Task migration optimization method and device

Abstract

The present invention provides a task migration optimization method and device, belonging to the field of power network. The method includes: obtaining communication parameters in a power convergence network; based on the communication parameters, establishing a first objective function with the goal of minimizing the task migration delay; decomposing the first objective function into a second objective function and a third objective function; iteratively solving the second objective function and the third objective function until the second objective function and the third objective function converge to obtain an optimal task migration solution; based on the optimal task migration solution, performing task migration in the power convergence network. The present invention can reduce the complexity of optimization solution and improve the efficiency of optimization solution.

Description

Technical Field

[0001] The present invention relates to the technical field of electric power networks, and in particular to a task migration optimization method and device. Background Art

[0002] A large number of intelligent power terminals (IPTs) are deployed in the power convergence network, including devices such as temperature sensors, voltage sensors, and cameras. IPTs will generate a large number of computing tasks. Considering the limited computing power of IPTs themselves, unmanned aerial vehicles (UAVs) with high-speed mobility are usually used to collect the computing tasks of each IPT, and the computing tasks are migrated to the base station (BS) equipped with a mobile edge computing (MEC) server. The MEC server in the base station is used to process the above computing tasks.

[0003] In the prior art, when solving the optimal task migration solution, an objective function is usually established with the goal of minimizing energy consumption, computing resources or communication resources, and a centralized algorithm is used to solve the objective function to obtain the optimal task migration solution. Due to the excessive number of IPTs in the power fusion network, the complexity of using a centralized algorithm to optimize and solve the optimal task migration solution is greatly increased, which occupies too many computing resources, resulting in reduced optimization solution efficiency and slow task migration speed. Summary of the invention

[0004] The embodiment of the present invention provides a task migration optimization method and device to solve the problem in the prior art that the complexity of optimizing and solving the optimal task migration solution is too high, which occupies too many computing resources and leads to reduced optimization and solving efficiency and slow task migration speed.

[0005] In a first aspect, an embodiment of the present invention provides a task migration optimization method, comprising:

[0006] Obtain communication parameters in power convergence networks;

[0007] Based on the communication parameters, a first objective function is established with the goal of minimizing the task migration delay;

[0008] Decomposing the first objective function into a second objective function and a third objective function;

[0009] Iteratively solving the second objective function and the third objective function until both the second objective function and the third objective function converge to obtain an optimal task migration solution;

[0010] Based on the optimal task migration solution, task migration is performed in the power fusion network.

[0011] In a second aspect, an embodiment of the present invention provides a task migration optimization device, comprising:

[0012] An acquisition module, used to acquire communication parameters in the current power convergence network;

[0013] An optimization module, configured to establish a first objective function based on the communication parameters and with the goal of minimizing the task migration delay;

[0014] The optimization module is used to decompose the first objective function into a second objective function and a third objective function;

[0015] The optimization module is further used to iteratively solve the second objective function and the third objective function until the second objective function and the third objective function converge to obtain an optimal task migration solution;

[0016] The migration module is used to perform task migration in the power convergence network based on the optimal task migration solution.

[0017] The embodiment of the present invention provides a task migration optimization method and device, which decomposes a non-convex optimization problem with high solution complexity by decomposing a first objective function into a second objective function and a third objective function, and obtains an optimal task migration solution by iteratively solving the second objective function and the third objective function, which can effectively reduce the solution complexity and improve the optimization solution efficiency. At the same time, the embodiment of the present invention establishes the first objective function with minimizing the task migration delay as the goal, which can effectively reduce the task migration delay. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. 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 creative work.

[0019] Figure 1 It is an application scenario diagram of the task migration optimization method provided by an embodiment of the present invention;

[0020] Figure 2 is a flow chart of an implementation of a task migration optimization method provided by an embodiment of the present invention;

[0021] Figure 3 is a flowchart for implementing iterative solution of the second objective function and the third objective function provided by an embodiment of the present invention;

[0022] Figure 4 is a flowchart for solving the fifth objective function provided by an embodiment of the present invention;

[0023] Figure 5 is a schematic diagram of the relationship between task migration delay and bandwidth under different task migration schemes provided by an embodiment of the present invention;

[0024] Figure 6 It is a structural diagram of a task migration optimization device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0025] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0026] In order to make the purpose, technical solutions and advantages of the present invention more clear, specific embodiments will be described below in conjunction with the accompanying drawings.

[0027] Figure 1 FIG. 1 is an application scenario diagram of the task migration optimization method provided by an embodiment of the present invention. Figure 1 As shown in Figure 1, the power convergence network contains K IPTs, N UAVs, and a base station equipped with a MEC server. Each UAV can be associated with a corresponding IPT. The set of IPTs connected to the nth UAV can be expressed as K n ={1,…k n ,…,K n Each IPT in the power convergence network will generate a task to be processed in each time slot. These tasks can be processed by the IPT itself or transmitted to the BS through the UAV for processing. Therefore, it is necessary to provide a task migration optimization method to optimize and solve an optimal task migration plan to guide the UAV, IPT and BS in the power convergence network to perform task migration.

[0028] Figure 2 The implementation flow chart of the task migration optimization method provided by the embodiment of the present invention is detailed as follows:

[0029] Step 201: Acquire communication parameters in the power convergence network.

[0030] The cycle of the power fusion network is T, which is divided into M time slots, and the duration of each time slot is Δ, then T = MΔ. The set of time slots can be expressed as: kthn The horizontal position of an IPT can be expressed as in, Indicates the kth n The two-dimensional abscissa of an IPT, Indicates the kth n The motion trajectory of the nth UAV in the mth time slot can be expressed as: n (m) = (x n (m),y n (m),H uav ). Among them, x n (m) represents the two-dimensional horizontal coordinate of the nth UAV in the mth time slot, y n (m) represents the two-dimensional ordinate of the nth UAV in the mth time slot, H uav Indicates the flight altitude of each UAV.

[0031] For information transmission between IPTs and UAVs, all UAVs use different frequency bands, and each IPT is orthogonally allocated frequency band resources within the coverage area of ​​each UAV. Therefore, there is no interference in the power fusion network. The wireless channel between UAVs and IPTs can be regarded as a line-of-sight channel. Therefore, the nth UAV and the kth n The channel gain of an IPT in the mth time slot can be expressed as Wherein, h represents the channel gain when the distance is 1m.

[0032] The nth UAV and the kth n The information transmission rate of an IPT in the mth time slot can be expressed as in, Assign the nth UAV to the kth time slot n The bandwidth of each IPT. Indicates the kth n is the transmission power of an IPT, and σ represents the noise power.

[0033] The nth UAV and the kth n The information transmission delay of an IPT in the mth time slot can be expressed as in, Indicates the kth n The data size of the task in the mth time slot of an IPT.

[0034] kth n The transmission energy of an IPT in the mth time slot can be expressed as

[0035] The information transmission delay between the nth UAV and the BS in the mth time slot can be expressed as Among them, Rn,BS (m) represents the data transmission rate between the nth UAV and the BS in the mth time slot.

[0036] kth n The task of an IPT in the mth time slot can be expressed as in, Indicates the calculation of the kth n The computing resources of the task of an IPT in the mth time slot, Indicates the kth n The data size of the task in the mth time slot of an IPT is Indicates the kth n The maximum delay tolerance of the task of an IPT in the mth time slot.

[0037] Task migration marker To characterize the kth n Whether an IPT uses the nth UAV to migrate tasks in the mth time slot. When n The task of an IPT in the mth time slot is transmitted to the MEC server for processing via the nth UAV as a relay. When n The task of an IPT in the mth time slot is handled by the IPT itself.

[0038] When the kth n When the task of the mth time slot of an IPT is handled by the IPT itself, the kth time slot n The local computation delay of an IPT in processing its own task in the mth time slot can be expressed as in, Indicates the computing resources of its own task. The corresponding kth n The energy consumption of an IPT can be expressed as in, Indicates the kth n The energy consumption coefficient of the kth IPT depends on n The chip architecture of the IPT. According to actual measurements, the general value is

[0039] When the kth n When the task of an IPT in the mth time slot is migrated to the MEC server in the BS for processing, the task must first be transmitted to the nth UAV, and then the nth UAV transmits the task to the BS. The MEC server in the BS processes k n The computation time of an IPT task can be expressed as in, Indicates that the MEC server is assigned to the kth time slot in the mth time slot nSince the output data size after calculation is much smaller than the data size before calculation, the data transmission time and energy consumption after calculation are ignored in the embodiment of the present invention.

[0040] α represents the task migration mark of each IPT in each time slot b represents the bandwidth allocated to each IPT by each UAV in each time slot. f represents the computing resources of the MEC server in the BS allocated to each IPT task in each time slot. Q represents the motion trajectory of each UAV in each time slot. n (m) collection.

[0041] Based on the above, the communication parameters in the embodiment of the present invention include but are not limited to: a set of task migration marks of each IPT in each time slot, a set of bandwidths allocated to each IPT by each UAV in each time slot, a set of computing resources of tasks allocated to each IPT by the MEC server in the BS in each time slot, a set of motion trajectories of each UAV in each time slot, a set of motion trajectories of the nth UAV in the mth time slot, and a set of motion trajectories of the nth UAV in the mth time slot. n (m), task migration mark The MEC server allocates the kth n Computational resources for IPT tasks The number of UAVs N in the power convergence network and the number of IPTs K served by the nth UAV n , the nth UAV and the kth n The information transmission delay of an IPT in the mth time slot Information transmission delay between the nth UAV and the BS in the mth time slot The MEC server in the BS processes the kth n The computation time of an IPT task and kth n The local computation delay of an IPT in processing its own task in the mth time slot wait.

[0042] Step 202: Based on the communication parameters, a first objective function is established with the goal of minimizing the task migration delay.

[0043] Based on the above communication parameters,

[0044]

[0045] Calculate the kth n The task migration delay of the task of the IPT in the mth time slot is calculated, so as to establish a first objective function with the goal of minimizing the task migration delay.

[0046] In some embodiments, the

[0047]

[0048] Establish the first objective function.

[0049] Where α represents the set of task migration marks of each IPT in each time slot, b represents the set of bandwidths allocated to each IPT by each UAV in each time slot, f represents the computing resources of the tasks allocated to each IPT by the MEC server in the BS in each time slot, Q represents the set of motion trajectories of each UAV in each time slot, n represents the nth UAV in the power convergence network, N represents the number of UAVs in the power convergence network, and k n Indicates the kth UAV served by the nth UAV n IPT, K n represents the number of IPTs served by the nth UAV, Represents the task migration marker, used to characterize the kth n Whether an IPT uses the nth UAV to migrate tasks in the mth time slot, Indicates the relationship between the nth UAV and the kth n The information transmission delay of an IPT in the mth time slot is: represents the information transmission delay between the nth UAV and the BS in the mth time slot, Indicates that the MEC server in the BS processes the kth n The computation time of an IPT task, Indicates the kth n The local computational delay of an IPT in processing its own task in the mth time slot.

[0050] The constraints of the first objective function are established accordingly:

[0051]

[0052]

[0053]

[0054]

[0055]

[0056]

[0057]

[0058] ‖q n (m+1)-q n (m)‖2 ≤(V max Δ) 2 ,

[0059] q n [1] = q n [M],

[0060]

[0061] in, Indicates the kth n The task migration delay of the task in the mth time slot of an IPT is Indicates the kth n The maximum delay tolerance of the task in the mth time slot of an IPT, B n represents the available bandwidth of the nth UAV, F max Indicates the maximum computing resources of the MEC server in the BS, Indicates the kth n The maximum energy of an IPT, q n (m+1) represents the motion trajectory of the nth UAV in the m+1th time slot, V max Indicates the maximum moving speed of the UAV, V max Δ represents the moving distance of the UAV in each time slot, q n (1) represents the motion trajectory of the nth UAV in the first time slot, q n (M) represents the motion trajectory of the nth UAV in the Mth time slot, q x (m) represents the motion trajectory of the x-th UAV in the m-th time slot, x≠m, d min Indicates the minimum safe distance between UAVs.

[0062] The embodiment of the present invention takes into account that each UAV needs to return to the starting point after a complete time period to serve other IPTs, so the constraint condition q is established. n [1] = q n [M]. The embodiment of the present invention takes into account the safe distance between UAVs and establishes the constraint condition

[0063] Step 203: decompose the first objective function into a second objective function and a third objective function.

[0064] The task migration mark in the first objective function above is a 0,1 integer variable, and the multiple parameters in the first objective function are coupled to each other. Therefore, the first objective function is a non-convex optimization function. When the non-convex optimization function is directly solved, the solution complexity is high. Therefore, the embodiment of the present invention decomposes the first objective function.

[0065] In some embodiments, step 203 may include:

[0066] By fixing the set Q of motion trajectories of each UAV in each time slot in the first objective function, the first objective function is converted into a second objective function;

[0067] By fixing the set α of task migration marks of each IPT in each time slot in the first objective function, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of the task allocated to each IPT by the MEC server in the BS in each time slot, the first objective function is converted into the third objective function.

[0068] When the set Q of motion trajectories of each UAV in each time slot is fixed, the above-mentioned first objective function and its corresponding constraint conditions can be transformed into a second objective function and its corresponding constraint conditions.

[0069] The second objective function and its corresponding constraints are:

[0070]

[0071]

[0072]

[0073]

[0074]

[0075]

[0076]

[0077]

[0078] When the set α of task migration marks of each IPT in each time slot, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of the tasks allocated to each IPT by the MEC server in the BS in each time slot in the first objective function are fixed, the above-mentioned first objective function and its corresponding constraints can be converted into the third objective function and its corresponding constraints.

[0079] The third objective function and its corresponding constraints are:

[0080]

[0081] stC1:

[0082] C2:

[0083] C3:‖q n (m+1)-q n (m)‖ 2 ≤(V max Δ) 2 ,

[0084] C4:q n [1] = q n [M]

[0085] C5:

[0086] Step 204, iteratively solve the second objective function and the third objective function until both the second objective function and the third objective function converge, thereby obtaining an optimal task migration solution.

[0087] The second objective function and the third objective function are both obtained by fixing some parameters in the first objective function. Therefore, the second objective function and the third objective function can be iteratively solved until the second objective function and the third objective function converge, and the optimal solution of the first objective function, that is, the optimal task migration solution, is obtained.

[0088] In some embodiments, see Figure 3 , step 204 may include:

[0089] Step 241, initializing a set Q of motion trajectories of each UAV in each time slot;

[0090] Step 242, applying the set Q of the motion trajectories of each UAV in each time slot to the second objective function, and solving the second objective function to obtain the set α of the task migration mark of each IPT in each time slot, the set b of the bandwidth allocated to each IPT by each UAV in each time slot, and the computing resource f of the task allocated to each IPT by the MEC server in the BS in each time slot;

[0091] Step 243, applying the set α of task migration marks of each IPT in each time slot, the set b of bandwidths allocated to each IPT by each UAV in each time slot, and the computing resources f of the tasks allocated to each IPT by the MEC server in the BS in each time slot to the third objective function, and solving the third objective function to obtain a new set Q of motion trajectories of each UAV in each time slot;

[0092] Step 244 , jump to the step of applying the set Q of motion trajectories of each UAV in each time slot to the second objective function and continue to execute until both the second objective function and the third objective function converge.

[0093] By solving the problem through iterative loop until the second objective function and the third objective function converge, we can obtain the final set α of task migration marks of each IPT in each time slot, the set b of bandwidth allocated to each IPT by each UAV in each time slot, the computing resources f of the task allocated to each IPT by the MEC server in the BS in each time slot, and the set Q of motion trajectories of each UAV in each time slot.

[0094] Based on the final set α of task migration marks of each IPT in each time slot, the set b of bandwidth allocated to each IPT by each UAV in each time slot, the computing resources f of the tasks allocated to each IPT by the MEC server in the BS in each time slot, and the set Q of motion trajectories of each UAV in each time slot, the determined task migration plan is the optimal task migration plan.

[0095] In some embodiments, solving the second objective function to obtain a set α of task migration marks of each IPT in each time slot, a set b of bandwidths allocated by each UAV to each IPT in each time slot, and a computing resource f of tasks allocated by the MEC server in the BS to each IPT in each time slot may include:

[0096] According to the preset conversion rule, the second objective function is converted into the fourth objective function.

[0097] Based on the function variables in the fourth objective function, local variables of each UAV are established.

[0098] Based on local variables, the fourth objective function is transformed into the fifth objective function, and the fifth objective function is solved to obtain the set α of task migration marks of each IPT in each time slot, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of the tasks allocated to each IPT by the MEC server in the BS in each time slot.

[0099] The second objective function is converted into the fourth objective function according to a preset conversion rule, including:

[0100] make The second objective function can be transformed into the fourth objective function. The fourth objective function and its corresponding constraints can be expressed as:

[0101]

[0102]

[0103]

[0104]

[0105]

[0106]

[0107]

[0108]

[0109] Solving the fourth objective function based on the constraint conditions corresponding to the fourth objective function belongs to a convex optimization problem. When the fourth objective function is directly solved in a conventional manner, as the number of IPTs and UAVs increases, the complexity of the solution increases sharply. Therefore, the embodiment of the present invention further transforms the fourth objective function to further reduce the complexity of the solution.

[0110] Based on the fourth objective function and its constraints, the function variables in the fourth objective function include: α, and

[0111] in, Indicates the parameters corresponding to each IPT in each time slot A collection of Indicates that the nth UAV is assigned to the kth UAV in the mth time slot n The bandwidth of each IPT. Indicates the parameters corresponding to each IPT in each time slot A collection of Indicates that the MEC server in the BS is assigned to the kth time slot n The computing resources for the tasks of an IPT.

[0112] Local variables include: local task migration variables, local bandwidth allocation variables and local resource allocation variables.

[0113] In some embodiments, based on the function variables in the fourth objective function, local variables of each UAV are established, including:

[0114] For the nth UAV, according to Establish the local task migration variables of the nth UAV;

[0115] in, represents the local task migration variable of the nth UAV, represents the kth local UAV served in the mth time slot n Task migration marker for each IPT.

[0116] according to determining a local bandwidth allocation variable for the nth UAV;

[0117] in, represents the local bandwidth allocation variable of the nth UAV, Indicates that the local UAV is assigned to the kth time slot in the mth time slot n The bandwidth of each IPT.

[0118] according to Determining a local resource allocation variable for the nth UAV;

[0119] in, Represents the local resource allocation variable of the nth UAV.

[0120] In some embodiments, based on the local variables, converting the fourth objective function into a fifth objective function includes:

[0121] according to Establishing the fifth objective function;

[0122] Among them, g n Represents the local variable of the nth UAV, Indicates the kth n The task migration delay of an IPT in the mth time slot, Δ(g n ) represents the constraint function, when g n ∈Φ, Δ(g n )=0, otherwise, Δ(g n )=+∞, Φ represents a variable set, and Φ satisfies all constraints corresponding to the fourth objective function.

[0123] The variables in the fifth objective function are independent of each UAV, and the variables in the constraints are coupled to each other. Therefore, the embodiment of the present invention adopts the Alternating Direction Method of Multipliers (ADMM) algorithm to solve the fifth objective function to further reduce the complexity of the solution.

[0124] In some embodiments, see Figure 4 , solving the fifth objective function, obtaining the set α of task migration marks of each IPT in each time slot, the set b of bandwidths allocated by each UAV to each IPT in each time slot, and the computing resources f of the tasks allocated by the MEC server in the BS to each IPT in each time slot, may include:

[0125] Step 401, converting the fifth objective function into an augmented Lagrangian function, and determining local variables, global variables and Lagrangian variables in the augmented Lagrangian function;

[0126] Step 402, iteratively updating the local variables, global variables and Lagrangian variables, and solving the fifth objective function based on the updated local variables, global variables and Lagrangian variables until the fifth objective function converges;

[0127] Step 403, when the fifth objective function converges, based on the current local variables and the current global variables, determine the set α of task migration marks of each IPT in each time slot in the second objective function, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of the task allocated to each IPT by the MEC server in the BS in each time slot.

[0128] When using the ADMM algorithm to solve the fifth objective function, the fifth objective function is first converted into an augmented Lagrangian form:

[0129]

[0130] Among them, μ n and v n is the Lagrangian variable, l indicates the local UAV, and is the Lagrangian variable corresponding to the local UAV, Indicates the kth n When the UAV corresponding to the kth IPT is the local UAV, n IPT task migration marker, Indicates the kth n When the UAV corresponding to the IPT is the local UAV, the local UAV is assigned to the kth time slot. n The bandwidth of an IPT. ρ represents the penalty factor. By adjusting the penalty factor, the convergence speed of the ADMM algorithm can be adjusted.

[0131] The local variables in the above augmented Lagrangian form can be expressed as:

[0132]

[0133] in, Represents the local variable g at the τ+1th iteration n , and is the Lagrangian variable corresponding to the local UAV at the τth iteration, represents the kth local UAV served in the mth time slot at the τth iteration. n IPT task migration marker It means that the local UAV is assigned to the kth time slot in the mth time slot at the τth iteration. n Bandwidth of IPT

[0134] Global variables include: α and

[0135] The global variables in the above augmented Lagrangian form can be expressed as:

[0136]

[0137] Among them, α [τ+1] represents the set α of task migration marks of each IPT in each time slot at the τ+1th iteration, Indicates the parameters corresponding to each IPT in each time slot at the τ+1th iteration Collection

[0138] The Lagrangian variables include: μ n and v n .

[0139] The Lagrangian variables in the above augmented Lagrangian form can be expressed as:

[0140]

[0141] in, and is the Lagrangian variable at the τ+1th iteration, and is the Lagrangian variable at the τth iteration, represents the local task migration variable of the nth UAV at the τ+1th iteration represents the local bandwidth allocation variable of the nth UAV at the τ+1th iteration

[0142] The number of UAVs is N. The fifth objective function is decomposed into N independent problems. Each UAV updates its local variables by solving the following function.

[0143]

[0144] stg n ∈Φ

[0145] Solving the above function based on the above constraints to update the local variables is a convex optimization problem. The CVX algorithm can be used to solve the convex optimization problem. Users can also use other algorithms to solve the above convex optimization problem according to their own needs, and the embodiment of the present invention does not specifically limit this.

[0146] When updating global variables, α and The gradient of is set to 0, that is:

[0147]

[0148] Thus we get

[0149] in, represents the kth local UAV served in the mth time slot at the τ+1th iteration n IPT task migration marker Indicates that at the τ+1th iteration, the kth n When the UAV corresponding to the kth IPT is the local UAV, n IPT task migration marker Indicates that the local UAV is assigned to the kth time slot in the mth time slot at the τ+1th iteration n Bandwidth of IPT Indicates that at the τ+1th iteration, the kth n When the UAV corresponding to the IPT is the local UAV, the local UAV is assigned to the kth time slot. n Bandwidth of IPT

[0150] because Therefore, we can get Based on this formula the global variables can be updated.

[0151] When updating the Lagrangian variables, the following formula can be used for updating:

[0152] in, and It represents the Lagrangian variable corresponding to the local UAV at the τ+1th iteration.

[0153] The feasibility condition correlation difference of each UAV can be expressed as: Among them, pri Indicates the difference in feasibility conditions.

[0154] The relative difference of the dual feasibility condition can be expressed as: Among them, dual represents the difference in the dual feasibility condition.

[0155] Each UAV iteratively updates its corresponding local variables, global variables and Lagrangian variables respectively until each UAV satisfies the above feasibility condition related difference and dual feasibility condition related difference. The iteration is completed and the ADMM algorithm terminates.

[0156] It should be noted that in the above iterative update solution process, the parameters is relaxed to: Accordingly, after the ADMM algorithm is completed, it is necessary to Will Restore to integer 0, 1.

[0157] according to The value of g can be determined n , α and Then determine the set α of task migration marks of each IPT in each time slot, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of the tasks allocated to each IPT by the MEC server in the BS in each time slot.

[0158] The embodiment of the present invention takes into account that the ADMM algorithm can further simplify the solution complexity, so as to establish a preset conversion rule to transform the second objective function into the fourth objective function. And on the basis of the fourth objective function, a local variable is established to further transform the fourth objective function into the fifth objective function, so that the fifth objective function can be directly solved by the ADMM algorithm to reduce the solution complexity.

[0159] There are N UAVs and K IPTs in the power convergence network. When the traditional interior point method is used to directly optimize and solve the objective function, the solution complexity is For the method proposed in the embodiment of the present invention, which converts the objective function into the fifth objective function step by step and solves it using the ADMM algorithm, each UAV only needs to calculate the IPT task within its coverage area. For the update of local variables, the complexity is For the update of global variables, the complexity is Where y represents the number of updates of the global variable. For the update of the Lagrangian variable, the complexity is Where z is the number of updates of the Lagrangian variables. Therefore, the complexity of each iteration is It can be seen that the embodiment of the present invention converts the second objective function into the fifth objective function step by step, and the complexity of solving the fifth objective function by using the ADMM algorithm is lower than that of the traditional solution.

[0160] In some embodiments, solving the third objective function includes:

[0161] The third objective function is transformed into a convex optimization function, and the convex optimization function is solved.

[0162] The third objective function and its constraints C1 and C5 are non-convex with respect to parameter Q, but is convex. Considering that the lower bound of the convex problem can be obtained by using the first-order Taylor expansion, the embodiment of the present invention uses the method of successive convex approximation (SCA) to solve the third objective function. Let Indicates the relationship between the nth UAV and the kth n The distance between the IPTs at the τth iteration. n (m), The lower bound of can be expressed as in, Indicates the relationship between the nth UAV and the kth n The information transmission speed of an IPT at the τth iteration is for The first derivative of .

[0163] Considering ||q n (m)-q l (m)|| 2 For parameter Q, which is convex, the embodiment of the present invention uses the SCA method to expand the constraint condition C5 in the third objective function to obtain

[0164] in, represents the motion trajectory of the nth UAV at the τ'th iteration, represents the motion trajectory of the x-th UAV at the τ'th iteration.

[0165] On this basis, the third objective function and its corresponding constraints can be transformed into the following form:

[0166]

[0167]

[0168]

[0169]

[0170]

[0171] q n [1] = q n [M].

[0172] in,

[0173]

[0174] The above objective function This is a convex optimization function. The CVX algorithm can be used to solve the convex optimization function and obtain the set Q of the motion trajectories of each UAV in each time slot.

[0175] Step 205: Perform task migration in the power convergence network based on the optimal task migration solution.

[0176] By iteratively solving the second and third objective functions, we can obtain the set Q of motion trajectories of each UAV in each time slot, the set α of task migration marks of each IPT in each time slot, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of tasks allocated to each IPT by the MEC server in the BS in each time slot, which minimizes the task migration delay, and then determine the optimal task migration scheme. Based on the optimal task migration scheme, the UAV, IPT and MEC server in the power fusion network can be controlled to perform task migration to minimize the task migration delay.

[0177] The embodiments of the present invention respectively verify the relationship between the task migration delay and the bandwidth under different task migration schemes to prove the performance of the task migration scheme determined by the task migration optimization method in the embodiments of the present invention.

[0178] Scheme 1: All IPTs local computation (ASLC) In this scheme, all tasks are processed by the IPT itself.

[0179] Scheme 2: Random task offloading and bandwidth allocation (R-TOBA) In this scheme, all tasks are processed by the MEC server, and the bandwidth is randomly orthogonally allocated to each IPT.

[0180] Solution 3, Penalty block coordinate descent (P-BCD) scheme. In this scheme, the original problem is decomposed into four sub-problems and solved through a cross-iteration algorithm.

[0181] Solution 4 is an optimal task migration solution determined based on the task migration optimization method in the embodiment of the present invention.

[0182] For specific verification results, see Figure 5For the ALSC scheme, since all tasks are processed by the IPT itself, the task migration delay is independent of the bandwidth. Therefore, when the bandwidth increases, the total task migration delay does not change, and the task migration delay is the largest. For the other three schemes, when the bandwidth increases, the total task migration delay decreases. The task migration delay of the task migration scheme provided in the embodiment of the present invention is the smallest. Moreover, compared with Scheme 3, when the bandwidth increases from 5MHz to 5.5MHz, the task migration scheme provided in the embodiment of the present invention can reduce the task migration delay by 3.47%.

[0183] Compared with the prior art, the embodiment of the present invention decomposes the non-convex optimization problem with high solution complexity by decomposing the first objective function into the second objective function and the third objective function, and obtains the optimal task migration solution by iteratively solving the second objective function and the third objective function, which can effectively reduce the solution complexity and improve the optimization solution efficiency. At the same time, the embodiment of the present invention establishes the first objective function with the goal of minimizing the task migration delay, which can effectively reduce the task migration delay.

[0184] Moreover, when solving the second objective function, the embodiment of the present invention can further reduce the complexity of the optimization solution by converting the second objective function into the fifth objective function step by step and solving the fifth objective function by iteratively updating local variables, global variables and Lagrangian variables.

[0185] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0186] The following is an embodiment of the device of the present invention. For details not described in detail, reference may be made to the corresponding method embodiments described above.

[0187] Figure 6 The structure diagram of the task migration optimization device provided by the embodiment of the present invention is shown. For the convenience of explanation, only the part related to the embodiment of the present invention is shown, which is described in detail as follows:

[0188] like Figure 6 As shown, the task migration optimization device 6 includes: an acquisition module 61, an optimization module 62 and a migration module 63.

[0189] An acquisition module 61 is used to acquire communication parameters in the current power convergence network;

[0190] An optimization module 62, configured to establish a first objective function based on the communication parameters and with the goal of minimizing the task migration delay;

[0191] An optimization module 62, configured to decompose the first objective function into a second objective function and a third objective function;

[0192] The optimization module 62 is further used to iteratively solve the second objective function and the third objective function until the second objective function and the third objective function converge to obtain an optimal task migration solution;

[0193] The migration module 63 is used to perform task migration in the power convergence network based on the optimal task migration solution.

[0194] In a possible implementation, the optimization module 62 is used to

[0195] Establish the first objective function;

[0196] Where α represents the set of task migration marks of each IPT in each time slot, b represents the set of bandwidths allocated to each IPT by each UAV in each time slot, f represents the computing resources of the tasks allocated to each IPT by the MEC server in the BS in each time slot, Q represents the set of motion trajectories of each UAV in each time slot, n represents the nth UAV in the power convergence network, N represents the number of UAVs in the power convergence network, and k n Indicates the kth UAV served by the nth UAV n IPT, K n represents the number of IPTs served by the nth UAV, Represents the task migration marker, used to characterize the kth n Whether an IPT uses the nth UAV to migrate tasks in the mth time slot, Indicates the relationship between the nth UAV and the kth n The information transmission delay of an IPT in the mth time slot is: represents the information transmission delay between the nth UAV and the BS in the mth time slot, Indicates that the MEC server in the BS processes the kth n The computation time of an IPT task, Indicates the kth n The local computational delay of an IPT in processing its own task in the mth time slot.

[0197] In a possible implementation, the optimization module 62 is specifically configured to:

[0198] By fixing the set Q of motion trajectories of each UAV in each time slot in the first objective function, the first objective function is converted into a second objective function;

[0199] By fixing the set α of task migration marks of each IPT in each time slot in the first objective function, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of the task allocated to each IPT by the MEC server in the BS in each time slot, the first objective function is converted into the third objective function.

[0200] In a possible implementation, the optimization module 62 is specifically configured to:

[0201] Initialize the set Q of motion trajectories of each UAV in each time slot;

[0202] Apply the set Q of the motion trajectories of each UAV in each time slot to the second objective function, and solve the second objective function to obtain the set α of the task migration mark of each IPT in each time slot, the set b of the bandwidth allocated to each IPT by each UAV in each time slot, and the computing resource f of the task allocated to each IPT by the MEC server in the BS in each time slot;

[0203] Apply the set α of task migration marks of each IPT in each time slot, the set b of bandwidth allocated to IPT by each UAV in each time slot, and the computing resource f of the task allocated to each IPT by the MEC server in the BS in each time slot to the third objective function, and solve the third objective function to obtain the new set Q of motion trajectories of each UAV in each time slot;

[0204] Jump to the step of applying the set Q of motion trajectories of each UAV in each time slot to the second objective function and continue executing until both the second objective function and the third objective function converge.

[0205] In a possible implementation, the optimization module 62 is specifically configured to:

[0206] According to a preset conversion rule, the second objective function is converted into a fourth objective function;

[0207] Based on the function variables in the fourth objective function, establishing local variables of each UAV;

[0208] Based on local variables, the fourth objective function is transformed into the fifth objective function, and the fifth objective function is solved to obtain the set α of task migration marks of each IPT in each time slot, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of the tasks allocated to each IPT by the MEC server in the BS in each time slot.

[0209] In a possible implementation, the function variables in the fourth objective function include: α, and

[0210] in, Indicates the parameters corresponding to each IPT in each time slot A collection of Indicates that the nth UAV is assigned to the kth UAV in the mth time slot n The bandwidth of each IPT; Indicates the parameters corresponding to each IPT in each time slot A collection of Indicates that the MEC server in the BS is assigned to the kth time slot n The computing resources for each IPT task;

[0211] Local variables include: local task migration variables, local bandwidth allocation variables, and local resource allocation variables;

[0212] On this basis, the optimization module 62 is specifically used to:

[0213] For the nth UAV, according to Establish the local task migration variables of the nth UAV;

[0214] in, represents the local task migration variable of the nth UAV, represents the kth local UAV served in the mth time slot n Task migration mark of each IPT;

[0215] according to determining a local bandwidth allocation variable for the nth UAV;

[0216] in, represents the local bandwidth allocation variable of the nth UAV, Indicates that the local UAV is assigned to the kth time slot in the mth time slot n The bandwidth of each IPT;

[0217] according to Determining a local resource allocation variable for the nth UAV;

[0218] in, Represents the local resource allocation variable of the nth UAV.

[0219] In a possible implementation, the optimization module 62 is specifically configured to:

[0220] according to Establishing the fifth objective function;

[0221] Among them, g n Represents the local variable of the nth UAV, Indicates the kthn The task migration delay of an IPT in the mth time slot, Δ(g n ) represents the constraint function, when g n ∈Φ, Δ(g n )=0, otherwise, Δ(g n )=+∞, Φ represents a variable set, and Φ satisfies all constraints corresponding to the fourth objective function.

[0222] In a possible implementation, the optimization module 62 is specifically configured to:

[0223] Convert the fifth objective function into an augmented Lagrangian function, and determine local variables, global variables, and Lagrangian variables in the augmented Lagrangian function;

[0224] Iteratively updating local variables, global variables, and Lagrangian variables, and solving a fifth objective function based on the updated local variables, global variables, and Lagrangian variables until the fifth objective function converges;

[0225] When the fifth objective function converges, based on the current local variables and the current global variables, the set α of task migration marks of each IPT in each time slot in the second objective function, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of the tasks allocated to each IPT by the MEC server in the BS in each time slot are determined.

[0226] In a possible implementation, the optimization module 62 is used to convert the third objective function into a convex optimization function and solve the convex optimization function.

[0227] The embodiment of the present invention decomposes the first objective function into the second objective function and the third objective function through the optimization module 62, thereby decomposing the non-convex optimization problem with high solution complexity. The optimization module 62 iteratively solves the second objective function and the third objective function to obtain the optimal task migration solution, which can effectively reduce the solution complexity and improve the optimization solution efficiency. At the same time, the optimization module 62 establishes the first objective function with the goal of minimizing the task migration delay, which can effectively reduce the task migration delay.

[0228] Furthermore, when solving the second objective function, the optimization module 62 can further reduce the complexity of the optimization solution by converting the second objective function into the fifth objective function step by step and solving the fifth objective function by iteratively updating local variables, global variables and Lagrangian variables.

[0229] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0230] Those of ordinary skill in the art will appreciate that the templates, units, and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0231] If the module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by the processor, the steps of the above-mentioned task migration optimization method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal and software distribution medium.

[0232] The embodiments described above are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A task migration optimization method, characterized in that: include: Obtain communication parameters in the power convergence network; Based on the communication parameters, a first objective function is established with the goal of minimizing the task migration delay; Decomposing the first objective function into a second objective function and a third objective function; Iteratively solving the second objective function and the third objective function until both the second objective function and the third objective function converge to obtain an optimal task migration solution; Based on the optimal task migration solution, performing task migration in the power fusion network; Decomposing the first objective function into a second objective function and a third objective function comprises: By fixing the set Q of motion trajectories of each UAV in each time slot in the first objective function, the first objective function is converted into a second objective function; By fixing the set α of task migration marks of each IPT in each time slot in the first objective function, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of the task allocated to each IPT by the MEC server in the BS in each time slot, the first objective function is converted into the third objective function.

2. The task migration optimization method according to claim 1, characterized in that: The establishing of a first objective function based on the communication parameters with the goal of minimizing the task migration delay includes: according to Establish the first objective function; Where α represents the set of task migration marks of each IPT in each time slot, b represents the set of bandwidths allocated to each IPT by each UAV in each time slot, f represents the computing resources of the tasks allocated to each IPT by the MEC server in the BS in each time slot, Q represents the set of motion trajectories of each UAV in each time slot, n represents the nth UAV in the power convergence network, N represents the number of UAVs in the power convergence network, and k n Indicates the kth UAV served by the nth UAV n IPT, K n represents the number of IPTs served by the nth UAV, Represents the task migration marker, used to characterize the kth n Whether an IPT uses the nth UAV to migrate tasks in the mth time slot, Indicates the relationship between the nth UAV and the kth n The information transmission delay of an IPT in the mth time slot is: represents the information transmission delay between the nth UAV and the BS in the mth time slot, Indicates that the MEC server in the BS processes the kth n The computation time of an IPT task, Indicates the kth n The local computational delay of an IPT in processing its own task in the mth time slot.

3. The task migration optimization method according to claim 1 or 2, characterized in that: The iteratively solving the second objective function and the third objective function until the second objective function and the third objective function converge, comprises: Initialize the set Q of motion trajectories of each UAV in each time slot; Apply the set Q of the motion trajectories of each UAV in each time slot to the second objective function, and solve the second objective function to obtain the set α of the task migration mark of each IPT in each time slot, the set b of the bandwidth allocated to each IPT by each UAV in each time slot, and the computing resource f of the task allocated to each IPT by the MEC server in the BS in each time slot; Apply the set α of task migration marks of each IPT in each time slot, the set b of bandwidths allocated to each IPT by each UAV in each time slot, and the computing resources f of tasks allocated to each IPT by the MEC server in the BS in each time slot to the third objective function, and solve the third objective function to obtain a new set Q of motion trajectories of each UAV in each time slot; Jump to the step of applying the set Q of motion trajectories of each UAV in each time slot to the second objective function and continue executing until both the second objective function and the third objective function converge.

4. The task migration optimization method according to claim 3, characterized in that: Solving the second objective function, obtaining a set α of task migration marks of each IPT in each time slot, a set b of bandwidths allocated by each UAV to each IPT in each time slot, and a computing resource f of tasks allocated by the MEC server in the BS to each IPT in each time slot, including: According to a preset conversion rule, converting the second objective function into a fourth objective function; Establishing local variables of each UAV based on the function variables in the fourth objective function; Based on the local variables, the fourth objective function is converted into the fifth objective function, and the fifth objective function is solved to obtain the set α of task migration marks of each IPT in each time slot, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of the tasks allocated to each IPT by the MEC server in the BS in each time slot.

5. The task migration optimization method according to claim 4, characterized in that: The function variables in the fourth objective function include: α, and in, Indicates the parameters corresponding to each IPT in each time slot A collection of Indicates that the nth UAV is assigned to the kth UAV in the mth time slot n The bandwidth of each IPT; Indicates the parameters corresponding to each IPT in each time slot A collection of Indicates that the MEC server in the BS is assigned to the kth time slot n The computing resources for each IPT task; The local variables include: local task migration variables, local bandwidth allocation variables and local resource allocation variables; The establishing of local variables of each UAV based on the function variables in the fourth objective function includes: For the nth UAV, according to Establish the local task migration variables of the nth UAV; in, represents the local task migration variable of the nth UAV, represents the kth local UAV served in the mth time slot n Task migration mark of each IPT; according to Determining a local bandwidth allocation variable for the nth UAV; in, represents the local bandwidth allocation variable of the nth UAV, Indicates that the local UAV is assigned to the kth time slot in the mth time slot n The bandwidth of each IPT; according to Determining a local resource allocation variable for the nth UAV; in, Represents the local resource allocation variable of the nth UAV.

6. The task migration optimization method according to claim 4, characterized in that: Based on the local variables, converting the fourth objective function into a fifth objective function comprises: according to Establishing the fifth objective function; Among them, g n Represents the local variable of the nth UAV, Indicates the kth n The task migration delay of an IPT in the mth time slot, Δ(g n ) represents the constraint function, when g n ∈Φ, Δ(g n )=0, otherwise, Δ(g n )=+∞, Φ represents a variable set, and Φ satisfies all constraints corresponding to the fourth objective function, represents the local task migration variable of the nth UAV, represents the local bandwidth allocation variable of the nth UAV, represents the local resource allocation variable of the nth UAV, represents the kth local UAV served in the mth time slot n Task migration marker for each IPT.

7. The task migration optimization method according to claim 4, characterized in that: Solving the fifth objective function, obtaining a set α of task migration marks of each IPT in each time slot, a set b of bandwidths allocated by each UAV to each IPT in each time slot, and a computing resource f of tasks allocated by the MEC server in the BS to each IPT in each time slot, including: Converting the fifth objective function into an augmented Lagrangian function, and determining local variables, global variables and Lagrangian variables in the augmented Lagrangian function; Iteratively updating the local variables, the global variables and the Lagrangian variables, and solving the fifth objective function based on the updated local variables, global variables and Lagrangian variables until the fifth objective function converges; When the fifth objective function converges, based on the current local variables and the current global variables, the set α of task migration marks of each IPT in each time slot in the second objective function, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of the tasks allocated to each IPT by the MEC server in the BS in each time slot are determined.

8. The task migration optimization method according to claim 3, characterized in that: Solving the third objective function includes: The third objective function is converted into a convex optimization function, and the convex optimization function is solved.

9. A task migration optimization device, characterized in that: include: An acquisition module, used to acquire communication parameters in the current power convergence network; An optimization module, configured to establish a first objective function based on the communication parameters and with the goal of minimizing the task migration delay; The optimization module is used to decompose the first objective function into a second objective function and a third objective function; The optimization module is further used to iteratively solve the second objective function and the third objective function until the second objective function and the third objective function converge to obtain an optimal task migration solution; Migration module, used to migrate tasks in the power convergence network based on the optimal task migration solution; The step of decomposing the first objective function into a second objective function and a third objective function comprises: By fixing the set Q of motion trajectories of each UAV in each time slot in the first objective function, the first objective function is converted into a second objective function; By fixing the set α of task migration marks of each IPT in each time slot in the first objective function, the set b of bandwidth allocated to each IPT by each UAV in each time slot, and the computing resources f of the task allocated to each IPT by the MEC server in the BS in each time slot, the first objective function is converted into the third objective function.

Images