A joint offline and online drone-assisted mobile edge computing communication method

Abstract

The present invention provides a combined offline and online drone-assisted mobile edge computing communication method. Based on a probabilistic line-of-sight channel model, the method takes into account the start time of user computing tasks, constructs an optimization problem with the goal of minimizing the sum of all user computing delays, and obtains the optimal user scheduling, user resource allocation and drone flight trajectory for each section through offline iterative optimization; constructs a linear programming problem for each section of the optimal drone flight trajectory through an online process, and cyclically optimizes the constructed linear programming problem, adaptively adjusts the drone's flight speed and user computing task resource allocation in the remaining sections, and provides computing services for users of the corresponding sections according to the adjustment results; the method optimizes drone-assisted mobile edge computing communication by combining offline and online methods, which can significantly reduce the sum of the delays of all user computing tasks and improve communication efficiency.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle (UAV) mobile edge computing technology, and more specifically, to a combined offline and online UAV-assisted mobile edge computing communication method. Background Art

[0002] The increasing popularity of the Internet of Things (IoT) and the growing number of complex mobile applications, such as virtual and augmented reality, online gaming, and autonomous driving, have led to a dramatic increase in the computing demands of user devices. However, limited computing resources make it difficult for user devices to provide satisfactory quality of service. Therefore, effective technologies are needed to overcome the increasing computational workload. Mobile edge computing technology is a technology that can help user devices with limited computing resources handle computationally intensive tasks. By offloading computing tasks to nearby base stations, it can significantly reduce network burden and task execution latency.

[0003] Drones have the advantages of on-demand deployment, controllable maneuverability, security monitoring and surveillance, and low cost. They have received widespread attention in both civil and military applications. At the same time, drones are also expected to play an important role in future wireless communication systems.

[0004] Drone-assisted mobile edge computing is a new communication technology. Drones, equipped with powerful microservers, provide services to users on the ground. Each user can offload some of their computing tasks to the drone for processing, while the remaining tasks are processed locally. Drones can fly close to users on the ground to provide computing services. Communication between drones and users utilizes time-division multiple access (TDMA), meaning that a drone only communicates with one user at a time. This technology then minimizes the combined computational latency of all user tasks by optimizing user offloaded tasks, user scheduling, and drone flight paths.

[0005] The current prior art discloses a scheduling optimization method and system for drone-assisted mobile edge computing, including constructing an offloading model of a mobile edge computing system with a drone and several user devices, and calculating the energy consumption of completing each computing task; with the goal of minimizing the average energy consumption of user devices, establishing an optimization problem of joint drone trajectory and user device scheduling, converting it into a Markov decision process, defining the state space, action space and reward function of the mobile edge computing system offloading model, and using it to train a deep neural network constructed based on the SAC algorithm. The trained deep neural network can be used to perform scheduling optimization and obtain the optimal scheduling strategy, and the continuous action of the drone can be planned to obtain a reasonable and accurate flight trajectory and user device selection strategy; although the prior art can reduce the computing energy consumption and computing latency of user devices to a certain extent through drone-assisted mobile edge computing technology, it cannot The start time of the user's computing task, that is, the time when the user generates a computing task requirement, is not taken into account. Before this time, the user has no computing task requirement, and accordingly, there is no need for the drone to provide computing services for it. In addition, the model between the drone and the user in the existing technology is a line-of-sight channel model, which is too simple and ignores factors such as the existence of obstacles and small-scale fading. Therefore, the designed flight route may not meet the conditions for line-of-sight communication in some sections due to the presence of obstacles, and is not suitable for actual use. In addition, the existing technology uses an offline design method, which considers a deterministic expected channel gain. However, in the process of designing communication between the drone and the user, the communication channel state is random and uncertain. Therefore, the drone cannot perceive the instantaneous channel state for online adaptive adjustment, thereby increasing the total delay of the user computing task. Therefore, the current existing technology has the defect of a large total delay of the user computing task. Summary of the Invention

[0006] In order to overcome the defect of large total delay of the above-mentioned user computing tasks, the present invention provides a joint offline and online drone-assisted mobile edge computing communication method, which can significantly reduce the sum of the delays of all user computing tasks and improve communication efficiency.

[0007] In order to solve the above technical problems, the technical solutions of the present invention are as follows:

[0008] A combined offline and online UAV-assisted mobile edge computing communication method includes the following steps:

[0009] S1: Construct a probabilistic line-of-sight channel model;

[0010] S2: Offline process: Based on the probabilistic line-of-sight channel model, the start time of the user computing task is considered and an optimization problem is constructed. The goal is to minimize the sum of all user computing delays. The user scheduling, user computing task resource allocation and UAV flight trajectory of each road segment are obtained.

[0011] The obtained UAV flight trajectory is composed of a series of path points and line segments connecting the path points. Adjacent path points and the connecting line segments between them are defined as a segment.

[0012] S3: Iteratively optimize the optimization problem constructed in the offline process to obtain the optimal user scheduling for each road section, the optimal user computing task resource allocation for each road section, and the optimal UAV flight trajectory for each road section;

[0013] S4: Online process: Make the UAV fly along the optimal UAV flight trajectory for each road section, fix the optimal user scheduling for each road section, do not change the flight direction of the UAV at each path point, construct a linear programming problem, cyclically optimize the linear programming problem, adaptively adjust the flight speed of the UAV on the remaining road sections and the allocation of user computing task resources, and provide computing services to users on the corresponding road sections based on the adjustment results.

[0014] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0015] The present invention provides a combined offline and online UAV-assisted mobile edge computing communication method. Based on a probabilistic line-of-sight channel model, the method considers the start time of user computing tasks and, with the goal of minimizing the sum of all user computing delays, constructs an optimization problem. Through offline iterative optimization, the optimal user scheduling, optimal user resource allocation, and optimal UAV flight trajectory for each section are obtained. An online linear programming problem is constructed for each section of the optimal UAV flight trajectory, and the constructed linear programming problem is cyclically optimized. The flight speed of the UAV and the user computing task resource allocation in the remaining sections are adaptively adjusted, and computing services are provided to users in the corresponding sections based on the adjustment results. The method considers the start time of user computing tasks and uses a probabilistic line-of-sight channel model for communication between the UAV and ground users to obtain user scheduling, user resource allocation, and UAV flight trajectory. A combined offline and online optimization approach is used to achieve better optimization results, significantly reducing the sum of all user computing task delays and improving communication efficiency. Furthermore, the consideration of the randomness of the start time of user computing tasks and the use of a probabilistic line-of-sight channel model are closer to actual scenarios and communication environments, resulting in higher communication accuracy and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A flowchart of a combined offline and online drone-assisted mobile edge computing communication method provided in Example 1.

[0017] Figure 2 Schematic diagram of drone-assisted mobile edge computing communication provided in Example 2.

[0018] Figure 3 This is the online process cycle optimization flow chart provided for Example 2.

[0019] Figure 4 The two-dimensional trajectory diagram of different drone flight durations under the probabilistic line-of-sight channel provided in Example 2.

[0020] Figure 5 The three-dimensional trajectory diagram of different drone flight durations under the probabilistic line-of-sight channel provided in Example 2.

[0021] Figure 6 The two-dimensional trajectory diagram of the drone under different schemes when the drone flight duration provided in Example 2 is 25s.

[0022] Figure 7 The three-dimensional trajectory diagram of the drone under different schemes when the drone flight duration provided in Example 2 is 25 seconds.

[0023] Figure 8 The two-dimensional trajectory diagram of the drone provided in Example 2 is a 15s flight duration diagram under different user task start times.

[0024] Figure 9 The three-dimensional trajectory diagram of the drone provided in Example 2 is when the flight duration of the drone is 15 seconds and the user task start time is different.

[0025] Figure 10 Schematic diagram of online process performance optimization when the drone flight duration provided in Example 2 is 25 seconds.

[0026] Figure 11 This is a relationship diagram between the drone's continuous flight time and the sum of the computing delays of all users under different solutions provided in Example 2. DETAILED DESCRIPTION

[0027] The accompanying drawings are for illustrative purposes only and are not to be construed as limiting this patent;

[0028] In order to better illustrate this embodiment, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product size;

[0029] It is understandable to those skilled in the art that some well-known structures and descriptions thereof may be omitted in the drawings.

[0030] The technical solution of the present invention is further described below with reference to the accompanying drawings and embodiments.

[0031] Example 1

[0032] like Figure 1 As shown, this embodiment provides a joint offline and online drone-assisted mobile edge computing communication method, including the following steps:

[0033] S1: Construct a probabilistic line-of-sight channel model;

[0034] S2: Offline process: Based on the probabilistic line-of-sight channel model, the start time of the user computing task is considered and an optimization problem is constructed. The goal is to minimize the sum of all user computing delays. The user scheduling, user computing task resource allocation and UAV flight trajectory of each road segment are obtained.

[0035] The obtained UAV flight trajectory is composed of a series of path points and line segments connecting the path points. Adjacent path points and the connecting line segments between them are defined as a segment.

[0036] S3: Iteratively optimize the optimization problem constructed in the offline process to obtain the optimal user scheduling for each road section, the optimal user computing task resource allocation for each road section, and the optimal UAV flight trajectory for each road section;

[0037] S4: Online process: Make the UAV fly along the optimal UAV flight trajectory for each road section, fix the optimal user scheduling for each road section, do not change the flight direction of the UAV at each path point, construct a linear programming problem, cyclically optimize the linear programming problem, adaptively adjust the flight speed of the UAV on the remaining road sections and the allocation of user computing task resources, and provide computing services to users on the corresponding road sections based on the adjustment results.

[0038] In the specific implementation process, a probabilistic line-of-sight channel model is first constructed. Then, based on the probabilistic line-of-sight channel model, the start time of the user computing task is considered and an optimization problem is constructed. The goal is to minimize the sum of all user computing delays, and obtain the user scheduling, user computing task resource allocation and drone flight trajectory of each section. The obtained drone flight trajectory is composed of a series of path points and line segments connecting the path points. Adjacent path points and the connecting line segments between them are defined as a section. Then, the optimization problem constructed in the offline process is iteratively optimized to obtain the optimal user scheduling for each section, the optimal user computing task resource allocation for each section and the optimal drone flight trajectory for each section. Then, the drone is made to fly along the obtained optimal drone flight trajectory for each section, and the obtained optimal user scheduling for each section is fixed. At each path point, the flight direction of the drone is not changed, and a linear programming problem is constructed. The linear programming problem is cyclically optimized, and the flight speed and user computing task resource allocation of the drone in the remaining sections are adaptively adjusted. Computing services are provided to users of the corresponding section based on the adjustment results.

[0039] This method takes into account the start time of user computing tasks. The communication between UAVs and ground users adopts a probabilistic line-of-sight channel model to obtain user scheduling, user resource allocation and UAV flight trajectory, and adopts a combination of offline and online optimization methods to obtain better optimization results. It can significantly reduce the sum of the delays of all user computing tasks and improve communication efficiency.

[0040] Example 2

[0041] This embodiment provides a combined offline and online drone-assisted mobile edge computing communication method, including the following steps:

[0042] S1: Construct a probabilistic line-of-sight channel model;

[0043] S2: Offline process: Based on the probabilistic line-of-sight channel model, the start time of the user computing task is considered and an optimization problem is constructed. The goal is to minimize the sum of all user computing delays. The user scheduling, user computing task resource allocation and UAV flight trajectory of each road segment are obtained.

[0044] The obtained UAV flight trajectory is composed of a series of path points and line segments connecting the path points. Adjacent path points and the connecting line segments between them are defined as a segment.

[0045] S3: Iteratively optimize the optimization problem constructed in the offline process to obtain the optimal user scheduling for each road section, the optimal user computing task resource allocation for each road section, and the optimal UAV flight trajectory for each road section;

[0046] S4: Online process: Make the UAV fly along the optimal UAV flight trajectory for each road section, fix the optimal user scheduling for each road section, do not change the flight direction of the UAV at each path point, construct a linear programming problem, cyclically optimize the linear programming problem, adaptively adjust the flight speed of the UAV on the remaining road sections and the allocation of user computing task resources, and provide computing services to users on the corresponding road sections based on the adjustment results.

[0047] In the specific implementation process, a probabilistic line-of-sight channel model is first constructed, which includes an initial sub-model, a UAV trajectory sub-model, a communication sub-model, and a calculation sub-model;

[0048] The initial sub-model is specifically:

[0049] like Figure 2 As shown, Figure 2 Schematic diagram of drone-assisted mobile edge computing communication. A drone provides computing services to K users on the ground. The user set is represented as κ = {1,…,K}. The user's position relative to the ground remains unchanged. The coordinates of user k are represented as in represents the horizontal coordinate of user k; the total task computation amount of user k is L k bit, the corresponding user computing task start time is Where k∈κ; the drone is dispatched to provide computing services to all users within the total flight time T0. Each user offloads part of the computing tasks to the drone for computing, and the remaining computing tasks are calculated locally;

[0050] The UAV trajectory sub-model is specifically:

[0051] The total flight time T0 of the UAV is divided into N time slots of equal length. The UAV takes off from the initial position and lands at the end point. The flight trajectory of the UAV is converted into a (N+1)-length coordinate sequence, that is, in, is the initial coordinate of the UAV, is the endpoint coordinate of the UAV; the UAV independently controls the horizontal and vertical flight speeds under the premise of meeting the maximum flight speed constraint, where V xy,max and V z,max are the maximum horizontal speed and maximum vertical speed of the UAV, respectively, and the following trajectory constraints are obtained:

[0052]

[0053] in, S xy,max and S z,max are the maximum horizontal flight distance and maximum vertical flight distance of the UAV in each time slot respectively;

[0054] At the same time, set the constraints on the drone's flight altitude:

[0055]

[0056] Among them, H min and H max are the minimum and maximum flight altitudes of the UAV in each time slot respectively;

[0057] The communication sub-model is specifically:

[0058] The channel model between the UAV and the user is divided into two states: line-of-sight and non-line-of-sight. represents the channel state between the binary drone and user k at the nth time slot, where c k,n =1 and c k,n = 0 respectively indicates that the channel state between the UAV and user k in the nth time slot is line-of-sight and non-line-of-sight; the line-of-sight probability of the nth time slot

[0059]

[0060] Wherein, B1, B2, B3 and B4 are the first environmental constant, the second environmental constant, the third environmental constant and the fourth environmental constant respectively, satisfying B1<0, B2>0, B4>0 and B3+B4=1; θ k,n represents the elevation angle between the UAV and user k at time slot n, and is expressed as:

[0061]

[0062] Non-line-of-sight probability in the nth time slot

[0063] In the nth time slot, the real-time channel power gain between the UAV and the user is expressed as:

[0064]

[0065]

[0066]

[0067] in, and Denote the channel power gain in line-of-sight and non-line-of-sight states, d k,n is the distance between the UAV and user k in the nth time slot, β0 is the average channel power gain at the reference distance d0 = 1m in the line-of-sight state, μ represents the additional signal attenuation factor in the non-line-of-sight state, and α L and α N They represent the average path loss index under line-of-sight and non-line-of-sight conditions respectively; in this embodiment, μ<1, 2≤α L ≤α N ≤6;

[0068] User k transmits with the UAV at its maximum power P, using a k,n represents the binary communication scheduling variable between the UAV and user k at the nth time slot, a k,n =1 means the drone communicates with user k, a k,n =0 means the UAV does not communicate with user k;

[0069] In each time slot, the drone is only allowed to communicate with one user, and the following scheduling constraints are set:

[0070]

[0071] Define the transmission rate of user k in the nth time slot as r k,n , the unit is bps / Hz, the formula is as follows:

[0072]

[0073] Among them, h k,n represents the real-time channel power gain, σ 2 represents the additional Gaussian white noise power of the drone; the real-time transmission rate is:

[0074]

[0075] in, and Represent the transmission rate in line-of-sight and non-line-of-sight conditions, respectively, and

[0076] Define the first communication rate of user k in each time slot as E[r k,n ], specifically:

[0077]

[0078] The first communication rate E[r k,n ]Reconstructed to the second communication rate The second communication rate Specifically:

[0079]

[0080] The calculation sub-model is specifically:

[0081] The time spent by user k to perform the computing task locally in each time slot is:

[0082]

[0083] in, is the local computing time, L k,n is the computational workload of user k in the nth time slot, in bits, C k The number of CPU cycles required for user k to execute 1 bit of task, f k The local computing power of user k;

[0084] The time it takes for the drone to perform mobile edge computing tasks is:

[0085]

[0086]

[0087]

[0088] in, The time taken for mobile edge computing execution, The time it takes for user k to offload part of the computing tasks to the drone, The time it takes for the drone to calculate the offloaded computational tasks, f m is the computing power of the UAV, and B is the transmission bandwidth;

[0089] In this embodiment, the user's computing latency is the maximum of the time spent on local computing and the time spent on edge computing. Based on the probabilistic line-of-sight channel model, the start time of the user's computing task is considered to construct a first optimization problem R. The first optimization problem R and its constraints are specifically as follows:

[0090]

[0091]

[0092]

[0093]

[0094]

[0095]

[0096]

[0097]

[0098]

[0099]

[0100]

[0101]

[0102] Among them, δ = T0 / N, which represents the length of a time slot, Indicates the number of time slots included in the start time of the user computing task, A set of horizontal coordinates representing the flight trajectory of the drone, A set of vertical coordinates representing the flight trajectory of the drone, represents the set of binary communication scheduling variables between the UAV and user k in each time slot, represents the set of computing tasks of user k in each time slot;

[0103] constraint and They respectively indicate that the UAV will not schedule communication with the user and the user will not assign computing tasks before the user's computing task starts. Indicates that the sum of the time spent by the user offloading the computing task to the drone in each time slot and the time spent by the drone performing the offloaded task is less than the length of the time slot;

[0104] The first optimization problem R is difficult to solve directly. A convex optimization problem is established based on the first optimization problem R. The convex optimization problem and its constraints are specifically as follows:

[0105]

[0106]

[0107]

[0108]

[0109]

[0110]

[0111]

[0112]

[0113]

[0114]

[0115]

[0116] The constructed convex optimization problem is difficult to solve. First, the convex optimization problem is transformed into a first intermediate problem. The first intermediate problem and its constraints are specifically:

[0117]

[0118]

[0119]

[0120]

[0121]

[0122]

[0123]

[0124]

[0125]

[0126]

[0127]

[0128]

[0129]

[0130] Where η represents the maximum value of the local computing time and the time spent on mobile edge computing execution;

[0131] Then relax the user scheduling constraints The binary integer variables in the result are the second intermediate problem, which is specifically:

[0132]

[0133]

[0134]

[0135]

[0136]

[0137]

[0138]

[0139]

[0140]

[0141]

[0142]

[0143]

[0144]

[0145] Then relax the second intermediate problem and solve the constraint The non-affine constraints in , we get the second optimization problem, which is specifically:

[0146]

[0147]

[0148]

[0149]

[0150]

[0151]

[0152]

[0153]

[0154]

[0155]

[0156]

[0157]

[0158]

[0159] The second optimization problem is difficult to solve directly. Given the user computing task resource allocation and the drone 3D trajectory set {L, Q, Z}, the second optimization problem is reconstructed into a third optimization problem. The third optimization problem and its constraints are specifically:

[0160]

[0161]

[0162]

[0163]

[0164]

[0165]

[0166]

[0167] The third optimization problem is a standard linear programming problem, and the user scheduling of each road segment is obtained by solving the third optimization problem using a solver;

[0168] Given the user schedule and the UAV's three-dimensional trajectory {A, Q, Z}, the second optimization problem is reconstructed into a fourth optimization problem. The fourth optimization problem and its constraints are specifically:

[0169]

[0170]

[0171]

[0172]

[0173]

[0174]

[0175]

[0176] The fourth optimization problem is a standard linear programming problem, and the fourth optimization problem is solved by a solver to obtain the user computing task resource allocation of each road segment;

[0177] The UAV flight trajectory includes a UAV horizontal flight trajectory and a UAV vertical flight trajectory;

[0178] Given user scheduling, user computing task resource allocation, and the UAV vertical trajectory {A, L, Z}, the second optimization problem is restructured into a third intermediate problem. The third intermediate problem and its constraints are specifically:

[0179]

[0180]

[0181]

[0182]

[0183]

[0184]

[0185]

[0186] Before solving the third intermediate problem, we first introduce Theorem 1:

[0187] Theorem 1: Given γ>0 and α>2, is a convex function with respect to x>0 and y>0;

[0188] Using Theorem 1, for the second communication rate Perform a first-order Taylor expansion to obtain the second communication rate The lower bound of :

[0189]

[0190] in, and is a given initial feasible value,

[0191] Then introduce the slack variable ξ to constrain and Transformed into the following form:

[0192]

[0193]

[0194]

[0195] Then constrain Transformed into the following form:

[0196]

[0197] in, and is the given initial feasible value;

[0198] The third intermediate problem is reconstructed into a fifth optimization problem by introducing a slack variable ξ and a partial Taylor expansion. The fifth optimization problem and its constraints are specifically:

[0199]

[0200]

[0201]

[0202]

[0203]

[0204]

[0205]

[0206]

[0207]

[0208] The fifth optimization problem is solved by the solver to obtain the horizontal flight trajectory of the UAV in each section;

[0209] Given user scheduling, user computing task resource allocation, and drone horizontal trajectory {A, L, Q}, the second optimization problem is restructured into a fourth intermediate problem. The fourth intermediate problem and its constraints are specifically:

[0210]

[0211]

[0212]

[0213]

[0214]

[0215]

[0216]

[0217]

[0218] The fourth intermediate problem is reconstructed into a sixth optimization problem by introducing a slack variable ξ and a partial Taylor expansion. The sixth optimization problem and its constraints are specifically:

[0219]

[0220]

[0221]

[0222]

[0223]

[0224] z1=z I , z N+1 =z F

[0225]

[0226]

[0227]

[0228]

[0229]

[0230] The sixth optimization problem is solved using the solver to obtain the vertical flight trajectory of the UAV in each section;

[0231] The obtained UAV flight trajectory is composed of a series of path points and line segments connecting the path points. Adjacent path points and the connecting line segments between them are defined as a segment.

[0232] This embodiment iteratively optimizes the optimization problem constructed in the offline process to obtain the optimal user scheduling for each road section, the optimal user computing task resource allocation for each road section, and the optimal drone flight trajectory for each road section. The specific method includes the following steps:

[0233] S3.1: Set the initial value of the number of iterations i to 0 and the threshold ε0 to 1×10 -4 , initialize the first optimization problem R (i) and its parameter A (i) , L (i) , Q (i) and Z (i) ;

[0234] S3.2: Using the L (i) , Q (i) and Z (i) , obtain A by solving the third optimization problem (i+1) ;

[0235] S3.3: Using the A (i+1) , Q (i) and Z (i) , obtain L by solving the fourth optimization problem (i+1) ;

[0236] S3.4: Using the A (i+1) , L (i+1) , Q (i) and Z (i) , obtain Q by solving the fifth optimization problem (i+1) ;

[0237] S3.5: Using the A (i+1) , L (i+1) , Q (i+1) and Z (i) , obtain Z by solving the sixth optimization problem (i+1) ;

[0238] S3.6: Using the A (i+1) , L (i+1) , Q (i+1) and Z (i+1) Get the first optimization problem R after the next iteration (i +1) ;

[0239] S3.7: When When the sum of all user computing delays is minimized, the optimal user scheduling, optimal user computing task resource allocation and optimal UAV flight trajectory for each section are obtained; otherwise, the first optimization problem R after the next iteration is (i+1) and its parameter A (i+1) , L(i+1) , Q (i+1) and Z (i+1) Repeat steps S3.2-S3.6 as the initial value;

[0240] Then, the UAV is made to fly along the optimal UAV flight trajectory of each road section, the optimal user scheduling of each road section is fixed, and the task start time T is calculated according to each user. k start Divide the optimal UAV flight trajectory into several path points. At each path point, without changing the UAV's flight direction, construct a linear programming problem, iteratively optimize the linear programming problem, adaptively adjust the UAV's flight speed and user computing task resource allocation on the remaining sections, and provide computing services to users on the corresponding sections based on the adjustment results.

[0241] The linear programming problem and its constraints are specifically:

[0242]

[0243]

[0244]

[0245]

[0246]

[0247]

[0248]

[0249]

[0250]

[0251]

[0252]

[0253] in, represents the resource allocation set of user computing tasks for each road segment from n to N, It represents the set of drone flight durations in each section from n to N. Represents the sum of the remaining flight time of the drone starting from the nth segment, satisfying n>1, T k,n Indicates the start time of the calculation task of user k on the nth road segment, satisfying T k,n =T k,n-1 -tn-1 ,n>1, Indicates the remaining computational tasks for user k starting from the nth road segment, satisfying represents the sum of the time spent by user k on local computation before the nth road segment. It represents the sum of the time spent on the edge computing execution of the drone before the nth road segment;

[0254] like Figure 3 As shown, the specific method of loop optimization of the linear programming problem includes the following steps:

[0255] S4.1: Construct N corresponding linear programming problems based on the N segments of the optimal UAV flight trajectory;

[0256] S4.2: Optimize the constructed N linear programming problems in a loop, adaptively adjust the flight speed of the UAV on the remaining road sections and the allocation of user computing task resources;

[0257] S4.3: Providing calculation services to users of the corresponding road section based on the adjustment results;

[0258] like Figure 4 and Figure 5 As shown, combined Figure 4 and Figure 5 The two-dimensional and three-dimensional trajectory diagrams of the drone show that as the drone's flight duration increases, the drone can fly as close to the user as possible to communicate with it. When the drone's flight duration is 20 seconds, the drone will fly above four users to communicate with them. The drone's close communication with the user can increase the communication rate, thereby reducing the time it takes for the user to offload computing tasks to the drone. When the drone cannot fly close to communicate with the user, the drone will increase its flight altitude to increase the probability of line-of-sight communication with the user and increase the communication rate, thereby reducing the time it takes for the user to offload computing tasks to the drone.

[0259] like Figure 6 and Figure 7 As shown, Figure 6 and 7 The trajectory comparison diagram of different schemes when the UAV flight duration is 25s, among which the LOS scheme represents the optimization scheme based on the line-of-sight channel model, PLLA represents the optimization scheme based on the probabilistic line-of-sight channel model and the fixed UAV flight altitude is a fixed value, and PLOS represents the offline optimization scheme based on the probabilistic line-of-sight channel model. Figure 6 and Figure 7The two-dimensional and three-dimensional trajectory diagrams of the drone show that when the drone flies for a long time, the two-dimensional trajectories of the three different schemes are almost the same, because the drone will choose to fly above the user to achieve a higher communication rate; however, there is a certain difference in the three-dimensional trajectory. The PLOS scheme has a significant difference in the drone's flight altitude compared to the LOS and PLLA schemes. When the drone is far away from the user, the PLOS scheme will increase the drone's flight altitude to increase the elevation angle and improve the probability of line-of-sight communication, thereby increasing the communication rate and reducing the time it takes for the user to offload computing tasks to the drone.

[0260] like Figure 8 and Figure 9 As shown, Figure 8 and 9 This is a comparison chart of the drone's trajectories for different user task start times when the flight duration is 15 seconds. The figure shows the drone's trajectories for four different user task start times. It can be seen from the figure that the drone's flight trajectory changes according to the user task start time. In other words, the drone will prioritize flying to the user with the first service request, and then serve the next user with the next request, providing computing services to users in sequence.

[0261] like Figure 10 As shown, Figure 10 This is a performance graph for online optimization when the drone's flight duration is 25 seconds. The graph shows the drone's flight duration for each road section, the communication rate between the drone and the user, and the amount of tasks offloaded by the user. Analysis of the data in the graph shows that when the communication rate on a certain road section is high, the drone's flight duration is longer and the amount of tasks offloaded by the user is also correspondingly higher. Because the communication rate on this road section is high, the drone chooses to fly slowly, allowing the user to offload as many computing tasks as possible to the drone, which can then perform task calculations with the help of the drone's more powerful edge computing server, reducing the time spent on task calculations.

[0262] like Figure 11 As shown, Figure 11 It is the relationship between the sum of the computing delays of all users and the continuous flight time of the drone. The four curves in the figure respectively represent the optimization scheme based on the line-of-sight channel model (LOS), the optimization scheme based on the probabilistic line-of-sight channel model and the fixed drone flight altitude (PLLA), the offline optimization scheme based on the probabilistic line-of-sight channel model (PLOS), and the offline and online combined optimization scheme provided in this embodiment. From the figure, we can see that using the method provided in this embodiment, the sum of the computing delays of all users is the smallest, followed by the PLOS scheme and the PLLA scheme. The sum of the computing delays of all users under the LOS scheme is the largest. Analyzing the data in the figure, it can be seen that the offline and online combined optimization scheme proposed in this embodiment can significantly reduce the sum of the computing delays of all users, which also shows the effectiveness of the algorithm provided in this embodiment.

[0263] This method first obtains user scheduling, user resource allocation and UAV trajectory under an offline algorithm, and then performs online optimization based on the UAV trajectory, which can significantly reduce the sum of the delays of all user computing tasks and improve communication efficiency.

[0264] The same or similar reference numerals correspond to the same or similar components;

[0265] The terms used in the drawings to describe positional relationships are for illustrative purposes only and should not be construed as limiting this patent;

[0266] Obviously, the above embodiments of the present invention are merely examples for the purpose of clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the claims of the present invention.

Claims

1. A joint offline and online drone-assisted mobile edge computing communication method, characterized in that: The following steps are involved: S1: Construct a probabilistic line-of-sight channel model; The probabilistic line-of-sight channel model includes an initial sub-model, specifically: A UAV provides computing services to K users on the ground. The user set is represented as , the user's relative position to the ground remains unchanged, and the coordinates of user k are expressed as ,in represents the horizontal coordinate of user k; the total task computation amount of user k is bit, the corresponding user computing task start time is ,in ; Drones were dispatched in total flight time Providing computing services to all users, each user offloads part of the computing tasks to the drone for computing, and the remaining computing tasks are calculated locally; The probabilistic line-of-sight channel model also includes a drone trajectory sub-model, specifically: The total flight time of the drone Divided into The drone takes off from the initial position and lands at the end point in equal time slots. The flight trajectory of the drone is converted into Long coordinate sequence, i.e. ,in, is the initial coordinate of the UAV, is the endpoint coordinate of the UAV; the UAV independently controls the horizontal and vertical flight speeds under the premise of meeting the maximum flight speed constraint, where and are the maximum horizontal speed and maximum vertical speed of the UAV, respectively, and the following trajectory constraints are obtained: in, , and are the maximum horizontal flight distance and maximum vertical flight distance of the UAV in each time slot respectively; At the same time, set the constraints on the drone's flight altitude: in, and are the minimum and maximum flight altitudes of the UAV in each time slot respectively; The probabilistic line-of-sight channel model also includes a communication sub-model, specifically: The channel model between the UAV and the user is divided into two states: line-of-sight and non-line-of-sight. represents the channel state between the binary UAV and user k at the nth time slot, where and They represent the channel states between the UAV and user k at the nth time slot as line-of-sight and non-line-of-sight respectively; the line-of-sight probability of the nth time slot is : in, 、 、 and They are the first environmental constant, the second environmental constant, the third environmental constant and the fourth environmental constant, satisfying 、 、 and ; Indicates time slot Drone and user The elevation angle between is expressed as: Non-line-of-sight probability in the nth time slot ; In the nth time slot, the real-time channel power gain between the UAV and the user is expressed as: in, and They represent the channel power gains in line-of-sight and non-line-of-sight states, is the distance between the UAV and user k in the nth time slot, It is the reference distance in the line of sight state The average channel power gain at m, Indicates the additional signal attenuation factor under non-line-of-sight conditions, and They represent the average path loss index under line-of-sight and non-line-of-sight conditions respectively; User k has its maximum power Transmit with drone, Indicates in The binary communication scheduling variable between the UAV and user k at time slots is: Indicates that the drone communicates with user k, Indicates that the drone does not communicate with user k; In each time slot, the drone is only allowed to communicate with one user, and the following scheduling constraints are set: ， ， Define the transmission rate of user k in the nth time slot as , the unit is , the formula is as follows: in, represents the real-time channel power gain, represents the additional Gaussian white noise power of the UAV; The real-time transmission rates are as follows: in, and , respectively represent the transmission rate in line-of-sight and non-line-of-sight states, and ; Define the first communication rate of user k in each time slot as , specifically: ， The first communication rate of user k in each time slot Reconfiguration to the second communication rate , the second communication rate Specifically: The probabilistic line-of-sight channel model also includes a calculation sub-model, specifically: The time spent by user k to perform the computing task locally in each time slot is: in, is the local computation time, is the computational workload of user k in the nth time slot, in units of , Execute 1 for user k The number of CPU cycles required for the task, The local computing power of user k; The time it takes for the drone to perform mobile edge computing tasks is: in, The time taken for mobile edge computing execution, The time it takes for user k to offload part of the computing tasks to the drone, The time it takes to calculate the offloaded computational tasks for the drone, For the computing power of drones, is the transmission bandwidth; S2: Offline process: Based on the probabilistic line-of-sight channel model, the start time of the user computing task is considered and an optimization problem is constructed. The goal is to minimize the sum of all user computing delays, and obtain the user scheduling of each road section, the user computing task resource allocation and the UAV flight trajectory; specifically: Based on the probabilistic line-of-sight channel model, a first optimization problem R is constructed by considering the start time of the user computing task. The first optimization problem R and its constraints are specifically as follows: R: s.t. , ， in, , represents the length of a time slot, Indicates the number of time slots included in the start time of the user computing task, , represents the horizontal coordinate set of the UAV flight trajectory, , represents the vertical coordinate set of the UAV flight trajectory, , represents the set of binary communication scheduling variables between the UAV and user k in each time slot, , represents the set of computing tasks of user k in each time slot; The first optimization problem R is reconstructed and solved using the solver to obtain the user scheduling, user computing task resource allocation and UAV flight trajectory of each road segment; The obtained UAV flight trajectory is composed of a series of path points and line segments connecting the path points. Adjacent path points and the connecting line segments between them are defined as a segment. S3: Iteratively optimize the optimization problem constructed in the offline process to obtain the optimal user scheduling for each road section, the optimal user computing task resource allocation for each road section, and the optimal UAV flight trajectory for each road section; S4: Online process: The UAV is flown along the obtained optimal UAV flight trajectory for each road section, the obtained optimal user scheduling for each road section is fixed, and at each path point, the UAV's flight direction is not changed. A linear programming problem is constructed, the constructed linear programming problem is solved using a solver, and the linear programming problem is cyclically optimized. The UAV's flight speed and user computing task resource allocation on the remaining road sections are adaptively adjusted, and computing services are provided to users on the corresponding road sections based on the adjustment results. The linear programming problem and its constraints are specifically: s.t. in, , represents the resource allocation set of user computing tasks for each road section from n to N, , represents the set of drone flight durations in each section from n to N, Represents the sum of the remaining flight time of the drone starting from the nth segment, satisfying , Indicates the start time of the calculation task of user k on the nth road segment, satisfying , Indicates the remaining computational tasks for user k starting from the nth road segment, satisfying , Indicates the The sum of the time spent on local computation execution by user k before the road segment, Indicates the The sum of the time spent on drone edge computing execution before each road section.

2. The method for combined offline and online UAV-assisted mobile edge computing communication according to claim 1, characterized in that: The first optimization problem R is reconstructed and solved using the solver to obtain the user scheduling of each road section, the user computing task resource allocation, and the UAV flight trajectory as follows: The first optimization problem R is reconstructed into a second optimization problem. The second optimization problem and its constraints are specifically: s.t. ， in, It represents the maximum value of the local computing time and the time spent on mobile edge computing execution; The second optimization problem is reconstructed into a third optimization problem, and the third optimization problem is solved by a solver to obtain the user scheduling of each road segment. The third optimization problem and its constraints are specifically: s.t. The second optimization problem is reconstructed into a fourth optimization problem. The fourth optimization problem is solved by a solver to obtain the user computing task resource allocation for each road segment. The fourth optimization problem and its constraints are specifically: s.t. The UAV flight trajectory includes the UAV horizontal flight trajectory and the UAV vertical flight trajectory; The second optimization problem is reconstructed into a fifth optimization problem. The fifth optimization problem is solved by a solver to obtain the horizontal flight trajectory of the UAV in each section. The fifth optimization problem and its constraints are specifically: s.t. , ， ， ， in, 、 and For the given initial value, is the slack variable; The second optimization problem is reconstructed into the sixth optimization problem. The sixth optimization problem is solved by the solver to obtain the vertical flight trajectory of the UAV in each section. The sixth optimization problem and its constraints are specifically: s.t. , ， ， ， 。 3. The method for combined offline and online UAV-assisted mobile edge computing communication according to claim 2, characterized in that: In step S3, the optimization problem constructed in the offline process is iteratively optimized to obtain the optimal user scheduling for each road section, the optimal user computing task resource allocation for each road section, and the optimal UAV flight trajectory for each road section. The specific method includes the following steps: S3.1: Set the initial value of the iteration number i to 0 and set the threshold , initialize the first optimization problem and its parameters 、 、 and ; S3.2: Utilize the 、 and , obtained by solving the third optimization problem ; S3.3: Utilize the 、 and , obtained by solving the fourth optimization problem ; S3.4: Utilize the 、 、 and , obtained by solving the fifth optimization problem ; S3.5: Utilize the 、 、 and , obtained by solving the sixth optimization problem ; S3.6: Utilize the 、 、 and Get the first optimization problem after the next iteration ; S3.7: When When the sum of all user computing delays is minimized, the optimal user scheduling, optimal user computing task resource allocation and optimal UAV flight trajectory for each section are obtained; otherwise, the first optimization problem after the next iteration is and its parameters 、 、 and Repeat steps S3.2-S3.6 as initial values.

4. The method for combined offline and online UAV-assisted mobile edge computing communication according to claim 1, characterized in that: In step S4, the linear programming problem is constructed and the linear programming problem is cyclically optimized. The specific method includes the following steps: S4.1: Construct N corresponding linear programming problems based on the N segments of the optimal UAV flight trajectory; S4.2: Optimize the constructed N linear programming problems in a loop, adaptively adjust the flight speed of the UAV on the remaining road sections and the allocation of user computing task resources; S4.3: Provide calculation services to users of the corresponding road section based on the adjustment results.

Images