Joint trajectory and deep learning task optimization method for multiple unmanned aerial vehicles

Abstract

The invention relates to a joint trajectory and deep learning task optimization method for multiple unmanned aerial vehicles, and belongs to the field of communication, and the method comprises the steps: initializing the configuration information of a ground terminal and unmanned aerial vehicles, setting an iteration frequency variable, and initializing the trajectory of the unmanned aerial vehicles; obtaining an initial target function value, and optimizing the link between the unmanned aerial vehicle and the ground terminal, the trajectory of the unmanned aerial vehicle and the calculation resource allocation of the unmanned aerial vehicle; and judging whether the target function value obtained during the ith iteration reaches a preset target value, if so, obtaining the optimal link between the unmanned aerial vehicle and the ground terminal, the optimal computing resource allocation of the unmanned aerial vehicle, the optimal track of the unmanned aerial vehicle and the minimum energy consumption value of the unmanned aerial vehicle, and if not, continuing iteration. The method combines mobile edge calculation, joint optimization of unmanned aerial vehicle tracks and dynamic unloading of deep learning tasks, and makes full use of multi-unmanned aerial vehicle edge calculation resources, so that the wireless communication rate is maximized while the energy consumption of the unmanned aerial vehicle is minimized, and the communication energy efficiency of the unmanned aerial vehicle and the ground terminal is improved.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a joint trajectory and deep learning task optimization method for multiple unmanned aerial vehicles. Background Art

[0002] In recent years, drones have been widely used in various fields such as combat, surveillance, media and rescue. Especially in the field of wireless communications, drones can build efficient ground communication systems and fly quickly to designated locations with their superior maneuverability and high flexibility. Since drones fly at a low altitude and are close to ground terminals, they can reduce distortion caused by long channels. Drones can also flexibly select locations in three-dimensional space according to needs and establish a communication network with ground terminals. In addition, when there are a large number of intensive ground terminal tasks, edge computing can be performed by drones to process ground terminal tasks, and each ground terminal task can be offloaded to drones to improve work efficiency. In summary, in natural disasters such as earthquakes where a large amount of infrastructure is destroyed, wireless communications supported by drones can greatly improve the accuracy, reliability and speed of disaster relief.

[0003] Despite these advantages, the air-to-ground wireless communication network constructed by drones has four unresolved technical problems. First, unlike base stations, drones are not fixed on the ground during flight and are easily affected by environmental factors such as strong winds in the air, which may lead to unstable communication links and information distortion, reducing the reliability of ground-to-air communication; second, drones move in three-dimensional space and are also moving when transmitting information to ground terminals. The communication link is always affected by the trajectory of the drone, so the optimization design of the trajectory of the drone is very difficult; at the same time, during information transmission and task processing, drones and ground terminals consume a lot of energy, and the energy consumption of drones during flight and information transmission is affected by various factors; finally, during task unloading, it is not easy to determine how many tasks to unload to drones, and when to complete the entire work process efficiently and quickly. Usually, the above problems are interrelated and cannot be solved at the same time. Therefore, how to optimize the airspace communication system to achieve the best effect is a very important issue. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of the prior art and provide a joint trajectory and deep learning task optimization method for multiple UAVs, thereby solving the problems existing in the prior art.

[0005] The object of the present invention is achieved by the following technical solution: a joint trajectory and deep learning task optimization method for multiple unmanned aerial vehicles, the method comprising:

[0006] Initialize the configuration information of the ground terminal and the drone, set the iteration number variable, and initialize the drone trajectory;

[0007] According to the initialization conditions, the initial objective function value is obtained, and the energy consumption problem of the UAV is described as And set constraints, and optimize the connection between the UAV and the ground terminal, optimize the UAV trajectory, and optimize the allocation of UAV computing resources;

[0008] Determine whether the objective function value obtained at the i-th iteration reaches the preset target value E val , if reached, then the optimal link α between the UAV and the ground terminal is obtained un , Optimal UAV computing resource allocation Optimal drone trajectory And the minimum drone energy consumption value, otherwise continue to iterate.

[0009] The setting of the iteration number variable and the initialization of the drone trajectory include:

[0010] Set the iteration number variable i with an initial value of 1;

[0011] The initial trajectory of the drone is set to an elliptical trajectory with the head and tail connected, and the drone maintains a constant speed during the flight. The starting position of the initial drone trajectory is The position of the u-th UAV in the m-th segment is

[0012] The obtaining of the initial objective function value comprises:

[0013] Get the propulsion energy consumption of the u-th UAV in the m-th segment of flight

[0014] Get the energy consumption E generated by the u-th UAV serving the n-th ground terminal un ;

[0015] Get the total energy consumption of the u-th drone Combined with the initialization trajectory of the u-th UAV Get the energy consumption of the initialized drone Among them, β is the weight value.

[0016] The energy consumption problem of drones is described as And set constraints including:

[0017] The energy consumption problem of drones is described as

[0018] Set the link limit of the ground base station to That is, a base station can only select one drone for connection;

[0019] Set the maximum value of the drone computing resources to

[0020] Set the service demand constraint for each task as and That is, the delay time must be less than the task deadline, S n is the amount of data for the intermediate results, D off The amount of tasks assigned to the edge for ground terminals, is the channel rate, t s is the gap duration;

[0021] Set the starting point of the drone trajectory to

[0022] Set the flight speed of the drone to

[0023] The optimization of the link between the drone and the ground terminal includes:

[0024] The problem Disassemble and follow the given drone trajectory Use convex optimization tools to calculate the optimal link between the UAV and the ground terminal at the i-th iteration α un [i], get the decomposed sub-problems for And solve it using convex optimization technology through CVX, and finally get the suboptimal solution α at the i-th iteration un [i], and the following constraints are met:

[0025] The link limit of the ground base station is That is, a base station can only select one drone to connect to; the service demand constraint for each task is and That is, the delay time must be less than the task deadline, where M n The number of time slots required for the ground terminal to complete the local deep learning task, M' n The number of time slots required for the nth ground terminal to upload the intermediate results to the UAV.

[0026] The optimization of the UAV trajectory includes:

[0027] The problem Disassemble and obtain the disassembled sub-problems for And the following constraints are met:

[0028] The service demand constraint for each task is and That is, the delay time must be less than the task deadline; the starting point of the drone trajectory is The flight speed of the drone is θ represents the relaxation parameter.

[0029] For the channel rate in the constraint condition Convert to get To achieve suboptimal solution using convex optimization tools, is the intermediate parameter for formula simplification, constant Constant A<0, h m is the altitude of the drone in the current time slot, d nm is the distance between the UAV and the nth ground terminal in the current time slot, N0 is the noise variable, P is the fixed transmission power of the UAV, B is the bandwidth, and c is the speed of light;

[0030] The channel rate is further converted into And order Simplified

[0031] Select a local point Substitution Convert it into a convex form and perform Taylor expansion to obtain Will and Substitution and The new constraints are obtained and

[0032] The optimization of the UAV trajectory includes:

[0033] The problem Disassemble and obtain the disassembled sub-problems for The optimal UAV trajectory is obtained directly through the CVX tool And satisfy the following constraints:

[0034] The starting point of the drone trajectory is The flight speed of the drone is θ represents the relaxation parameter; and

[0035] The optimization of the allocation of UAV computing resources includes:

[0036] According to the given link between the drone and the ground terminal α un [i], through the formula Using CVX to calculate the optimal suboptimal solution for the allocation of UAV computing resources at the i-th iteration

[0037] The present invention has the following advantages:

[0038] 1. Considering the construction of an air-to-ground communication system with multiple drones, more computing resources can be allocated to ground terminals to offload tasks, which can solve the dense ground terminal task requirements more quickly.

[0039] 2. Offload part of the deep learning tasks of the ground terminal to the drone for calculation, make full use of the drone's computing resources, and reduce the task pressure of the ground terminal.

[0040] 3. Jointly optimize the connection between multiple drones and ground terminals, the trajectory of drones, and the problem of deep learning task offloading, so as to minimize the energy consumption of drones while maximizing the wireless communication rate and improve the communication energy efficiency between drones and ground terminals. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic diagram of the scene model structure of the present invention;

[0042] Figure 2 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The components of the embodiments of the present application described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of the present application provided below in conjunction with the drawings is not intended to limit the scope of protection of the application claimed for protection, but merely represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present application. The present invention is further described below in conjunction with the drawings.

[0044] like Figure 1As shown, the present invention specifically relates to a joint trajectory and deep learning task optimization method in a multi-UAV system based on mobile edge computing. The multi-UAV communication system consists of a ground terminal and an edge computing server mounted on a UAV. Each ground terminal has a certain computing power and each UAV has relatively more computing resources, which can provide services for the ground terminal, so the ground terminal can partially offload the deep learning task to the UAV for completion. .

[0045] The present invention jointly optimizes the connection between the UAV and the ground terminal, the UAV trajectory and the allocation of UAV computing resources to minimize the energy consumption of the UAV. The energy consumption problem of the UAV can be described as

[0046]

[0047]

[0048]

[0049]

[0050]

[0051]

[0052]

[0053] exist In the formula (1.1), the connection restriction of the ground base station is specified, that is, a base station can only select one UAV for connection; formula (1.2) specifies the maximum value of the UAV computing resources; formulas (1.3) and (1.4) are constraints on the service requirements of each task, that is, the delay time must be less than the task deadline, S n is the data volume of the intermediate result; formula (1.5) specifies the starting point of the UAV trajectory; formula (1.6) limits the flight speed of the UAV.

[0054] Because of the problem The objective function is a non-convex function and cannot be solved directly. The present invention proposes a block coordinate descent algorithm to iteratively optimize the objective function, optimizing only one variable at a time to reduce the complexity of the problem. The present invention decomposes the target problem into three sub-problems and solves them respectively through convex optimization. After iteration, the suboptimal solution of each variable can be obtained, and finally the minimum value of the energy consumption of the drone is obtained.

[0055] according to Related to the trajectory of the drone, E unIt is related to the connection between the drone and the ground terminal and the allocation of the drone's computing resources. At the same time, the drone's trajectory is also subject to the constraints of the connection between the drone and the terminal, which leads to the non-convexity of the target problem. First, the two energy consumption values ​​are formally expanded:

[0056]

[0057]

[0058] The target problem is decomposed into two sub-problems. Problem one is to optimize the connection between the UAV and the ground terminal and the allocation of UAV computing resources, and problem two is to optimize the UAV trajectory.

[0059] like Figure 2 As shown, the specific process of the present invention includes:

[0060] S1: Initialize the configuration information of the ground terminal and the UAV, where the configuration information of the ground terminal and the UAV includes the mission duration T n , the number of ground terminals Ground terminal location Ground terminal mission volume D n , the amount of tasks assigned to the edge by the ground terminal D off , UAV computing resources, Number of drones Drone location The number of gaps in the UAV flight trajectory m = 1, ..., M, and the duration of the gaps Horizontal flight speed of drone Maximum horizontal speed of drone UAV altitude H, constant blade power P0, hovering induced power P1, rotor blade speed U tip , average rotor induced speed v0 during hovering, fuselage drag ratio d0, rotor solidity S, air density ρ, rotor disk area G, fixed transmission power P of the drone, bandwidth B, noise variable N0, environmental constants a, b, speed of light c, and the loss of LoS link and non-LoS link according to different environments is η LoS and η NLoS , carrier frequency f c , constants A and C are A=η LoS -η NLoS and The distance between the drone and the ground base station at the mth time slot The blocking probability of the link between the UAV and the kth ground terminal in the nth time slot Path loss of data transmission link between base station and drone Channel Rate

[0061] S2: Set an iteration number variable i, whose initial value i = 1; set the initial trajectory of the drone to an elliptical trajectory connected end to end, and the drone maintains a constant speed during the flight. The starting position of the initial drone trajectory is expressed as The position of the u-th UAV in the m-th segment is expressed as

[0062] S3: According to the initialization conditions, the initial objective function value is obtained, including the following steps:

[0063] S3.1: Obtain the propulsion energy consumption of the u-th UAV in the m-th flight segment:

[0064] S3.2: Obtain the energy consumption generated by the u-th UAV serving the n-th ground terminal: E un ;

[0065] S3.3: The total energy consumption of the u-th UAV is obtained through S3.1-S3.2: Where β is the weight value, which is substituted into the trajectory of the u-th drone initialized by S2 Get the energy consumption of the initialized drone

[0066] S4: Optimize the connection between the drone and the ground terminal. The specific method is to first solve the problem Disassemble and follow the given drone trajectory Use convex optimization tools to calculate the optimal link between the UAV and the ground terminal at the i-th iteration α un [i], the sub-problems after decomposition In the following form:

[0067]

[0068] st(1.1)(1.3)(1.4

[0069] Target problem after decomposition is a convex function, which can be solved by CVX using convex optimization techniques, and finally the suboptimal solution α at the i-th iteration is obtained. un [i].

[0070] S5: The optimization of the trajectory can be rewritten as follows:

[0071]

[0072] st(1.3)(1.4)(1.5)(1.6)

[0073] Among them, the constraint condition (1.6) can be added with a relaxation parameter θ to make the speed range larger, which is conducive to finding the best drone position. The constraint after substituting the relaxation parameter is as follows:

[0074]

[0075] However, since constraints (1.3) and (1.4) are still non-convex, we cannot directly solve the problem The convex optimization tool is used to find a suboptimal solution, so continue to To convert the form: in is the intermediate parameter for formula simplification, constant Constant A<0, h m is the altitude of the drone in the current time slot, d nm is the distance between the UAV and the nth ground terminal in the current time slot, so the above expression can be changed into the form: Let's simplify the expression further, It can be simplified to:

[0076]

[0077] In order to transform formula (7) into a convex form, a local point is selected Substitute it into formula (7) and perform Taylor expansion to convert it into the following form:

[0078]

[0079] in, Substituting back into the constraints (1.3) and (1.4) we can get the new constraints:

[0080]

[0081]

[0082] Therefore, the optimization problem of the UAV trajectory can be written as follows:

[0083]

[0084] st(1.5)(7)(10)(11)

[0085] The above issues It has a convex property and can be directly solved by the CVX tool to obtain the optimal UAV trajectory

[0086] S6: The specific method for optimizing the allocation of UAV computing resources is to: un [i], using formula (1.2) to calculate the optimal suboptimal solution for the optimal UAV computing resource allocation at the i-th iteration using CVX

[0087] S7: Compare the objective function E obtained at the i-th iteration with the target function value E val Compare, if less than E val , the iteration ends and executes S8; otherwise, return to step S4 to execute the i+1th iteration.

[0088] S8: Obtain the best link between the drone and the ground terminal α un , Optimal UAV computing resource allocation Optimal drone trajectory And the minimum drone energy consumption value.

[0089] The above is only a preferred embodiment of the present invention. It should be understood that the present invention is not limited to the form disclosed herein, and should not be regarded as excluding other embodiments, but can be used for various other combinations, modifications and improvements, and can be modified within the scope of the concept described herein through the above teachings or the technology or knowledge of the relevant field. The changes and modifications made by those skilled in the art do not deviate from the spirit and scope of the present invention, and should be within the scope of protection of the claims attached to the present invention.

Claims

1. A joint trajectory and deep learning task optimization method for multiple UAVs, characterized by: The method comprises: Initialize the configuration information of the ground terminal and the drone, set the iteration number variable, and initialize the drone trajectory; According to the initialization conditions, the initial objective function value is obtained, and the energy consumption problem of the UAV is described as And set constraints, and optimize the connection between the UAV and the ground terminal, optimize the UAV trajectory, and optimize the allocation of UAV computing resources; Determine whether the objective function value obtained at the i-th iteration reaches the preset target value E val , if reached, then the optimal link α between the UAV and the ground terminal is obtained un Optimal UAV computing resource allocation Optimal drone trajectory And the minimum drone energy consumption value, otherwise continue to iterate.

2. The method for joint trajectory and deep learning task optimization for multiple UAVs according to claim 1, characterized in that: The setting of the iteration number variable and the initialization of the drone trajectory include: Set the iteration number variable i with an initial value of 1; The initial trajectory of the drone is set to an elliptical trajectory with the head and tail connected, and the drone maintains a constant speed during the flight. The starting position of the initial drone trajectory is The position of the u-th UAV in the m-th segment is 3. The method for joint trajectory and deep learning task optimization for multiple UAVs according to claim 2, characterized in that: The obtaining of the initial objective function value comprises: Get the propulsion energy consumption of the u-th UAV in the m-th segment of flight Get the energy consumption E generated by the u-th UAV serving the n-th ground terminal un ; Get the total energy consumption of the u-th drone Combined with the initialization trajectory of the u-th UAV Get the energy consumption of the initialized drone Among them, β is the weight value.

4. The method for joint trajectory and deep learning task optimization for multiple UAVs according to claim 3, characterized in that: The energy consumption problem of drones is described as And set constraints including: The energy consumption problem of drones is described as Set the link limit of the ground base station to That is, a base station can only select one drone for connection; Set the maximum value of the drone computing resources to Set the service demand constraint for each task as and That is, the delay time must be less than the task deadline, S n is the amount of data for the intermediate results, D off The amount of tasks assigned to the edge for ground terminals, is the channel rate, t s is the gap duration; Set the starting point of the drone trajectory to Set the flight speed of the drone to 5. The method for joint trajectory and deep learning task optimization for multiple UAVs according to claim 4, characterized in that: The optimization of the link between the drone and the ground terminal includes: The problem Disassemble and follow the given drone trajectory Use convex optimization tools to calculate the optimal link between the UAV and the ground terminal at the i-th iteration α un [i], get the decomposed sub-problems for And solve it using convex optimization technology through CVX, and finally get the suboptimal solution α at the i-th iteration un [i], and the following constraints are met: The link limit of the ground base station is That is, a base station can only select one drone to connect to; the service demand constraint for each task is and That is, the delay time must be less than the task deadline, where M n The number of time slots required for the ground terminal to complete the local deep learning task, M' n The number of time slots required for the nth ground terminal to upload the intermediate results to the UAV.

6. The method for joint trajectory and deep learning task optimization for multiple UAVs according to claim 4, characterized in that: The optimization of the UAV trajectory includes: The problem Disassemble and obtain the disassembled sub-problems for And the following constraints are met: The service demand constraint for each task is and That is, the delay time must be less than the task deadline; the starting point of the drone trajectory is The flight speed of the drone is θ represents the relaxation parameter.

7. The method for joint trajectory and deep learning task optimization for multiple UAVs according to claim 6, characterized in that: For the channel rate in the constraint condition Convert to get To achieve suboptimal solution using convex optimization tools, is the intermediate parameter for formula simplification, constant Constant A<0, h m is the altitude of the drone in the current time slot, d nm is the distance between the UAV and the nth ground terminal in the current time slot, N0 is the noise variable, P is the fixed transmission power of the UAV, B is the bandwidth, and c is the speed of light; The channel rate is further converted into And order Simplified Select a local point Substitution Convert it into a convex form and perform Taylor expansion to obtain Will and Substitution and Get new constraints and 8. The method for joint trajectory and deep learning task optimization for multiple UAVs according to claim 7, characterized in that: The optimization of the UAV trajectory includes: The problem Disassemble and obtain the disassembled sub-problems for The optimal UAV trajectory is obtained directly through the CVX tool And satisfy the following constraints: The starting point of the drone trajectory is The flight speed of the drone is θ represents the relaxation parameter; and 9. The method for joint trajectory and deep learning task optimization for multiple UAVs according to claim 4, characterized in that: The optimization of the allocation of computing resources of the UAV includes: According to the given link between the drone and the ground terminal α un [i], through the formula Using CVX to calculate the optimal suboptimal solution for the allocation of UAV computing resources at the i-th iteration

Images