Robust task unloading method for unmanned aerial vehicle assisted mobile edge calculation

Abstract

The invention provides a robust task unloading method for unmanned aerial vehicle assisted mobile edge computing, which comprises the following steps of: establishing an unmanned aerial vehicle assisted mobile edge computing system scene model, and constructing a problem model for minimizing the maximum task completion time of the unmanned aerial vehicle assisted mobile edge computing system under the condition of satisfying energy consumption constraint and time slot limitation. And decomposing the problem model, approaching a non-convex constraint condition by using a continuous convex approximation method, and carrying out alternate iterative solution by combining a block coordinate descent frame. According to the method, the potential of partial unloading of the unmanned aerial vehicle can be fully excavated, and the system delay can be effectively reduced and the overall energy efficiency can be effectively improved under the condition of conforming to the actual NLOS channel condition.

Description

Technical Field

[0001] The present invention belongs to the combination of information and communication engineering technology and Internet of Things technology, and specifically relates to a robust task offloading method for unmanned aerial vehicle (UAV)-assisted mobile edge computing. Background Art

[0002] With the development of the sixth-generation 6G mobile network technology, new mobile devices such as augmented reality (AR), virtual reality (VR), and intelligent navigation will be spawned. These applications have extremely high requirements for computing resources and latency. However, ground user equipment (such as smartphones and Internet of Things devices) is limited by limited computing power and battery budget and is difficult to independently complete complex computing tasks. The UAV-assisted mobile edge computing system (MEC) uses UAVs as airborne mobile edge servers to provide flexible computing offloading services for ground devices, which has become an effective method to solve this problem.

[0003] Many studies have been devoted to the optimization of UAV-assisted MEC systems, mainly by jointly optimizing UAV trajectories, task offloading decisions, and resource allocation to reduce the overall latency and energy consumption of task completion. For example, in special cases where the number of ground devices surges or the infrastructure is sparse, some studies minimize the latency between all devices and UAVs; some studies minimize the system energy consumption while meeting task latency and UAV dynamic constraints. However, the existing studies have the following problems: many studies assume that task offloading is binary and ignore the potential of partial offloading; existing algorithms force UAVs to frequently adjust their trajectories to reduce device energy consumption, resulting in a decrease in the overall energy efficiency of the system. Summary of the Invention

[0004] The present invention proposes a robust task offloading method for UAV-assisted mobile edge computing.

[0005] The technical solution for achieving the object of the present invention is: a robust task offloading method for UAV-assisted mobile edge computing, including:

[0006] Establish a scenario model of a UAV-assisted mobile edge computing system, and construct a problem model that minimizes the maximum task completion time of the UAV-assisted mobile edge computing system under the conditions of meeting energy consumption constraints and time slot limitations;

[0007] Decompose the problem of minimizing the maximum task completion time of the UAV-assisted mobile edge computing system into three sub-problems, namely, the UAV trajectory optimization sub-problem, the ground device transmission power and the proportion of data offloaded to the UAV optimization sub-problem, and the binary offloading decision and UAV computing power optimization sub-problem;

[0008] Sequentially and iteratively solve the three sub-problems, and continuously repeat the iteration until the convergence condition is met to complete the robust task offloading, where the continuous convex approximation method is used to solve each sub-problem.

[0009] Compared with the prior art, the significant advantages of the present invention are as follows:

[0010] The present invention breaks through the limitations of the binary hypothesis of task offloading in traditional research, explores the potential of partial offloading of unmanned aerial vehicles (UAVs), and through the optimization of coupled variables such as the task offloading rate, enables the system to flexibly allocate computing tasks according to the actual situation, improving the efficiency and rationality of task processing.

[0011] In response to the problem that the frequent adjustment of the UAV trajectory leads to a decrease in the overall energy efficiency of the system, the present invention innovatively constructs a three-dimensional joint optimization model by fully considering the strong coupling relationship among trajectory - offloading - power - computing ability. Through the nested solution of the block coordinate descent method (BCD) and the successive convex approximation method (SCA), by jointly optimizing parameters such as the UAV trajectory, computing ability, and device transmission power, it avoids the waste of UAV energy consumption caused by excessive pursuit of device energy conservation, and while ensuring reasonable device energy consumption, significantly improves the overall energy efficiency of the system.

[0012] Based on the NLOS channel for research, it is more in line with the actual deployment scenario. Compared with the existing methods based on the ideal LOS channel, it is more in line with the communication situation in the real environment, effectively reducing the system delay and enhancing the practicability and reliability of the system.

[0013] The following further describes the present invention in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a system diagram of a robust task offloading method model for UAV-assisted mobile edge computing according to an embodiment of the present invention.

[0015] Figure 2 It is a flowchart of an algorithm for a robust task offloading method for UAV-assisted mobile edge computing according to an embodiment of the present invention.

[0016] Figure 3 It is an algorithm convergence effect diagram of a robust task offloading method for UAV-assisted mobile edge computing according to an embodiment of the present invention.

[0017] Figure 4 It is a comparison diagram of the total system energy consumption corresponding to different numbers of ground devices in a robust task offloading method for UAV-assisted mobile edge computing according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] A robust task offloading method for unmanned aerial vehicle (UAV)-assisted mobile edge computing, mainly targeting scenarios where the computing capabilities and battery budgets of ground user devices are limited, and there are many limitations in the existing research on UAV-assisted mobile edge computing (MEC) systems. Considering energy consumption constraints and time slot limitations, it studies how to jointly optimize UAV trajectories, task offloading rates, UAV computing capabilities, device transmission powers, and binary offloading decisions to minimize the maximum delay of the system and improve the overall energy efficiency of the system. It mainly decomposes the problem, uses the successive convex approximation (SCA) method to approximate non-convex constraint conditions, and combines the block coordinate descent (BCD) framework to solve it by alternating iterations. The present invention can not only fully exploit the potential of partial offloading of UAVs, but also effectively reduce the system delay and improve the overall energy efficiency under actual non-line-of-sight (NLOS) channel conditions.

[0019] (1) Establish a UAV-assisted MEC system model based on the NLOS channel, and analyze and derive the objective function for minimizing the maximum delay of the system under the conditions of meeting energy consumption constraints and time slot limitations; (2) Decompose the original problem into three sub-problems, namely, optimizing UAV trajectories , the transmission power of ground devices and the proportion of data offloaded to the UAV optimization, binary offloading decision and UAV computing capability optimization of three sub-problems; (3) Sequentially iterate and optimize, continuously repeat the iteration until the convergence condition is met, and obtain the final optimization result. The specific implementation process is as follows:

[0020] Step 1: Establish a scenario model of a UAV-assisted mobile edge computing (MEC) system, namely, a channel model, a local computing model, and a UAV-assisted edge computing model, and determine the problem model for minimizing the maximum task completion time of the UAV-assisted mobile edge computing (MEC) system under the conditions of meeting energy consumption constraints and time slot limitations.

[0021] As Figure 1 shown in the UAV-assisted mobile edge computing (MEC) system model, it consists of a UAV integrated with an edge server as an aerial base station and ground users, denoted as . The UAV can communicate with ground mobile devices and provide computing services simultaneously. Each user can choose whether to offload a part of its computing tasks to the UAV and execute the rest locally. Assume that the UAV flies at a fixed altitude , is the length of an equidistant time slot, , the horizontal coordinate of the UAV is , and the horizontal coordinate of the ground user is . The maximum speed of the UAV is , and then the UAV trajectory constraint can be shown as follows:

[0022] ; ;

[0023] Among them, the constraint means that the flight speed of the UAV cannot exceed its maximum speed, and the constraint means that the UAV returns to its initial position at the end of the period . The energy consumption caused by the UAV flight is related to its speed. The UAV flight energy consumption model can be expressed as:

[0024]

[0025] Among them is related to the payload of the UAV.

[0026] Channel model:

[0027] The distance between the device and the UAV in the th time slot is expressed as:

[0028]

[0029] The device and the average channel gain between the UAV and the UAV are expressed as:

[0030]

[0031]

[0032] Among them represents the probability that the LoS channel appears between the device and the UAV at the th time slot in the NLoS link, is the probability that the NLoS channel appears between the device and the UAV at the th time slot, and are parameters determined by the external environment, is the elevation angle, in degrees.

[0033] Assume that the maximum transmit power of the device is , then the power constraint is shown as follows:

[0034]

[0035] The information rate between the device and the UAV can be expressed as:

[0036]

[0037] wherein is the total channel bandwidth in Hertz (Hz), represents the noise power spectral density, and due to the actual modulation and coding scheme, there is a gap in the channel capacity, denoted by

[0038] The above distance formula, the channel gain between the device and the UAV, the transmission power, and the information rate constitute the channel model of the present invention.

[0039] Local computing model:

[0040] Use to represent the task that the mobile device needs to execute in each time slot wherein is the proportion of data unloaded to the UAV, is the discrete binary user scheduling variable, the computing task volume in the th time slot, is the number of CPU cycles required for each user to execute each bit. For each time slot , the device has the following local computing delay and local energy consumption:

[0041]

[0042]

[0043] wherein is the maximum computing power (cycles / second) of user , is a constant, depending on the chip architecture of the mobile device.

[0044] The above local computing delay and local computing energy consumption of the device constitute the local computing model of the present invention.

[0045] UAV-assisted edge computing model:

[0046] Based on the above model, in the time slot for user the delay and energy consumption of task offloading are as follows:

[0047]

[0048]

[0049] After task offloading, the time and energy consumption for the UAV to execute the task are:

[0050] ​

[0051]

[0052] Among them It depends on the chip architecture of the UAV. Taking the maximum computing energy of the UAV as, the computing power constraint, time constraint, and energy constraint can be obtained as follows:

[0053]

[0054]

[0055]

[0056]

[0057]

[0058] The offloading computing delay and computing energy consumption of the above device, as well as the time and energy consumption of the UAV to execute the task after task offloading, constitute the UAV-assisted computing offloading model of the present invention.

[0059] Problem description:

[0060] Since the computing result is relatively small compared to the task upload, and the transmission bandwidth allocated to the uplink is much larger than that of the downlink, therefore, the downlink transmission of sending the computing result to the ground device is negligible. Based on the above model, the following problem can be established:

[0061] ;

[0062] Among them, the objective function is the minimum value of the maximum of the local processing time and the task completion time.

[0063] Step 2: Based on the alternating optimization algorithm, the original problem is decomposed into three sub-problems.

[0064] This algorithm decomposes the original problem into three sub-problems: using the successive convex approximation method (SCA) to convert non-convex constraints into convex constraints, and combining the block coordinate descent method (BCD) to decompose the original problem into three sub-problems, namely optimizing the UAV trajectory and the transmit power of the ground device and the proportion of data offloaded to the UAV for optimization, the binary offloading decision and the computing power of the UAV for optimization of the three sub-problems.

[0065] Step 3: Sequentially and iteratively solve three sub-problems (Steps 3.1 - 3.3), and continuously repeat the iterative solution process until the convergence condition is met to obtain the final optimization result.

[0066] The present invention proposes an optimization algorithm for establishing a UAV-assisted MEC system model based on the NLOS channel, analyzes and derives the objective function for minimizing the maximum system delay under the conditions of meeting the energy consumption constraint and time slot limit. This algorithm decomposes the original problem into three sub-problems: UAV trajectory , ground device transmission power , and the proportion of data unloaded to the UAV , binary offloading decision , and UAV computing power .

[0067] First, initialize each variable: the maximum number of time slots , the maximum number of ground users , the proportion of data unloaded to the UAV , the binary user scheduling variable , the th time slot's computing task volume , the number of CPU cycles required for each user to execute each bit , the maximum transmission power , the maximum flight speed of the UAV , the current iteration number ;

[0068] 1) Given the transmission power at the th iteration, the proportion of data unloaded to the UAV , the binary user scheduling variable , the UAV computing power , optimize and solve the trajectory ;

[0069] 2) Using the trajectory obtained in the previous step at the th iteration, given the binary user scheduling variable at the th iteration, the UAV computing power , optimize and solve the ground device transmission power and the proportion of data unloaded to the UAV ;

[0070] 3) Using the trajectory obtained in the previous two steps at the th iteration, the transmission power and the proportion of data unloaded to the UAV , optimize and solve the discrete binary user scheduling variable and the computing power of the UAV ;

[0071] 4) Determine whether the convergence condition is satisfied: , if not, let , return to 1), otherwise use the current result as the final optimization result, being the convergence threshold.

[0072] In a further embodiment, given other variables , the specific method for optimizing and solving the trajectory is as follows:

[0073] The original problem can be transformed into :

[0074]

[0075] Constraints and are non-convex constraints and contain multiple variables (such as and ). To solve this problem, by considering the most likely elevation angle, uniformly approximate and fix it as . Then, the information rate between the device and the UAV can be approximately expressed as:

[0076]

[0077] where , . Introduce an auxiliary variable , set it to 2, such that , at this time it is non-concave, and perform the following transformation . Then, perform Taylor expansion on to obtain:

[0078]

[0079] where . At this time, the problem can be reformulated as as follows:

[0080]

[0081] At this time, both the objective function and the constraint conditions are convex and can be solved using the relevant CVX solver.

[0082] In a further embodiment, using the above steps to obtain , and given other variables , the specific method for optimizing the transmission power of the ground equipment and the data ratio unloaded to the UAV is as follows:

[0083] The transmission power of the ground equipment is optimized and the data ratio unloaded to the UAV , and other variables are given . At this time, the original problem can be transformed into .

[0084]

[0085] Similar to sub-problem 1, introduce an auxiliary variable , such that , that is , and because . At this time, the problem can be reformulated as as follows:

[0086]

[0087] where is jointly convex. At this time, Both the objective function and the constraint conditions are convex and can be solved using the relevant CVX solver.

[0088] In a further embodiment, using the above two steps to obtain , the specific method for optimizing the discrete binary user scheduling variable and the UAV computing power is as follows:

[0089] Optimize the discrete binary device scheduling variable and the UAV computing power given other variables . At this time, the original problem can be transformed into .

[0090]

[0091] Introduce a continuous variable , that is the continuous variable of, that is, introduce the constraint:

[0092]

[0093] Among them , that is , at this time the time constraint can be (replace in other constraints with ;

[0094]

[0095] Define auxiliary variables , , then the energy constraint can be expressed as:

[0096]

[0097] At this time, the problem can be reformulated as as follows:

[0098]

[0099] At this time, both the objective function and the constraint conditions are convex, and the CVX solver is used to solve them.

[0100] In a further embodiment, it is judged whether the convergence condition is satisfied: , if not satisfied, let , return 1), otherwise use the current result as the final optimization result, is the convergence threshold.

[0101] Result description:

[0102] Assume that mobile devices are randomly and uniformly distributed in a 500m×500m area, the fixed height of the UAV , the flight speed does not exceed 50m / s, the elevation angle is fixed at 90 o , the path loss exponent is set to 2.5, the maximum transmit power of the user = 1W, the total channel bandwidth = 1MHZ, the maximum computing power of the mobile device is randomly distributed within [300,400] MHZ, the maximum computing power of the UAV = 1200 MHZ, the noise power spectral density = -100dBm, the payload of the UAV = 6kg, at the reference distance the channel power gain is set to = -40dB, the task size follows (bits) uniform distribution, set cycles / bit, , the battery energy of the device = 0.01 KWh, the battery energy of the drone = 0.01 KWh.

[0103] The performance of the proposed joint optimization strategy (Algorithm 1) is verified through simulation experiments. It is compared with three benchmark schemes: the device transmit power and drone computing energy optimization algorithm (Algorithm 2), the genetic algorithm (Algorithm 3), and the full offloading (Algorithm 4).

[0104] Figure 3 It shows that our proposed solution gradually converges to a sub-optimal solution after several iterations. Further observation of this figure reveals that compared with Algorithm 3, our proposed solution exhibits a faster convergence speed, which means that under the same conditions, it can approach the ideal result more efficiently, saving a large amount of time and computational cost for practical applications.

[0105] Figure 4 A comparison of four algorithms showing the total energy consumption of the drone, including flight energy consumption, edge computing energy consumption, and device energy consumption such as upload energy consumption and local computing energy consumption, clearly shows that the energy consumption of the drone under the proposed method in this paper is lower than that of other methods. Except for Algorithm 4, the device energy consumption is lower than that of the other two algorithms. However, the total energy consumption of the drone and the device under the proposed method in this paper is much lower than that of Algorithm 4 because Algorithm 4 chooses to offload all tasks to the drone for execution, and the device only has upload energy consumption.

[0106] The present invention aims at the problem of multi-variable coupling optimization in a drone-assisted mobile edge computing system and proposes a joint optimization method based on successive convex approximation (SCA) and block coordinate descent (BCD). The proposed joint optimization method takes into account both energy consumption efficiency and real-time requirements, providing theoretical support and technical reference for the Internet of Things scenarios with high-density users and dynamic task requirements in the drone-assisted edge computing system.

Claims

1. A robust task offloading method for unmanned aerial vehicle assisted mobile edge computing, characterized in that Including: Establish a scenario model of a UAV-assisted mobile edge computing system. Under the conditions of meeting the energy consumption constraint and time slot limit, construct a problem model that minimizes the maximum task completion time of the UAV-assisted mobile edge computing system; Decompose the problem of minimizing the maximum task completion time of the UAV-assisted mobile edge computing system into three sub-problems, namely the UAV trajectory optimization sub-problem, the optimization sub-problem of the transmit power of ground devices and the data ratio offloaded to the UAV, and the binary offloading decision and UAV computing ability optimization sub-problem; Iteratively solve the three sub-problems in sequence and continuously repeat the iteration until the convergence condition is met to complete robust task offloading. Among them, the continuous convex approximation method is used to solve each sub-problem.

2. The robust task offloading method for drone-assisted mobile edge computing according to claim 1, wherein The described drone-assisted mobile edge computing system includes a drone integrated with an edge server and ground users, denoted as . The drone serves as an aerial base station, capable of communicating with ground mobile devices and providing computing services simultaneously. Each user chooses whether to offload a portion of their computing tasks to the drone and execute the remaining portion locally.

3. The robust task offloading method for drone-assisted mobile edge computing according to claim 1, wherein The established scenario model of the UAV-assisted mobile edge computing system includes a channel model, a local computing model, and a UAV-assisted edge computing model. The channel model includes the channel gain, transmit power, and information rate between the device and the UAV; the local computing model includes the local computing delay and the energy consumption of local computing; the UAV-assisted edge computing model includes the offloading computing delay, the energy consumption of computing, and the time and energy consumption of the UAV executing tasks after task offloading.

4. The robust task offloading method for drone-assisted mobile edge computing according to claim 1, characterized in that, Under the conditions of meeting the energy consumption constraint and time slot limit, the constructed problem model that minimizes the maximum task completion time of the UAV-assisted mobile edge computing system is specifically: ； Among them, ; Wherein, is the time slot length, is the maximum computing power of the UAV, is the computing power of the UAV, is the battery energy of the device, is the battery energy of the UAV, is the maximum tolerable time for task offloading and computing to complete the task, is the maximum tolerable time for local computing, is the maximum flight speed of the UAV, is the trajectory of the UAV, is the maximum transmission power of the ground device, is the transmission power of the ground device, is the proportion of data offloaded to the UAV, is the discrete binary user scheduling variable, and are the time and energy consumption of device offloading respectively, and are the time and energy consumption of local computing of the device respectively, and are the time and energy consumption of the UAV for processing offloaded tasks respectively, is the energy consumption of the UAV during flight, , , , is the set of the number of ground users, is the maximum number of devices, is the set of the number of time slots, is the maximum number of time slots.

5. The robust task offloading method for drone-assisted mobile edge computing according to claim 4, wherein The specific process of iteratively solving the three sub-problems in sequence is: Initialize the maximum number of variable time slots , the maximum number of ground users , the proportion of data unloaded to the UAV , the binary user scheduling variable , the th computing task volume of the time slot , the number of CPU cycles required for each user to execute each bit , the maximum transmission power , the maximum flight speed of the UAV , the current iteration number ; 1) Given the transmission power at the th iteration, the proportion of data unloaded to the UAV, the binary user scheduling variable, the computing ability of the UAV, the optimized solution trajectory ; ; ; 2) Using the trajectory at the th iteration obtained, given the binary user scheduling variable at the th iteration, the computing power of the UAV, optimally solve for the transmission power of the ground device and the proportion of data unloaded to the UAV; ; 3) Using the trajectory at the th iteration obtained, the transmit power and the proportion of data unloaded to the UAV, optimally solve the discrete binary user scheduling variable and the computing power of the UAV ; 4) Determine whether the convergence condition is satisfied: , if not satisfied, let , return to 1), otherwise use the current result as the final optimization result, is the convergence threshold.

6. The robust task offloading method for drone-assisted mobile edge computing according to claim 5, characterized in that Given variable value, optimize the solution trajectory The specific process is as follows: Given variable , the problem is transformed into , specifically as follows: ； The device and the probability that the LoS channel appears at the th time slot of the UAV are uniformly approximated and fixed as . The information rate between the device and the UAV is approximately expressed as ​ ； Among them , ; Introduce auxiliary variables , set to 2.5, such that , at this time it is non-concave, perform the following transformation ; Then, for performing Taylor expansion gives: ； Among them ; Problem Restated as , specifically as follows: ； At this time, both the objective function and the constraints are convex, and the CVX solver is used to solve them.

7. The robust task offloading method for drone-assisted mobile edge computing according to claim 5, characterized in that, The trajectory at the th iteration obtained by using the optimization solution , and given the value of the variable , the specific method for optimizing the transmission power of the ground equipment and the data ratio unloaded to the UAV is as follows: Optimized the transmission power of ground equipment and the data ratio unloaded to the UAV , and other variables are given below ; At this time, the original problem is transformed into : ； Introduce auxiliary variables such that i.e., and since ; at this time, the problem is reformulated as specifically: ； Among them is jointly convex. At this time both the objective function and the constraints are convex, and the CVX solver is used to solve them.

8. The robust task offloading method for UAV-assisted mobile edge computing according to claim 5, characterized in that Using the obtained , the specific method for optimizing the discrete binary user scheduling variable and the computing power of the UAV is as follows: Optimize the scheduling variables of discrete binary devices and the computing power of drones for optimization. Given other variables at this time, the original problem is transformed into , specifically: ； Introduce continuous variables , which is 's continuous variable, that is, introduce constraints ； Among them , that is , the time constraint is as follows: ； Define auxiliary variables , , the energy constraint is expressed as: ； The problem at this time is reformulated as , specifically as follows: ； At this time, both the objective function and the constraints are convex, and the CVX solver is used to solve them.

Images