CN115134370A - Multi-unmanned-aerial-vehicle-assisted mobile edge calculation unloading method - Google Patents

Abstract

The invention discloses a multi-unmanned aerial vehicle assisted mobile edge calculation unloading method, which comprises the following steps: collecting position information of each user on the ground through a software defined network; segmenting different user groups through a clustering algorithm; each unmanned aerial vehicle is set to be in charge of one group; data are transmitted between the users in each area and the corresponding unmanned aerial vehicles in a time division multiple access communication mode, and the data are processed in a partial unloading mode; constructing a mobile edge calculation unloading model; optimizing the constructed mobile edge calculation unloading model to obtain an optimal unmanned aerial vehicle flight track, an optimal calculation task amount of each time slot and an optimal partial unloading variable of each time slot; and carrying out multi-unmanned aerial vehicle-assisted mobile edge calculation unloading according to the optimal unmanned aerial vehicle flight trajectory, the optimal calculation task amount of each time slot and the optimal partial unloading variable of each time slot. The invention can solve the problems of multiple unmanned aerial vehicles and multiple users in a centralized manner, and reduce the energy consumption in the data processing process.

Description

A multi-UAV-assisted offloading method for mobile edge computing

technical field

The invention relates to the technical field of edge computing, in particular to a method for offloading mobile edge computing assisted by multiple unmanned aerial vehicles.

Background technique

A large number of technological products appear in people's lives. However, most technology products are computationally sensitive, and not only have higher requirements on the processing power of the CPU, but also have higher requirements on user experience such as latency, such as smart cities, driverless cars, and virtual reality. (AR) and other applications. Most local users do not have enough computing power to complete the task with the required delay, and may need to consume a lot of energy to complete the computing task, which is very bad for the user experience. Although cloud computing is a solution, if a large number of users choose to upload to cloud computing, it will cause serious network congestion and the user's (Qos) user experience will be very poor. Mobile edge computing is a good potential solution to the above problems. Compared with the traditional communication network architecture, one or more edge servers can be placed at the edge of the network, with both computing resources and storage. resource. The birth of mobile edge computing is not to replace cloud computing, but to complement cloud computing, effectively improving the shortcomings of cloud computing.

Software-defined networking (SDN) can be well combined with edge computing. SDN is a centralized network architecture that organically separates the control layer and data layer of network management. SDN can dynamically collect global network information, collect network data, and globally regulate the deployment and scheduling of the network through the control layer, optimize the network status, and effectively improve the management efficiency of the network.

UAVs are used in various application scenarios in edge computing. UAVs can act as an edge server. In places with a large number of computing-sensitive tasks and densely populated places, UAVs can be deployed in a convenient and flexible way, and can quickly fly to the target location, allowing users to offload computing tasks to unmanned aircraft in a timely manner. Human-computer computing can improve network throughput and effectively improve user experience. The UAV can act as a relay node. Since the UAV is flying in the air, it has good line of sight communication (Line Of Sight) characteristics. It can be used as a bridge between users and edge servers or base stations to restore the network in time. state. Based on software-defined network (SDN), drones and ground users can act as a data layer. SDN can collect network information of drones and users, environmental location, etc., and then use the controller to allocate the trajectory and resources of the drone. and the user's uninstall decision and other variables to optimize. Even with the change of the network environment, the network configuration can be flexibly carried out.

At present, the unloading strategy in a single drone has been fully studied. However, in the case of multiple drones, there is a lack of judgment to make the unloading decision based on the user's location, and some unloading schemes are not considered, so the user needs to consume more energy of. The trajectory scheduling of multiple drones can also make the drones closer to the user and provide better computing services, while reducing the energy consumed by the drone itself. In the application scenario of large-scale users, there is a lack of a centralized controller to collect user and UAV information to make better scheduling strategies and make the user experience better.

SUMMARY OF THE INVENTION

The purpose of the present invention is to overcome the deficiencies of the prior art and provide a multi-UAV-assisted mobile edge computing offloading method.

For achieving the above object, the technical scheme provided by the present invention is:

A multi-UAV-assisted mobile edge computing offloading method, comprising the following steps:

S1. Collect location information of each user on the ground through a software-defined network;

S2. Based on the location information of each user, different user groups are segmented through a clustering algorithm, and one user group is an area;

S3. Set up multiple drones, each drone is responsible for one area;

S4. Data is transmitted between the users in each area and the corresponding drones through time division multiple access communication, and the data is processed by partial unloading, that is, part of the data is processed locally by the user, and the other part of the data is unmanned. processing at the machine;

S5. Build a mobile edge computing offloading model;

S6, optimizing the mobile edge computing unloading model constructed in step S5, to obtain the optimal UAV flight trajectory, the optimal computing task amount for each time slot, and the optimal partial unloading variable for each time slot;

S7. Perform multi-UAV-assisted mobile edge computing offloading according to the optimal UAV flight trajectory, the optimal computing task amount per time slot, and the optimal partial unloading variable of each time slot.

Further, in the step S5, the objective function of the mobile edge computing offloading model is the sum of the energy required by all users to transmit data and the energy required by the user's local computing energy, that is, the following optimization problem P is obtained:

P:

s.t.

C3: l _k [i]>Γ, k∈K

C8: 0＜=ρ _k,u [i]＜=1,k∈K,u∈U,i∈I

C9:C _k ρ _k,u [i]l _k [i]/f _k ＜=δ,k∈K,i∈I

为用户k在第i个时隙内用于本地计算而损耗的能量，

为用户k在第i个时隙内输送数据至无人机u所消耗的能量；In the objective function,

is the energy dissipated by user k for local computation in the ith slot,

is the energy consumed by user k to transmit data to UAV u in the ith time slot;

为无人机u在第i个时隙内为用户k完成卸载任务所需要的能量；

为无人机u在第i个时隙内所需要消耗的飞行能量；τ为电池容量；The constraint C1 is used to ensure that the unloading calculation energy consumption and flight energy consumption of the drone u are less than the capacity of its battery during the working hours; among them, U is the set of drones; K is the set of users; The number of time slots obtained after the average division of the duration T; α _{k, u} is the unloading decision variable, which takes the value of 0 or 1;

is the energy required for the drone u to complete the unloading task for the user k in the ith time slot;

is the flight energy that the drone u needs to consume in the ith time slot; τ is the battery capacity;

The constraint C2 is used to ensure that the amount of data required by the user can be processed within the service duration T; l _k [i] is the amount of calculation that the user k needs to perform in the i-th time slot; L _k is the user k The amount of computation that needs to be completed within the service duration T;

The constraint C3 formula is used to constrain the calculation amount of each time slot of the user must be greater than a certain set value; Γ is the set value;

为无人机u的初始位置；

为无人机u最终时隙停靠的位置；Constraints C4-C5 are used to specify the initial position of UAV u and the final time slot docking position;

is the initial position of the drone u;

is the position where the drone u stops in the final time slot;

为无人机u所能允许达到的最高速度；The constraint C6 is used to ensure that the maximum speed of the drone u cannot exceed the maximum speed that the drone u can allow; δ is the time length of a time slot,

is the maximum speed allowed by the drone u;

Constraint C7 is used to ensure that user k can only select at most one drone as the edge server to be uninstalled in the entire time T;

Constraint C8 is used to ensure that the amount of partial unloading performed by user k cannot exceed the amount of tasks that user _k needs to process in the i-th time slot; Partial factor of the unloading amount at machine u;

Constraint C9 is used to ensure that the time for user k to process part of the local data in the ith time slot cannot exceed the time slot time; C _k is the computer cycle required to calculate 1 bit when user k calculates locally.

Further, in the step S5, the calculation process of the energy lost by the user k for the local calculation in the ith time slot includes:

The calculation time that user k needs to spend according to the ith slot is:

In formula (1), C _k is the computer cycle required to calculate 1 bit when user k calculates locally; f _k is the calculation frequency of user _k ; U is the set of UAVs; L _k [i] is the calculation amount that user k needs to perform in the i-th time slot; ρ _k,u [i] is the i-th time slot that user k transmits to Partial factor of the unloading amount when the drone is u;

According to the energy consumption formula, the energy consumed by user k for local calculation in the i-th time slot is:

In formula (2), γ _{k, u} is the capacitance-related energy consumption factor of the CPU processor transistor of user k.

Further, in the step S5, the calculation process of the energy consumed by the user k to transmit data to the drone u in the ith time slot includes:

After the user k selects the unloading task to the drone u in the initial time slot, the drone is regarded as the unloading object in the whole time T. In one time slot, multiple users transmit data to the drone u. And the time division multiple access communication method is adopted; assuming that W _k,u [i] is the amount of tasks to be uploaded in the i-th time slot, according to the Shannon formula, the transmission data size is:

In formula (3), B is the transmission bandwidth, N ₀ is the Gaussian white noise when transmitting the signal, δ is the time length of a time slot; Ku is the number of users unloaded by the UAV _u , P _k,u is the transmission power of the user from the user k to the UAV u in the i-th time slot, and g _{k, u} is the gain of the corresponding transmission power, which is related to the distance and is defined as

In formula (4), g ₀ is the channel gain at a distance of 1 m, λ is the path fading index, and d _{k, u} is the Euclidean distance from user k to UAV u:

In formula (5), x _k and y _k are the coordinates of the user k, x _u and y _u are the coordinates of the drone u, and H is the height of the drone u from the ground;

As can be seen from the above definition, ignoring the data size required to carry the communication protocol in order to send data packets, there are:

W _k,u =α _k,u (1-ρ _k,u [i])l _k [i],k∈K,u∈U (6)

Perform variable substitution to obtain the transmission power, and then according to the energy equal to the power multiplied by the time, the transmission energy required by the user to send the data packet is obtained as:

K _v the number of users served by each drone;

If the user unloads the task to the drone, and the amount of data that the drone needs to return to the user after processing the data is small, the energy consumption required by the user to accept the data is ignored.

Further, in the step S5, the calculation process of the flight energy that the drone u needs to consume in the ith time slot includes:

In order for the UAV to complete the data transmitted by the user in the previous time slot within one time slot, the UAV needs to meet the calculation frequency of at least f _u [i] or more to complete the calculation unloading within the defined time; In order to satisfy multiple users and ensure the normal progress of tasks, it is necessary to sum the unloaded object tasks:

In formula (8), C _u is the CPU cycle required by the UAV processor to calculate 1 bit;

In the i-th time slot, the energy required by the UAV u to complete the task is:

Considering that the flight energy consumption of the UAV is related to the flight speed of the UAV, it is assumed that the speed of the UAV in a time slot is constant, that is, the speed in a time slot is only related to the speed of the UAV at both ends of the time slot. position, the velocity is:

In formula (10), q _u [i] is the position of the drone u in the i-th time slot;

Therefore, the flight energy that the drone u needs to consume in the ith time slot is:

In formula (11), g is the flight constant related to the force area of the drone, and m is the mass of the drone u.

Further, the step S6 includes:

S6-1, initialize the number of iterations n=0, set the maximum number of iterations N;

S6-2. Calculate the flight trajectory of the UAV;

S6-3, calculate the calculation task amount of each time slot;

S6-4, calculate the partial unloading variable of each time slot;

S6-5. Determine whether n reaches the maximum number of iterations N. If so, output the latest UAV flight trajectory, the amount of computing tasks in each time slot, and the partial unloading variable in each time slot, and use the outputs as the maximum number of iterations. The optimal flight trajectory of the UAV, the optimal amount of computing tasks per time slot, and the optimal partial unloading variable in each time slot, if not, n=n+1, and return to step S6-2.

Further, the step S6-2 includes:

Fixed part of the unloading variable and the unloading data size variable of each time slot user to obtain the P1 problem, which is a convex problem, and the trajectory of the UAV is obtained by solving the CVX toolbox:

Further, the fixed trajectory variables and partial unloading variables are used to obtain the P2 problem, and the CVX toolbox is used to solve the task amount that needs to be calculated for each time slot:

Further, the step S6-4 includes:

Fixing the trajectory variables and the amount of computation per time slot, find the partial unloading variables:

Compared with the prior art, the principle and advantages of this scheme are as follows:

1. Using the software-defined network (SDN) architecture to collect data and make algorithmic decisions, centrally solve the problem of multiple drones and multiple users

2. The users are classified and preprocessed by the global Kmeans algorithm, which not only solves the user's uninstall decision problem, but also reduces the time complexity of the subsequent algorithm.

3. Use the method of block coordinate descent to solve the optimization problem of multi-variable strong coupling, so that the user's energy consumption is reduced.

Description of drawings

In order to illustrate the embodiments of the present invention or the technical solutions in the prior art more clearly, the following briefly introduces the services required in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only For some embodiments of the present invention, for those of ordinary skill in the art, other drawings can also be obtained according to these drawings without creative efforts.

Fig. 1 is the principle flow chart of a kind of multi-UAV-assisted mobile edge computing offloading method of the present invention;

Figure 2 is a data transmission and processing system for multiple drones and multiple users;

FIG. 3 is a schematic diagram of user group division.

Detailed ways

Below in conjunction with specific embodiment, the present invention will be further described:

Due to insufficient computing power, ground users need to partially offload computing tasks to UAVs carrying edge servers by means of wireless communication, that is, UAVs help users perform calculations. Software-Defined Networking (SDN) is responsible for collecting the geographic location information of terrestrial users and the information needed for computing services. In this scenario, it is necessary to make a decision on the unloading strategy of the ground user, that is, which drone the user needs to unload to, and how many tasks need to be unloaded. At the same time, the trajectory of the drone needs to be scheduled so that the user's The energy consumption and calculation energy of the transmission task are minimized. In order to solve this multivariate strong coupling problem, the global K-means mean clustering algorithm is used to solve the one-to-many problem between the UAV and the user. The problem in turn becomes a convex optimization problem to be solved.

As shown in FIG. 1 , the method for offloading mobile edge computing assisted by multiple UAVs described in this embodiment includes the following steps:

S1. Collect the location information of each user on the ground through a software-defined network; the system shown in Figure 2 is used, where UAVs are unmanned aerial vehicles, and SDN controller is the controller;

S2. Based on the location information of each user, different user groups are segmented through a clustering algorithm, and one user group is an area, as shown in Figure 3;

S3. Set up multiple drones, each drone is responsible for one area;

S4. Data is transmitted between the users in each area and the corresponding drones through time division multiple access communication, and the data is processed by partial unloading, that is, part of the data is processed locally by the user, and the other part of the data is unmanned. processing at the machine;

S5. Build a mobile edge computing offloading model;

Among them, the objective function of the mobile edge computing offloading model is the sum of the energy required by all users to transmit data and the energy required by the user's local computing energy, that is, the following optimization problem P is obtained:

P:

s.t.

C3: l _k [i]>Γ, k∈K

C8: 0＜=ρ _k,u [i]＜=1,k∈K,u∈U,i∈I

C9:C _k ρ _k,u [i]l _k [i]/f _k ＜=δ,k∈K,i∈I

为用户k在第i个时隙内用于本地计算而损耗的能量，

为用户k在第i个时隙内输送数据至无人机u所消耗的能量；In the objective function,

is the energy dissipated by user k for local computation in the ith slot,

is the energy consumed by user k to transmit data to UAV u in the ith time slot;

为无人机u在第i个时隙内为用户k完成卸载任务所需要的能量；

为无人机u在第i个时隙内所需要消耗的飞行能量；τ为电池容量；The constraint C1 is used to ensure that the unloading calculation energy consumption and flight energy consumption of the drone u are less than the capacity of its battery during the working hours; among them, U is the set of drones; K is the set of users; The number of time slots obtained after the average division of the duration T; α _{k, u} is the unloading decision variable, which takes the value of 0 or 1;

is the energy required for the drone u to complete the unloading task for the user k in the ith time slot;

is the flight energy that the drone u needs to consume in the ith time slot; τ is the battery capacity;

The constraint C2 is used to ensure that the amount of data required by the user can be processed within the service duration T; l _k [i] is the amount of calculation that the user k needs to perform in the i-th time slot; L _k is the user k The amount of computation that needs to be completed within the service duration T;

The constraint C3 formula is used to constrain the calculation amount of each time slot of the user must be greater than a certain set value; Γ is the set value;

为无人机u的初始位置；

为无人机u最终时隙停靠的位置；Constraints C4-C5 are used to specify the initial position of UAV u and the final time slot docking position;

is the initial position of the drone u;

is the position where the drone u stops in the final time slot;

为无人机u所能允许达到的最高速度；The constraint C6 is used to ensure that the maximum speed of the drone u cannot exceed the maximum speed that the drone u can allow; δ is the time length of a time slot,

is the maximum speed allowed by the drone u;

Constraint C7 is used to ensure that user k can only select at most one drone as the edge server to be uninstalled in the entire time T;

Constraint C8 is used to ensure that the amount of partial unloading performed by user k cannot exceed the amount of tasks that user _k needs to process in the i-th time slot; Partial factor of the unloading amount at machine u;

Constraint C9 is used to ensure that the time for user k to process part of the local data in the ith time slot cannot exceed the time slot time; C _k is the computer cycle required to calculate 1 bit when user k calculates locally.

In the above, the calculation process of the energy lost by user k for local calculation in the ith time slot includes:

The calculation time that user k needs to spend according to the ith slot is:

In formula (1), C _k is the computer cycle required to calculate 1 bit when user k calculates locally; f _k is the calculation frequency of user _k ; U is the set of UAVs; L _k [i] is the calculation amount that user k needs to perform in the i-th time slot; ρ _k,u [i] is the i-th time slot that user k transmits to Partial factor of the unloading amount when the drone is u;

According to the energy consumption formula, the energy consumed by user k for local calculation in the i-th time slot is:

In formula (2), γ _{k, u} is the capacitance-related energy consumption factor of the CPU processor transistor of user k.

In the above, the calculation process of the energy consumed by the user k to transmit data to the drone u in the ith time slot includes:

After the user k selects the unloading task to the drone u in the initial time slot, the drone is regarded as the unloading object in the whole time T. In one time slot, multiple users transmit data to the drone u. And the time division multiple access communication method is adopted; assuming that W _k,u [i] is the amount of tasks to be uploaded in the i-th time slot, according to Shannon's formula, the transmission data size is:

In formula (3), B is the transmission bandwidth, N ₀ is the Gaussian white noise when transmitting the signal, δ is the time length of a time slot; Ku is the number of users unloaded by the UAV _u , P _k,u is the transmission power of the user from the user k to the UAV u in the i-th time slot, and g _{k, u} is the gain of the corresponding transmission power, which is related to the distance and is defined as

In formula (4), g ₀ is the channel gain at a distance of 1 m, λ is the path fading index, and d _{k, u} is the Euclidean distance from user k to UAV u:

In formula (5), x _k and y _k are the coordinates of the user k, x _u and y _u are the coordinates of the drone u, and H is the height of the drone u from the ground;

As can be seen from the above definition, ignoring the data size required to carry the communication protocol in order to send data packets, there are:

W _k,u =α _k,u (1-ρ _k,u [i])l _k [i],k∈K,u∈U (6)

Perform variable substitution to obtain the transmission power, and then according to the energy equal to the power multiplied by the time, the transmission energy required by the user to send the data packet is obtained as:

K _v the number of users served by each drone;

If the user unloads the task to the drone, and the amount of data that the drone needs to return to the user after processing the data is small, the energy consumption required by the user to accept the data is ignored.

In the above, the calculation process of the flight energy that the drone u needs to consume in the ith time slot includes:

In order for the UAV to complete the data transmitted by the user in the previous time slot within one time slot, the UAV needs to meet the calculation frequency of at least f _u [i] or more to complete the calculation unloading within the defined time; In order to satisfy multiple users and ensure the normal progress of tasks, it is necessary to sum the unloaded object tasks:

In formula (8), C _u is the CPU cycle required by the UAV processor to calculate 1 bit;

In the i-th time slot, the energy required by the UAV u to complete the task is:

Considering that the flight energy consumption of the UAV is related to the flight speed of the UAV, it is assumed that the speed of the UAV in a time slot is constant, that is, the speed in a time slot is only related to the speed of the UAV at both ends of the time slot. position, the velocity is:

In formula (10), q _u [i] is the position of the drone u in the i-th time slot;

Therefore, the flight energy that the drone u needs to consume in the ith time slot is:

In formula (11), g is the flight constant related to the force area of the drone, and m is the mass of the drone u.

S6, optimizing the mobile edge computing unloading model constructed in step S5, to obtain the optimal UAV flight trajectory, the optimal computing task amount for each time slot, and the optimal partial unloading variable for each time slot;

This step specifically includes:

S6-1, initialize the number of iterations n=0, set the maximum number of iterations N;

S6-2. Calculate the flight trajectory of the UAV:

Fixed part of the unloading variable and the unloading data size variable of each time slot user to obtain the P1 problem, which is a convex problem, and the trajectory of the UAV is obtained by solving the CVX toolbox:

S6-3. Calculate the amount of computing tasks in each time slot:

Fixed trajectory variables and partial unloading variables get the P2 problem, and the CVX toolbox is used to solve the task amount that needs to be calculated for each time slot:

S6-4. Calculate the partial unloading variable of each time slot:

Fixing the trajectory variables and the amount of computation per time slot, find the partial unloading variables:

S6-5. Determine whether n reaches the maximum number of iterations N. If so, output the latest UAV flight trajectory, the amount of computing tasks in each time slot, and the partial unloading variable in each time slot, and use the outputs as the maximum number of iterations. The optimal flight trajectory of the UAV, the optimal amount of computing tasks per time slot, and the optimal partial unloading variable in each time slot, if not, n=n+1, and return to step S6-2.

S7. Perform multi-UAV-assisted mobile edge computing offloading according to the optimal UAV flight trajectory, the optimal computing task amount per time slot, and the optimal partial unloading variable of each time slot.

The above-mentioned embodiments are only preferred embodiments of the present invention, and are not intended to limit the scope of implementation of the present invention. Therefore, any changes made according to the shape and principle of the present invention should be included within the protection scope of the present invention.

Claims (9)

1. A multi-unmanned aerial vehicle assisted mobile edge computing unloading method is characterized by comprising the following steps:

s1, collecting the position information of each user on the ground through a software defined network;

s2, based on the position information of each user, segmenting different user groups through a clustering algorithm, wherein one user group is an area;

s3, arranging a plurality of unmanned aerial vehicles, wherein each unmanned aerial vehicle is in charge of an area;

s4, data are transmitted between the users in each area and the corresponding unmanned aerial vehicles in a time division multiple access communication mode, and data are processed in a partial unloading mode, namely, part of data are processed locally by the users, and the other part of data are processed by the unmanned aerial vehicles;

s5, constructing a mobile edge calculation unloading model;

s6, optimizing the moving edge calculation unloading model constructed in the step S5 to obtain an optimal unmanned aerial vehicle flight path, an optimal calculation task amount of each time slot and an optimal partial unloading variable of each time slot;

and S7, carrying out multi-unmanned aerial vehicle assisted mobile edge calculation unloading according to the optimal unmanned aerial vehicle flight path, the optimal calculation task amount of each time slot and the optimal partial unloading variable of each time slot.

2. The method for offloading computation of mobile edge assisted by multiple drones according to claim 1, wherein in step S5, the objective function of the model for offloading computation of mobile edge is the sum of energy required for data transmission of all users and energy required for local computation of energy of users, which is to say, the following optimization problem P is obtained:

P:

s.t.

C1:

C2:

C3:l _k [i]＞Γ，k∈K

C4:

C5:

C6:

C7:

C8:0＜＝ρ _k,u [i]＜＝1,k∈K,u∈U,i∈I

C9:C _k ρ _k,u [i]l _k [i]/f _k ＜＝δ,k∈K,i∈I

in the objective function, the target function is,

the energy lost for user k to use for local computation in the ith slot,

energy consumed by user k to transmit data to unmanned aerial vehicle u in ith time slot;

the constraint C1 type is used for ensuring that the energy consumption of unloading calculation and flight energy consumption of the unmanned aerial vehicle u are smaller than the capacity of a battery of the unmanned aerial vehicle u in the working time; wherein U is a set of unmanned aerial vehicles; k is a set of users; i denotes to average the service duration TThe number of the time slots obtained after the averaging; alpha is alpha _k,u Taking the value of 0 or 1 for the unloading decision variable;

the energy required by the unmanned aerial vehicle u to complete the unloading task for the user k in the ith time slot;

flight energy required to be consumed by the unmanned aerial vehicle u in the ith time slot; τ is the battery capacity;

constraint C2 is used to ensure that the data amount required by the user can be processed within the service duration T; l _k [i]Calculating the amount of calculation required to be carried out in the ith time slot for the user k; l is a radical of an alcohol _k Calculating the amount of calculation required to be completed by a user k within a service time length T;

constraint C3 is used to constrain the amount of computation per time slot of the user to be greater than a set value; gamma is a set value;

constraints C4-C5 are used to specify the initial position of drone u and the position of the final slotted landing;

is the initial position of drone u;

the position of the unmanned plane u for the final time slot stop;

constraint C6 is used to ensure that the maximum speed of drone u cannot exceed the maximum speed that drone u can allow to reach; delta is the length of time of one time slot,

the highest speed that the unmanned plane u can allow to reach;

constraint C7 is used to ensure that user k can only select one drone at most as an edge server to choose to offload during the entire time T;

constraint C8 is used to ensure that user k partially offloadsThe amount of the user k cannot exceed the task amount required to be processed by the user k in the ith time slot; rho _k,u [i]The partial factors of the unloading amount when the user k transmits to the unmanned plane u in the ith time slot are shown;

constraint C9 is used to ensure that the time for user k to process partial local data at the ith slot cannot exceed the slot time; c _k The computer period required for calculating 1bit when the user k calculates locally.

3. The method of claim 2, wherein in step S5, the step of calculating the energy consumed by user k for local calculation in the ith time slot includes:

the computation time that the user k needs to spend according to the ith time slot is as follows:

in the formula (1), C _k Calculating a computer period required by 1bit during local calculation for a user k; f. of _k Calculating a frequency for user k; alpha is alpha _k,u Taking the value of 0 or 1 for the unloading decision variable; k is a set of users; u is the set of unmanned aerial vehicles; l is _k [i]The calculated amount needed to be calculated for the user k in the ith time slot; rho _k,u [i]Partial factors of the unloading capacity when the user k in the ith time slot transmits to the unmanned plane u;

according to the energy consumption formula, the energy consumed by the user k for local calculation in the ith time slot is obtained as follows:

in the formula (2), gamma _k,u The capacitance dependent energy dissipation factor of the CPU processor transistor for user k.

4. The method of claim 2, wherein the step S5 of calculating the energy consumed by the user k to deliver data to the drone u in the ith time slot includes:

after a user k selects an unloading task to an unmanned aerial vehicle u in an initial time slot, taking the unmanned aerial vehicle as an unloading object in the whole T time, transmitting data to the unmanned aerial vehicle u by a plurality of users in one time slot, and adopting a time division multiple access communication mode; suppose W _k,u [i]Selecting the uploaded task amount in the ith time slot, and according to a shannon formula, the size of the transmitted data is as follows:

in formula (3), B is the transmission bandwidth, N ₀ Is white gaussian noise when transmitting signals, delta is the time length of one time slot; k _u Number of users, P, responsible for offloading for UAV u _k,u For the transmission power of user k to drone u in ith slot, g _k,u The gain for the corresponding transmission power, which is distance-dependent, is defined as

In the formula (4), g ₀ Is the channel gain at a distance of 1m, λ is the path fading index, d _k,u The Euclidean distance from the user k to the unmanned plane u is as follows:

in the formula (5), x _k And y _k As the coordinates of user k, x _u And y _u Is the coordinate of the unmanned plane u, and H is the height of the unmanned plane u from the ground;

ignoring the data size needed to carry the communication protocol in order to send a packet, there are:

W _k,u ＝α _k,u (1-ρ _k,u [i])l _k [i],k∈K,u∈U (6)

carrying out variable substitution to obtain transmission power, and then obtaining the transmission energy required by a user for sending a data packet according to the fact that the energy is equal to the power multiplied by time:

K _v the number of users served for each drone;

if the data amount required to be returned to the user after the unmanned aerial vehicle processes the data is small after the user unloads the task to the unmanned aerial vehicle, the energy consumption required by the user to receive the data is ignored.

5. The method of claim 2, wherein in step S5, the calculation process of the flight energy required to be consumed by the drone u in the ith time slot includes:

in order for an unmanned aerial vehicle to complete data transmitted in the last time slot by a user within the time of one time slot, the unmanned aerial vehicle needs to satisfy at least f _u [i]The calculation frequency is above, the calculation unloading can be completed within the defined time; in order to satisfy multiple users and ensure the normal operation of tasks, the unloaded object tasks need to be summed:

in formula (8), C _u Calculating the CPU period required by 1bit for the unmanned aerial vehicle processor;

in the ith time slot, the energy required by the unmanned plane u to complete the task is as follows:

considering that flight energy consumption of the unmanned aerial vehicle is related to the flight speed of the unmanned aerial vehicle, assuming that the speed of the unmanned aerial vehicle in a time slot is unchanged, that is, the speed in a time slot is only related to the positions of the unmanned aerial vehicles at the two ends of the time slot, the speed is:

in the formula (10), q _u [i]The position of the unmanned plane u in the ith time slot is determined;

therefore, the flight energy that the unmanned plane u needs to consume in the ith time slot is:

in the formula (11), g is a flight constant related to the stress area of the unmanned aerial vehicle, and m is the mass of the unmanned aerial vehicle u.

6. The method of claim 2, wherein the step S6 includes:

s6-1, setting the number of initialization iterations N as 0, and setting the maximum number of iterations N;

s6-2, calculating the flight track of the unmanned aerial vehicle;

s6-3, calculating the calculation task amount of each time slot;

s6-4, calculating each time slot partial unloading variable;

and S6-5, judging whether N reaches the maximum iteration number N, if so, outputting the latest flight path of the unmanned aerial vehicle, the calculation task amount of each time slot and each time slot partial unloading variable, and outputting the output to be used as the optimal flight path of the unmanned aerial vehicle, the optimal calculation task amount of each time slot and the optimal each time slot partial unloading variable respectively, if not, N is N +1, and returning to the step S6-2.

7. The method of claim 6, wherein the step S6-2 includes:

the fixed partial offload variable and the offload data size variable for each time slot user result in a P1 problem, which is a convex problem that is solved by the CVX toolbox to arrive at the trajectory of the drone:

8. the method of claim 6, wherein the step S6-3 includes:

and (3) obtaining a P2 problem by fixing the track variables and the partial unloading variables, and solving by a CVX tool box to obtain the task quantity required to be calculated in each time slot:

9. the method of claim 6, wherein the step S6-4 comprises:

and (3) fixing the track variable and the calculated amount of each time slot, and solving a partial unloading variable:

Images