 9); 3) Add the drones in the drone set J that meet the access conditions of user i to the access drone set if Jump to step 4), otherwise, jump to step 2); 4) If There is no UAV that meets the first VNF ​​type requirement of SFC, jump to step 5), otherwise, Remove all drones that do not meet the first VNF ​​type requirement of SFC and jump to step 8); 5) According to the communication distance between user i and drone j, the set Arrange the elements in ascending order, take out the first drone j in J′, and let user i’s access drone j a =j; 6) According to the access drone j a and the UAV network status, construct the mapping decision of the SFC required by user i {A i ,B i }, if successful, jump to step 7), otherwise, jump to step 2); 7) According to the mapping decision {A i ,B i } and the UAV network status, construct the resource allocation decision of the SFC required by user i {C i }, jump to step 2); 8) According to the collection The available computing resources c of each UAV j in j The value of the set The elements in are arranged in descending order, and the ones with the same c are sorted according to the communication distance between user i and drone j. j Arrange the elements of the value in ascending order and take out The first drone j in the list, allowing user i to access drone j a =j, jump to step 6); 9) For the available computing resources c in set J j >0, each drone j takes turns to add the minimum unit computing resources to each VNF instance on drone j and update the resource allocation decision {C i }, until c j =0; 10) Output SFC deployment decision {A, B, C} = {A i ,B i ,C i | i∈I }, the algorithm ends; The mapping decision {A} of the SFC required by user i is constructed in step 6) i ,B i } method is as follows: 11) Initialize temporary variable j t =j a , temporary set J′=J, Count variables n=0, l=0; 12) Let n = n + 1, and find the VNF type that meets the SFC nth requirement and has available computing resources c in J′. j UAVs with >c0 are added to the set J″, where c0 represents the basic computing resources required for VNF instantiation, and the node mapping cost τ of each UAV in the set J″ is updated. j ; 13) If According to τ j , sort the elements in the set J″ in ascending order, take out the first drone j in J″ as the nth VNF mapping node and add it to the node mapping decision A i , the drone j t The minimum cost path to UAV j Join link mapping decision B as the lth virtual link i , let j t = j, remove j from the set J′ and jump to step 14), otherwise, jump to step 15); 14) If n<|S i |, let l=l+1, Jump to step 12), otherwise, output the mapping decision of user i {A i ,B i }, jump to step 15); 15) The algorithm ends.

2. The SFC deployment method for UAV network based on multifunctional task offloading according to claim 1 is characterized in that: In step 101, the k-means algorithm is used to generate the 3D position vector set {D g,m | m≤M The methods of} specifically include: Let the 3D position vector of drone j be D g,m (j) = (x j ,y j ,z j ), the drone group position vector D g,m ={D g,m (j)| j∈J }, where (x j ,y j ), z j Denote the horizontal coordinate and vertical coordinate of drone j respectively, J is the drone set, the population individual count variable m≤M, the population iteration number g≤G, let g=0, according to the location distribution of ground users, use the k-means algorithm to find K cluster centers, and randomly deploy K drones carrying different VNF ​​types to the K cluster centers to generate the M individual drone swarm 3D position vector set {D g,m | m≤M }, where the number of individuals in the population M≤K!.

3. The SFC deployment method for UAV network based on multifunctional task offloading according to claim 1 is characterized in that: The node mapping cost τ of the drone in step 12) j The calculation method is shown in formula (1): In formula (1), 0≤α≤1, represents the normalized available computing resources of UAV j, as shown in formula (2), Indicates the path The normalized distance is shown in formula (3).

4. The SFC deployment method for UAV network based on multifunctional task offloading according to claim 1 is characterized in that: In step 7), the resource allocation decision {C i The method comprises the following steps: 21) Decision A based on node mapping i , add the mapping node to the temporary set J′, and allocate the basic computing resources c0 required for VNF instantiation to each drone in J′, let c j =c j -c0 Decision B based on link mapping i , the bandwidth b required by the virtual link i Add resource allocation decision C i , and obtain the total transmission delay of user i's offloading path in, is the cumulative transmission delay of task data through each physical link on the offloading path, and the counting variable n is set to 0; 22) Let the temporary set J″=J′. For each UAV j in the set J′, calculate the task computing resources that UAV j needs to allocate to user i according to the task offloading request made by user i. And add resource allocation decision C i ,make 23) If the task calculates the delay Among them, T i max =Tolerance delay of task of user i, jump to step 24), otherwise, output resource allocation decision C i , jump to step 25); 24) c j = 0 are removed from the set J″ if According to c j The value of , sort the elements in the set J″ in descending order, and for the first drone j in J″, add a unit of computing resource c1 to the VNF instance of user i and update the resource allocation decision C i , let c j =c j -c1, jump to step 23), otherwise, jump to step 25); 25) The algorithm ends.

5. The SFC deployment method for UAV network based on multifunctional task offloading according to claim 4 is characterized in that: In step 22), the task computing resources that drone j needs to allocate to user i are calculated. The method is shown in formula (4): In formula (4), u i represents the amount of task data for user i, w i represents the task complexity of user i.

6. The SFC deployment method for UAV network based on multifunctional task offloading according to claim 4 is characterized in that: The task calculation delay in step 23) The calculation method is shown in formula (5):

7. The SFC deployment method for UAV network based on multifunctional task offloading according to claim 1 is characterized in that: The definition of the objective function in step 103, step 104, step 105, step 106 and step 107 is shown in formula (6): In formula (6), (A, B, C, D) is expressed as (A g,m ,B g,m ,C g,m ,D g,m )or or (A g ,B g ,C g ,D g ) or (A * ,B * ,C * ,D * ), where 0≤β≤1, ψ represents the system task completion rate, Indicates the normalized system unit task latency.

8. The SFC deployment method for UAV network based on multifunctional task offloading according to claim 1 is characterized in that: In step 104, the position vector D of the drone group is g,m The methods for performing mutation and crossover operations are as follows: 1) Mutation operation: According to formula (7), by g,m The position vector D of UAV j g,m (j) Perform mutation processing to obtain the mutation position vector Thus, variant individuals are obtained In formula (7), ξ1, ξ2, ξ3 are randomly generated integers in the range [1, M], where ξ1≠ξ2≠ξ3≠m; 2) Crossover operation: According to formula (8), by g,m The position vector D of UAV j g,m (j) Perform cross processing to obtain the cross position vector Thus, we obtain crossover individuals In formula (8), rand(0,1) is a random number generated in the range of (0,1).

Images