Method, system, device and medium for planning and scheduling of a drone MEC network

Abstract

The application relates to the technical field of edge computing, and discloses a planning and scheduling method, system, device and medium for a UAV MEC network, the method comprising the following steps: constructing a network transmission model of an edge computing network according to total flight time delay and total energy consumption of a UAV; obtaining model task utility according to task priority and task processing time delay; constructing a planning function with the goal of maximizing the model task utility, and adopting a deep reinforcement learning algorithm to make a decision on the planning function to obtain an optimal user offloading strategy and an optimal UAV movement path. The application jointly optimizes the access path of the UAV and the task offloading strategy of the user, can maximize network performance under the constraints of UAV battery capacity and flight duration, effectively solves the complex task allocation and path planning problem, improves the problem solving efficiency, and further improves the network operation efficiency and network resource utilization.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of edge computing, in particular to a method, system, device and medium for planning and scheduling of a UAV MEC network. BACKGROUND

[0002] Mobile edge computing (MEC) technology is proposed to provide a low-latency, high-performance and high-bandwidth service environment for end users (EUs) by deploying computing servers at edge nodes. This technology plays a crucial role in Beyond 5G (B5G) networks. However, traditional MEC networks can only provide services within a limited coverage range, and do not take into account the changing needs of end users (EUs) over time. Therefore, one of the fundamental challenges for B5G is to effectively expand the service coverage of MEC networks and ensure reliable service delivery in the expanded range.

[0003] In recent years, the integration of unmanned aerial vehicles (UAVs) and MEC networks has become a promising solution to this challenge. This approach involves equipping UAVs with edge computing units and then using the flexibility of UAVs to reach areas beyond the coverage of traditional edge servers or where these servers are overloaded or malfunctioning. Therefore, by deploying UAVs into MEC networks, computing services can be provided in a timely manner to enhance MEC performance. However, the above work mainly focuses on UAVs and ignores the characteristics of tasks, i.e., task priority in MEC networks. Specifically, task priority determines the order of priority among different tasks, which is quantified by the importance, urgency and resource requirements of the task. In general, high-priority tasks, such as real-time video streaming and virtual reality applications, require fast response; while low-priority tasks, such as batch processing or non-real-time computing, can be executed during system idle periods.

[0004] Currently, the existing problems of the UAV path planning method are that, on the one hand, there is no dynamic adjustment strategy for task priority and size, which leads to unreasonable allocation between urgent tasks and non-urgent tasks, reducing the efficiency of the network. On the other hand, the existing technology cannot meet the maximum system performance under the constraints of UAV battery capacity and flight duration, so it is difficult to solve the performance maximization problem of the access path of the UAV and the unloading strategy of the EUs. SUMMARY

[0005] To solve the above technical problems, the present application provides a method, system, device and medium for planning and scheduling of a UAV MEC network, which can solve the problem that the existing method cannot dynamically adjust the task priority, resulting in low network performance efficiency, and achieve the effect of improving network operation efficiency and resource utilization.

[0006] In a first aspect, the present application provides a planning and scheduling method for a UAV MEC network, the method comprising:

[0007] According to the total flight time delay and the total energy consumption of the UAV, a network transmission model of the edge computing network is constructed, the edge computing network comprising a plurality of hotspot areas, a plurality of users contained in each hotspot area, and a UAV-mounted mobile edge computing server;

[0008] According to the task priority and the task processing time delay, a model task utility is obtained;

[0009] A planning function is constructed with the goal of maximizing the model task utility, and a deep reinforcement learning algorithm is used to make decisions on the planning function to obtain an optimal user offloading strategy and an optimal UAV movement path.

[0010] Further, the step of constructing the network transmission model of the edge computing network according to the total flight time delay and the total energy consumption of the UAV comprises:

[0011] According to the reachable communication transmission rate of the users in the hotspot area, the total hovering time delay of the UAV is calculated;

[0012] According to the total hovering time delay, the horizontal movement time delay and the horizontal movement energy consumption of the UAV between adjacent hotspot areas, the total flight time delay and the total energy consumption of the UAV are calculated, and the total flight time delay and the total energy consumption of the UAV are taken as the network transmission model of the edge computing network.

[0013] Further, the step of calculating the total hovering time delay of the UAV according to the reachable communication transmission rate of the users in the hotspot area comprises:

[0014] The local computing time delay when the user selects local computing is obtained;

[0015] According to the reachable communication transmission rate of the users in the hotspot area, the offloading computing time delay when the user offloads local computing is calculated;

[0016] According to the local computing time delay and the offloading computing time delay, a first task processing time delay of the user is obtained;

[0017] According to the first task processing time delay, the hovering time delay of the UAV in each hotspot area is calculated, and according to the hovering time delay, the total hovering time delay of the UAV is obtained.

[0018] Further, the step of obtaining the model task utility according to the task priority and the task processing time delay comprises:

[0019] According to the first task processing time delay, the horizontal movement time delay and the hovering time delay, a second task processing time delay of the user is obtained;

[0020] According to the second task processing delay and a preset task priority, a user task utility is obtained;

[0021] According to the user task utility, a model task utility is obtained.

[0022] Further, the deep reinforcement learning algorithm comprises a Markov decision algorithm.

[0023] Further, the total flight delay is expressed by the following formula:

[0024]

[0025] In the formula, UAV represents a UAV, represents a horizontal movement delay of the UAV from an (i-1)th hotspot area to an ith hotspot area, represents a hovering delay of the UAV in the ith hotspot area, and L represents a number of hotspot areas.

[0026] The total energy consumption is expressed by the following formula:

[0027]

[0028] In the formula, represents a horizontal movement energy consumption of the UAV from the (i-1)th hotspot area to the ith hotspot area, P(0) represents a horizontal flight propulsion power of the UAV at a speed of 0, ρ m,i represents an offloading rate of an mth user in the ith hotspot area, κ represents an energy consumption coefficient, d i-1,i represents a horizontal distance from the (i-1)th hotspot area to the ith hotspot area, ξ represents a CPU frequency required for calculating each bit of task, f u represents a UAV-borne CPU computing frequency.

[0029] Further, the model task utility is expressed by the following formula:

[0030]

[0031] In the formula, u m,i represents a user task utility of an mth user in an ith visited hotspot area.

[0032] In a second aspect, the application provides a planning and scheduling system of a UAV MEC network, the system comprising:

[0033] a transmission model construction module, configured to construct a network transmission model of an edge computing network according to a total flight delay and a total energy consumption of a UAV, the edge computing network comprising a plurality of hotspot areas, a plurality of users contained in each hotspot area, and a UAV-borne mobile edge computing server;

[0034] a task utility calculation module, configured to obtain model task utility according to task priority and task processing time delay;

[0035] an optimal planning and scheduling module, configured to construct a planning function with the goal of maximizing model task utility, and to make decisions on the planning function by using a deep reinforcement learning algorithm to obtain an optimal user offloading strategy and an optimal UAV movement path.

[0036] In a third aspect, an embodiment of the present application further provides a computer device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above method when executing the computer program.

[0037] In a fourth aspect, an embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executable on a processor to implement the steps of the above method.

[0038] The present application provides a planning and scheduling method, system, device and storage medium for a UAV MEC network. According to the priority and size of the task, the present application dynamically allocates resources and optimizes task distribution to meet different task requirements, and jointly optimizes the access path of the UAV and the task offloading strategy of the end user under the constraint condition of maximizing system performance, which can effectively solve the complex task allocation and path planning problem, improve the problem solving efficiency, and improve the network operation efficiency and resource utilization. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 is a flowchart of the planning and scheduling method in the embodiment of the present application;

[0040] Figure 2 is a system architecture diagram of the edge computing network in Figure 1

[0041] Figure 3 is a system convergence comparison diagram of different schemes in the simulation comparison experiment in the embodiment of the present application;

[0042] Figure 4 is a system performance comparison diagram of different schemes based on the number of users in the simulation comparison experiment in the embodiment of the present application;

[0043] Figure 5 is a system performance comparison diagram of different schemes based on user bandwidth in the simulation comparison experiment in the embodiment of the present application;

[0044] Figure 6 is a system performance comparison diagram of different schemes based on the computing capacity of the UAV in the simulation comparison experiment in the embodiment of the present application; ​

[0045] Figure 7 is a system performance trend diagram of the scheme under different energy consumption thresholds and different UAV speeds in the simulation comparison experiment in the embodiment of the application;

[0046] Figure 8 is a structure diagram of the planning and scheduling system in the embodiment of the application;

[0047] Figure 9 is an internal structure diagram of the computer device in the embodiment of the application. DETAILED DESCRIPTION

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the application clearer, the technical solutions in the embodiments of the application will be described below in conjunction with the accompanying drawings in the embodiments of the application. Obviously, the described embodiments are some but not all of the embodiments of the application. Based on the embodiments in the application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the application.

[0049] Referring to Figure 1 , the planning and scheduling method of the UAV MEC network proposed in the first embodiment of the application includes steps S10-S30:

[0050] Step S10, constructing a network transmission model of an edge computing network according to total flight delay and total energy consumption of a UAV, the edge computing network including a plurality of hot spot areas, a plurality of users contained in each hot spot area, and a UAV-mounted mobile edge computing server;

[0051] Step S20, obtaining a model task utility according to a task priority and a task processing delay;

[0052] Step S30, constructing a planning function with the goal of maximizing the model task utility, and using a deep reinforcement learning algorithm to make a decision on the planning function to obtain an optimal user offloading strategy and an optimal UAV movement path.

[0053] The application proposes an optimal path planning method for a new multi-user multi-hotspot MEC network adapted to UAV support for an integrated network of a UAV and a mobile edge computing EMC. Before the application is described in detail, the structure of the integrated network is first introduced.

[0054] The edge computing network in the application includes a plurality of hot spot areas, a plurality of users contained in each hot spot area, and a UAV-mounted mobile edge computing server, and the hot spot areas are set to L, and each hot spot area contains M users, that is, the user set in the hot spot area can be represented as M={u1,…,u M} can be expressed as L={v1, …, v L} where one rotary-wing UAV departs from a base station, visits L hotspots in a specified area in sequence at a fixed height, and provides computing services to M users (EUs) through wireless communication during each hotspot visit. Subsequently, the UAV returns to the base station to replenish energy. To ensure the rationality of the system, the energy consumption overhead and flight duration of the UAV should not exceed the corresponding thresholds respectively.

[0055] Based on the above edge computing network, the application uses free space path loss to construct a network transmission model, and the specific steps are as follows:

[0056] Step S101, according to the reachable communication transmission rate of the user in the hotspot area, the total hovering time delay of the UAV is calculated;

[0057] Step S102, according to the total hovering time delay, the horizontal movement time delay and the horizontal movement energy consumption of the UAV between adjacent hotspots, the total flight time delay and the total energy consumption of the UAV are calculated, and the total flight time delay and the total energy consumption of the UAV are taken as the network transmission model of the edge computing network.

[0058] In this embodiment, according to the reachable communication transmission rate of the user in the hotspot area, the task processing time delay of the user can be calculated, and then the total hovering time delay of the UAV is calculated, and the specific steps include:

[0059] The local computing time delay when the user selects local computing is obtained;

[0060] According to the reachable communication transmission rate of the user in the hotspot area, the offloading computing time delay when the user offloads local computing is calculated;

[0061] According to the local computing time delay and the offloading computing time delay, the first task processing time delay of the user is obtained;

[0062] According to the first task processing time delay, the hovering time delay of the UAV in each hotspot area is calculated, and according to the hovering time delay, the total hovering time delay of the UAV is obtained.

[0063] In this embodiment, the reachable communication transmission rate of the user in the hotspot area can be expressed as:

[0064]

[0065] In the formula, b represents the wireless bandwidth of the user, p Trans represents the transmission power of the user, represents the channel gain, μ represents the reference channel gain, H represents the flight height of the UAV, σ 2denotes the variance of the additive white Gaussian noise (AWGN).

[0066] In the embodiment, the user m adopts a binary offloading method, that is, the user can select a local computing mode and an edge computing mode. When the user selects the local computing, the local computing delay of the user is:

[0067]

[0068] wherein s m,l denotes the task quantity of the mth user in the lth hotspot area, ξ denotes the CPU frequency required for computing each bit of task, f m,l denotes the CPU frequency of the mth user in the lth hotspot area.

[0069] When the user selects to offload the local computing, the offloaded computing delay of the user is:

[0070]

[0071] wherein f u denotes the CPU computing frequency of the unmanned aerial vehicle.

[0072] According to the formulas (2) and (3), the first task processing delay of the user can be obtained:

[0073]

[0074] wherein ρ m,l ∈{0,1} denotes the offloading rate of the mth user in the lth hotspot area.

[0075] When the user in the hotspot area offloads the local computing and adopts the edge computing, the task is completed by the mobile edge computing server on board of the unmanned aerial vehicle, therefore, the hovering delay of the unmanned aerial vehicle UAV in the hotspot area v l can be expressed as:

[0076]

[0077] According to the hovering delay of the unmanned aerial vehicle in each hotspot area, the total hovering delay of the unmanned aerial vehicle can be obtained:

[0078]

[0079] After the total hovering delay of the unmanned aerial vehicle, that is, the first task processing delay of the user, is calculated, the flight consumption of the unmanned aerial vehicle is also considered in the embodiment, wherein the flight consumption includes the horizontal movement delay and the horizontal movement energy consumption of the unmanned aerial vehicle. Specifically, the horizontal movement delay and the horizontal movement energy consumption of the unmanned aerial vehicle from the hotspot area v l to the hotspot area v l+1 can be respectively expressed as:

[0080]

[0081]

[0082] wherein d i,i+1 represents the horizontal distance from the ith hotspot region visited by the UAV to the (i+1)th hotspot region visited by the UAV, V u represents the horizontal flight speed of the UAV, P(V u ) represents the horizontal flight propulsion power of the UAV at the speed of V u .

[0083] On the basis of the horizontal movement time delay and the horizontal movement energy consumption of the UAV, in combination with the total hovering time delay of the UAV, the total flight time delay and the total energy consumption of the UAV can be obtained:

[0084] The total flight time delay is expressed by the following formula:

[0085]

[0086] wherein UAV represents the UAV, represents the horizontal movement time delay of the UAV from the (i-1)th hotspot region to the ith hotspot region, represents the hovering time delay of the UAV in the ith hotspot region, and L represents the number of hotspot regions.

[0087] The total energy consumption is expressed by the following formula:

[0088]

[0089] wherein, represents the horizontal movement energy consumption of the UAV from the (i-1)th hotspot region to the ith hotspot region, P(0) represents the horizontal flight propulsion power of the UAV at the speed of 0, and p m,i represents the offloading rate of the mth user in the ith hotspot region, and κ represents the energy consumption coefficient, d i-1，i represents the horizontal distance from the (i-1)th hotspot region to the ith hotspot region, and ξ represents the CPU frequency required for calculating each bit of task, f u represents the UAV-borne CPU computing frequency.

[0090] Further, on the basis of the first task processing time delay of the user, in combination with the horizontal movement time delay and the total hovering time of the UAV, the actual time delay of the user processing task in the edge computing network, i.e. the second task processing time delay, can be obtained:

[0091]

[0092] Through the above modeling steps, the network transmission model of the edge computing network can be obtained, and in actual application, the priority of the task needs to be considered, the priority of the task determines the priority order between different tasks, and different orders will affect the network performance, in order to optimize the utilization rate of network resources and improve the network performance, the application provides a system utility model based on the priority of the task and the task processing delay to evaluate the system performance of the network, and the specific steps are as follows:

[0093] According to the first task processing delay, the horizontal movement delay and the hovering delay, the second task processing delay of the user is obtained;

[0094] According to the second task processing delay and the preset task priority, the user task utility is obtained;

[0095] According to the user task utility, the model task utility is obtained.

[0096] In the embodiment, the actual utility of task processing is calculated according to the urgency of the task and the actual delay of the user in processing the task, so as to evaluate the performance of the network, wherein β is set as the priority of the task, for the non-urgent task with β=0, the actual utility of task processing can be represented as:

[0097]

[0098] Wherein, Γ L represents the reward measure for completing the task within a given non-urgent delay threshold δ L , c is a decay factor, Γ L and c are usually positive real numbers, and Φ(x) represents a piecewise function, which is equal to 1 if x is true, otherwise 0.

[0099] For the urgent task with β=1, the actual utility of task processing, i.e. the task utility, can be represented as:

[0100]

[0101] Wherein, δ H is the delay threshold of the urgent task, and Γ H is a penalty factor.

[0102] When β∈(0,1), the task processing utility in this case is calculated using the interpolation method according to the task utility of the above two priorities.

[0103] Specifically, for the mth user in the ith visited hotspot area, the task utility can be represented as:

[0104]

[0105] Wherein, βm,i represents the task priority of the mth user in the ith visited hotspot region.

[0106] According to the task utility of each user, the task utility of the model, that is, the system performance expression of the network, can be obtained:

[0107]

[0108] Further, the present application considers the influence of the UAV access order x and the user offloading strategy p on the system performance, and formulates a system performance maximization problem, that is, a planning function, according to the above system performance expression:

[0109] object: max {x，ρ} U

[0110] constraints: EC UAV ≤ EC th , T UAV ≤ T th

[0111] wherein EC th represents the energy consumption threshold of the UAV, and T th represents the flight time threshold of the UAV.

[0112] Based on the above system performance maximization problem, the embodiment provides a deep reinforcement learning strategy for optimizing task offloading and access order. Specifically, a Markov Decision Process (MDP) framework is used to model the problem. In the MDP environment, the user and the UAV are regarded as agents. The agent can make actions by observing the environment, and the environment will give the agent a reward and make it enter a new state. The agent and the environment interact continuously until an optimal offloading strategy and access path are found. The environment that the agent can observe is:

[0113] s(t) = {x(t), p(t)}

[0114] wherein x(t) and p(t) represent the user offloading strategy and the access order of the hotspot region of the UAV at the tth time, respectively.

[0115] The set of actions that the agent can make is:

[0116] A(t) = {a(t) | a(t) = {a p (t), a o (t)}}

[0117] wherein a p (t) is the space of access paths, and a o (t) is the space of offloading decisions.

[0118] The agent can make the reward obtained by the following decision:

[0119]

[0120] Wherein, U(t-1) and U(t) represent the system performance at t-1 time and the system performance at t time respectively, c1, c2 and c3 are three positive real numbers. After the end of each time slot, the agent will store {s(t-1), s(t), a(t), r(t)} to the experience cache pool for updating the DRL algorithm. Specifically, after the agent and the environment interact for a certain number of times, the agent will start to randomly extract a fixed number of {s(t-1), s(t), a(t), r(t)} from the experience cache pool to update the loss function of the decision layer of the DRL algorithm:

[0121]

[0122] Wherein, H(s(t)) represents the degree of dispersion of state s at t time, Q w is used to evaluate the value of taking action a in state s.

[0123] Through the above deep reinforcement learning algorithm, the optimal user offloading strategy and the access order of the hot spot area, that is, the UAV movement trajectory, can be obtained.

[0124] The effect of the planning and scheduling method provided by the present application is verified through simulation comparison experiments. It is assumed that the system diagram of the airborne MEC server network model is as shown in Figure 2 Each hot spot area has a plurality of users. The task size and task priority of each user are variable.

[0125] In this experiment, the UAV is configured to operate at a flight height of 100m, and the reference channel gain at a distance of 1m is set to -60dB. The transmission power of the EU is uniformly set to 0.1W, and the channel noise is -110dBm. The allocated channel bandwidth of each EU is 6MHz, and the computing capacity of the UAV is 1x10 8 cycles / s. Moreover, the computing task size and task priority of the EU follow uniform distribution, and the distribution is represented as:

[0126]

[0127]

[0128] The computing task workload is 5 cycles / s, and the computing capacity of the EU is set to 5x10 6The flight speed of the UAV is set to 18 m / s. In addition, the task processing threshold of the non-emergency task δ L is set to 10 seconds, and the related reward factor Γ L is 0.1. The task processing threshold of the emergency task δ H is 7 seconds, and the penalty factor Γ H is 1. In addition, there are a total of 3 hotspot areas in this experiment, each with 5 users, and the service radius of the UAV is set to 100 m, and the energy consumption threshold and flight time threshold of the UAV are 5 kJ and 60 s respectively.

[0129] Based on the above settings, five different planning and scheduling schemes are simulated in the Python simulation environment, and the five schemes are the scheme, the fixed path optimization unloading scheme, the full unloading optimization path scheme, the full local scheme and the random scheme. The number of training rounds in the simulation is from 0 to 300, and according to the simulation experiment, the system performance comparison chart can be obtained as Figures 3 to 7 .

[0130] Figure 3 The system convergence of different schemes is shown, and it can be seen from Figure 3 that the system performance of the scheme increases at the beginning, and then converges to about 4.6 after 150 rounds, which shows that the scheme is efficient for the UAV MEC network. It can be seen that the scheme is obviously better than the other four methods in terms of system performance, which is because the scheme considers the influence of the UAV access path and the EU unloading strategy.

[0131] Figure 4 The system performance of the five schemes under different numbers of user EU is shown, and the number of EU increases from 3 to 7. As shown in Figure 4 , the system performance of the scheme, the fixed path optimization unloading scheme and the full unloading optimization path scheme increases, while the system performance of the full local scheme and the random scheme decreases. This is because the first three schemes effectively utilize the mobility of the UAV and the computing power of the EU. In addition, the system performance of the method proposed by the present application is higher than that of other schemes, because the scheme adopts the DRL algorithm to obtain a better UAV access path and a reasonable EU unloading strategy.

[0132] Figure 5 The influence of wireless bandwidth on the system performance of the five schemes is shown, and the wireless bandwidth varies from 2MHz to 10MHz. As can be seen from Figure 5 , the performance of all schemes increases with the increase of wireless bandwidth, except for the full local scheme. This is because a larger bandwidth will result in smaller transmission delay, thereby improving system performance.

[0133] Figure 6The impact of UAV computing power on system performance is described, where computing power ranges from 1×10 8 cycles / s increased to 5 × 10 8 cycles / s. From Figure 6 It can be seen that the performance of this scheme, the fixed path optimized unloading scheme, the full unloading optimized path scheme, and the random scheme all increase with increasing computing power. This is because the greater computing power of the UAV reduces the computational latency of the unloading task, further reducing the actual task processing latency of the EU, thereby improving system performance. Furthermore, the system performance of this scheme is significantly better than other schemes. Specifically, when the computing power is 5×10... 8 At cycles / s, the performance of this scheme is at least about 2.6 times better than other schemes, because this scheme optimizes the EU's offloading decision and the UAV's access path.

[0134] Figure 7 The paper depicts the system performance trend of this scheme with respect to the speed of the UAV under different energy consumption thresholds, where the UAV speed ranges from 0 m / s to 60 m / s. From Figure 7 It can be seen that when the speed changes from 0 m / s to 5 m / s, the system performance of this scheme remains stable at around -3, because at this speed the UAV is insufficient to provide service, and the EU can compute tasks locally. Furthermore, after 5 m / s, the system performance of the proposed methods with different energy consumption thresholds increases with increasing speed, then drops sharply to around -3 after reaching a peak. This is because once the UAV reaches the minimum speed at which it can provide service, higher speeds lead to increased system performance, but at the cost of higher energy consumption. However, when the UAV's energy consumption exceeds the energy threshold, it loses its service capability, and the EU must compute tasks locally. Moreover, because the UAV operates at higher speeds with larger energy consumption thresholds, the proposed methods with larger energy consumption thresholds achieve higher peak system performance.

[0135] This embodiment provides a planning and scheduling method for UAV MEC networks. Compared with traditional methods, which suffer from low network performance efficiency due to the inability to dynamically adjust task priorities, this invention maximizes network performance while satisfying the constraints of UAV battery capacity and flight duration by jointly optimizing the UAV access path and the user's task offloading strategy. This effectively solves the complex task allocation and path planning problems, improves the efficiency of problem solving, and further enhances network operating efficiency and network resource utilization.

[0136] Please see Figure 8 Based on the same inventive concept, the second embodiment of this invention proposes a planning and scheduling system for a drone MEC network, comprising:

[0137] The transmission model construction module 10 is configured to construct a network transmission model of an edge computing network according to total flight time delay and total energy consumption of the UAV, the edge computing network comprising a plurality of hotspot areas, a plurality of users contained in each hotspot area, and a UAV-mounted mobile edge computing server;

[0138] The task utility calculation module 20 is configured to obtain a model task utility according to a task priority and a task processing time delay.

[0139] The optimal planning and scheduling module 30 is configured to construct a planning function with the goal of maximizing the model task utility, and to make a decision on the planning function by using a deep reinforcement learning algorithm to obtain an optimal user offloading strategy and an optimal UAV movement path.

[0140] The technical features and technical effects of the planning and scheduling system of the UAV MEC network according to the embodiments of the present application are the same as those of the method according to the embodiments of the present application, and will not be repeated here. The various modules in the planning and scheduling system of the UAV MEC network can be realized by software, hardware, or a combination thereof, in whole or in part. The various modules can be embedded in or independent of the processor in the computer device in hardware form, or can be stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to the various modules.

[0141] In addition, the embodiments of the present application also propose a computer device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the above method when executing the computer program.

[0142] Please refer to Figure 9 , an internal structure diagram of a computer device in an embodiment, which can be a terminal or a server. The computer device comprises a processor, a memory, a network interface, a display, and an input device connected through a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The network interface of the computer device is configured to communicate with external terminals through a network connection. The computer program is executed by the processor to implement the planning and scheduling method of the UAV MEC network. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball, or touchpad arranged on the shell of the computer device, or an external keyboard, touchpad, or mouse, etc.

[0143] Those skilled in the art can understand that,Figure 9 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. A specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have the same arrangement of components.

[0144] In addition, the embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the method.

[0145] To sum up, the embodiment of the present application provides a method, system, device and medium for planning and scheduling of a UAV MEC network. The method constructs a network transmission model of an edge computing network according to total flight time delay and total energy consumption of a UAV, the edge computing network including a plurality of hot spot areas, a plurality of users contained in each hot spot area, and a UAV-mounted mobile edge computing server. A model task utility is obtained according to a task priority and a task processing time delay. A planning function is constructed with the goal of maximizing the model task utility, and a deep reinforcement learning algorithm is used to make decisions on the planning function to obtain an optimal user offloading strategy and an optimal UAV movement path. The present application jointly optimizes the access path of the UAV and the task offloading strategy of the user, can maximize the network performance under the constraints of the battery capacity and the flight duration of the UAV, effectively solves the complex task allocation and path planning problem, improves the problem solving efficiency, and further improves the network operation efficiency and network resource utilization.

[0146] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts of each embodiment can be referred to each other. Each embodiment focuses on the difference from other embodiments. Especially, the system embodiment is basically similar to the method embodiment, so the description is relatively simple, and the related parts can be referred to the part of the method embodiment. It should be noted that, each technical feature of the above-mentioned embodiments can be combined arbitrarily, in order to make the description simple, each technical feature of the above-mentioned embodiments is not described in all possible combinations, however, as long as the combination of these technical features does not exist contradictory, it should be considered that it is within the scope of the present application.

[0147] The above-described embodiments only express several preferred embodiments of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be noted that, for ordinary skilled in the art, without departing from the technical principles of the present application, a number of improvements and replacements can be made, which should also be considered as the protection scope of the present application. Therefore, the protection scope of the patent of the present application should be subject to the protection scope of the claims.

Claims

1. A planning and scheduling method for an unmanned aerial vehicle (UAV) MEC network, characterized in that, The method includes: Based on the total flight latency and total energy consumption of the UAV, a network transmission model for the edge computing network is constructed. The edge computing network includes several hotspot areas, several users in each hotspot area, and a mobile edge computing server on the UAV. The model's task utility is obtained based on task priority and task processing delay. A planning function is constructed with the goal of maximizing the utility of the model task, and a deep reinforcement learning algorithm is used to make decisions on the planning function to obtain the optimal user unloading strategy and the optimal UAV movement path. The step of constructing the network transmission model of the edge computing network based on the total flight latency and total energy consumption of the UAV includes: Calculate the total hovering latency of the drone based on the achievable communication transmission rate of users in the hotspot area; Based on the total hovering delay, the horizontal movement delay and horizontal movement energy consumption of the UAV between adjacent hotspot areas, the total flight delay and total energy consumption of the UAV are calculated, and the total flight delay and total energy consumption of the UAV are used as the network transmission model of the edge computing network. The step of calculating the total hovering delay of the UAV based on the achievable communication transmission rate of users in the hotspot area includes: Get the local computing latency when the user selects local computing; Calculate the unloading computation latency when a user unloads local computation based on the achievable communication transmission rate of the user in the hotspot area. The user's first task processing latency is obtained based on the local calculation latency and the unloading calculation latency; Based on the processing delay of the first task, the hovering delay of the UAV in each hotspot area is calculated, and the total hovering delay of the UAV is obtained based on the hovering delay. The steps for obtaining the model task utility based on task priority and task processing delay include: The user's second task processing latency is obtained based on the first task processing latency, the horizontal movement latency, and the hovering latency; The user task utility is obtained based on the second task processing delay and the preset task priority; Based on the user task utility, the model task utility is obtained.

2. The planning and scheduling method for UAV MEC networks according to claim 1, characterized in that, The deep reinforcement learning algorithm includes the Markov decision algorithm.

3. The planning and scheduling method for UAV MEC networks according to claim 1, characterized in that, The total flight delay is expressed by the following formula: In the formula, UAV represents unmanned aerial vehicle (UAV). This represents the horizontal movement delay of the drone from the (i-1)th hotspot area to the ith hotspot area. L represents the hovering delay of the drone in the i-th hotspot area, and L represents the number of hotspot areas. The total energy consumption is expressed by the following formula: In the formula, Let P(0) represent the energy consumption of the UAV moving horizontally from the (i-1)th hotspot area to the ith hotspot area, and let ρ represent the horizontal propulsion power of the UAV when its speed is 0. m,i Let κ represent the uninstallation rate of the m-th user within the i-th hotspot area, and d represent the energy consumption coefficient. i-1,i Let f represent the horizontal distance between the (i-1)th hotspot region and the ith hotspot region, ξ represent the CPU frequency required to compute each bit, and f represent the distance between the (i-1)th hotspot region and the ith hotspot region. u This indicates the CPU computing frequency on the drone.

4. The planning and scheduling method for UAV MEC networks according to claim 1, characterized in that, The utility of the model task is expressed by the following formula: In the formula, u m,i This represents the user task utility of the m-th user within the i-th visited hotspot area.

5. A planning and scheduling system for an unmanned aerial vehicle (UAV) MEC network, characterized in that, The system is applied to the method as described in any one of claims 1 to 4, the system comprising: The transmission model construction module is used to construct the network transmission model of the edge computing network based on the total flight latency and total energy consumption of the UAV. The edge computing network includes several hotspot areas, several users in each hotspot area, and a mobile edge computing server on the UAV. The task utility calculation module is used to obtain the model task utility based on task priority and task processing delay. The optimal planning and scheduling module is used to construct a planning function with the goal of maximizing the utility of the model task, and to make decisions on the planning function using a deep reinforcement learning algorithm to obtain the optimal user unloading strategy and the optimal UAV movement path.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images