Intelligent scheduling method for low-altitude economic activities

Abstract

The invention discloses an intelligent scheduling method for low-altitude economic activities, and relates to the technical field of unmanned aerial vehicle scheduling. The method comprises the following steps: constructing a task environment model, obtaining a task demand set, calculating an unmanned aerial vehicle set and a charging unmanned aerial vehicle set; dividing a plurality of task sub-regions according to the task region, the equipment position and the ground equipment and resource demand corresponding to the task; a calculation unmanned aerial vehicle is allocated to each task sub-region; assigning the tasks in the task sub-region to the calculation unmanned aerial vehicle through a market bidding method; and calculating a real-time scheduling strategy of the charging unmanned aerial vehicle generated through a DDPG algorithm in the task execution process of the unmanned aerial vehicle. According to the invention, clear task partitions are provided for the calculation unmanned aerial vehicle and the charging unmanned aerial vehicle by dividing the sub-regions, so that unnecessary frequent interaction and resource waste between devices are reduced; the scheduling strategy of the charging unmanned aerial vehicle is dynamically generated, so that the continuous and efficient operation of the calculation unmanned aerial vehicle is ensured in a dynamic environment.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle dispatching, and in particular to an intelligent dispatching method for low-altitude economic activities. Background Art

[0002] Traditional edge computing distributes data processing capabilities and storage resources at the edge of the network, close to the data source, to reduce the burden on remote data centers. This approach reduces transmission delays, improves data processing speed and response efficiency, and is very suitable for application scenarios that require real-time processing and low latency requirements, such as the Internet of Things and smart cities. However, fixed edge nodes are usually limited in distribution and cannot flexibly respond to dynamic and temporary activities.

[0003] In response to the above problems, applying drones to edge computing can improve the flexibility and coverage of the system. As mobile computing nodes, drones can respond quickly when mission requirements change, thereby providing computing resources and data collection functions. By carrying computing devices, drones can provide mobile edge computing services in a wide range of geographic spaces, solving the problem of areas that fixed nodes cannot cover, making the application of edge computing more flexible and extensive.

[0004] However, drone scheduling faces many challenges. Existing technologies make it difficult to accurately allocate tasks according to rapidly changing mission and environmental requirements, resulting in inefficient resource utilization. In addition, due to the limited battery life of drones, current technologies have not yet been able to provide sufficiently efficient charging strategies to ensure mission continuity, making it difficult to achieve efficient drone scheduling in dynamic and uncertain environments. Summary of the invention

[0005] In order to solve the problems existing in the above background technology, the present invention adopts the following technical solutions: An intelligent scheduling method for low-altitude economic activities, comprising the steps of: Construct a task environment model, and obtain a task requirement set, a computing drone set, and a charging drone set; the task environment model includes a task area and several equipment locations; the computing drone set includes several computing drones; the charging drone set includes several charging drones; the task requirement set includes ground equipment and resource requirements corresponding to several tasks; Divide the mission into several sub-areas according to the mission area, equipment location, and ground equipment and resource requirements corresponding to the mission; Allocate computing drones in each task sub-area; allocate tasks in the task sub-area to computing drones through market bidding; During the calculation of the UAV's mission execution, the DDPG algorithm is used to generate the real-time scheduling strategy for the charging UAV.

[0006] As a preferred solution of the present invention, the method of dividing the task sub-areas according to the task area, the equipment location, and the ground equipment and resource requirements corresponding to the task includes the following steps: S21. Obtain the total resource requirement of each ground device according to the task and its corresponding resource requirement S22, determining the number of task sub-regions, initializing the position and velocity of each particle in the particle swarm optimization algorithm; the position of the particle is used to represent the division scheme of the task sub-region; S23, calculate the fitness values ​​of all particles, and update the group's best position and each particle's personal best position; S24, updating the speed and position of each particle according to the individual best position and the group best position; S25. Repeat steps S23 to S24 until the maximum number of iterations is reached or the fitness value is less than a set threshold, and use the optimal position of the group as a division scheme for the task sub-areas.

[0007] As a preferred solution of the present invention, the total resource demand of the ground equipment is expressed as: ; in, represents the total resource demand of the hth ground equipment; represents the resource requirement of the qth task of the hth ground equipment; Represents the total number of tasks of the hth ground equipment.

[0008] As a preferred embodiment of the present invention, the position of the particles is expressed as: , in, is the number of task sub-areas; represents the cluster center of the k-th task sub-region, , and represent the longitude and latitude of the cluster center of the k-th task sub-region respectively; represents the clustering radius of the k-th task sub-region.

[0009] As a preferred embodiment of the present invention, the fitness value is expressed as: , in, represents the distance fitness function; represents the resource balance fitness function; and Respectively represent the distance fitness function and the weight parameter of the distance fitness function; represents the distance between the hth ground device and the kth cluster center; Represents the total number of tasks in the kth task sub-region.

[0010] As a preferred solution of the present invention, the tasks in the task sub-area are allocated to the computing drones by the market bidding method, specifically, the following steps are repeatedly performed until all the tasks in the task requirement set are allocated: Obtain the task with the largest unallocated resource demand in the task sub-region in the task demand set, and record it as the current task to be allocated; Calculate the quotation index based on the difference between the expected total revenue of the current task to be assigned after it is added to the task sequence and the current total revenue of the calculated drone, and generate a quotation request based on the quotation index; Each computing drone broadcasts the quotation request to other computing drones in all task sub-areas; After receiving the quotation request, each computing drone generates an evaluation result based on its own quotation index and transmits it to the dispatch center; The dispatch center assigns the current tasks to be assigned to the corresponding computing drones based on all evaluation results, and updates the task requirement set and the task sequence of the computing drones.

[0011] As a preferred solution of the present invention, the quotation index is expressed as: ; Among them, i represents the number of the calculated drone, represents the current tasks to be assigned in the kth task sub-area; represents the task sequence of the i-th computing drone; Indicates that the current task to be assigned is added to the task sequence The expected total return after Represents the current total benefit of the i-th calculated drone.

[0012] As a preferred solution of the present invention, the method for obtaining the expected total revenue after the current task to be assigned is added to the task sequence is as follows: Determine the optimal position of the current task to be assigned in the task sequence according to the device position corresponding to the current task to be assigned, and generate a reference task sequence; The expected total benefit is calculated based on the scheduling duration between adjacent tasks in the reference task sequence, the task benefit of each task, and the calculation task execution time.

[0013] As a preferred solution of the present invention, the real-time scheduling strategy for charging drones generated by the DDPG algorithm includes the following steps: Construct a Markov decision process for charging the drone; the Markov decision process consists of a state space, an action space, a transition probability, a reward function, and a discount factor; Set the optimization objective function of the charging drone implementation scheduling strategy; A real-time scheduling strategy for charging UAVs is generated using the DDPG algorithm based on the Markov decision process and the optimization objective function.

[0014] As a preferred solution of the present invention, the optimization objective function is set to minimize the task completion time of all computing drones; The reward function is expressed as: ; in, express Rewards for charging drones during time slots; express The amount of charge provided by the charging drone in the time slot; is a preset constant; represents the charging fairness indicator, , represents the number of drones calculated in the kth task sub-area; When The charging drone did not perform charging action during the time slot. When The charging drone performs charging actions within the time slot.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention can provide clear task partitions for computing drones and charging drones by building a task environment model and dividing sub-areas according to task requirements, thereby reducing unnecessary frequent interactions between devices and resource waste, and improving the overall task scheduling efficiency; by subdividing the task area into multiple sub-areas and accurately allocating them according to resource requirements, it is ensured that each task sub-area can obtain corresponding computing resources and charging services according to actual needs; during the task execution process, the DDPG algorithm is used to dynamically generate a scheduling strategy for the charging drone, so that the charging drone can adjust the flight path and charging action according to the real-time task, thereby ensuring the continuous and efficient operation of the computing drone in a dynamic environment.

[0016] The embodiment of the present invention divides a number of task sub-areas based on a particle swarm optimization algorithm according to the task area, equipment location, and ground equipment and resource requirements corresponding to the task; based on a distance fitness function, the concentration of the location of ground equipment in each task sub-area is evaluated; based on a resource balance fitness function, the equilibrium state of resource requirements in each task sub-area can be effectively evaluated and optimized; the fitness function of the particle swarm optimization algorithm is set as a weighted sum of a distance fitness function and a resource balance fitness function, and the task area division is continuously optimized by updating the speed and position of the particles, so as to comprehensively consider the geographical location and resource requirements, so that the distribution of equipment and tasks in the cluster area is more reasonable, the extra consumption of drones caused by distance problems and the drone scheduling conflicts caused by uneven resource allocation are reduced, and the overall task execution efficiency is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0018] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0019] Figure 1 A schematic diagram of a flow chart of an intelligent scheduling method for low-altitude economic activities provided by an embodiment of the present invention; Figure 2 A flow chart of dividing several task sub-areas according to the task area, equipment location, ground equipment and resource requirements corresponding to the task provided by an embodiment of the present invention; Figure 3 A schematic diagram of a process for allocating tasks in a task sub-area to a computing drone through a market bidding method provided by an embodiment of the present invention; Figure 4 A schematic diagram of a flow chart of generating a real-time scheduling strategy for charging drones using a DDPG algorithm provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0021] It should be noted that all directional indications (such as up, down, left, right, front, back, etc.) in the embodiments of the present invention are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0022] In addition, the descriptions of "first", "second", etc. in the present invention are only used for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0023] Traditional edge computing distributes data processing capabilities and storage resources at the edge of the network, close to the data source, to reduce the burden on remote data centers. This approach reduces transmission delays, improves data processing speed and response efficiency, and is very suitable for application scenarios that require real-time processing and low latency requirements, such as the Internet of Things and smart cities. However, fixed edge nodes are usually limited in distribution and cannot flexibly respond to dynamic and temporary activities.

[0024] In response to the above problems, applying drones to edge computing can improve the flexibility and coverage of the system. As mobile computing nodes, drones can respond quickly when mission requirements change, thereby providing computing resources and data collection functions. By carrying computing devices, drones can provide mobile edge computing services in a wide range of geographic spaces, solving the problem of areas that fixed nodes cannot cover, making the application of edge computing more flexible and extensive.

[0025] However, drone scheduling faces many challenges. Existing technologies make it difficult to accurately allocate tasks according to rapidly changing mission and environmental requirements, resulting in inefficient resource utilization. In addition, due to the limited battery life of drones, current technologies have not yet been able to provide sufficiently efficient charging strategies to ensure mission continuity, making it difficult to achieve efficient drone scheduling in dynamic and uncertain environments.

[0026] like Figure 1 As shown, an intelligent scheduling method for low-altitude economic activities includes the following steps: Construct a task environment model, and obtain a task requirement set, a computing drone set, and a charging drone set; the task environment model includes a task area and several equipment locations; the computing drone set includes several computing drones; the charging drone set includes several charging drones; the task requirement set includes ground equipment and resource requirements corresponding to several tasks; Divide the mission into several sub-areas according to the mission area, equipment location, and ground equipment and resource requirements corresponding to the mission; Allocate computing drones in each task sub-area; allocate tasks in the task sub-area to computing drones through market bidding; During the calculation of the UAV's mission execution, the DDPG algorithm is used to generate the real-time scheduling strategy for the charging UAV.

[0027] The present invention can provide clear task partitions for computing drones and charging drones by building a task environment model and dividing sub-areas according to task requirements, thereby reducing unnecessary frequent interactions between devices and resource waste, and improving the overall task scheduling efficiency; by subdividing the task area into multiple sub-areas and accurately allocating them according to resource requirements, it is ensured that each task sub-area can obtain corresponding computing resources and charging services according to actual needs; during the task execution process, the DDPG algorithm is used to dynamically generate a scheduling strategy for the charging drone, so that the charging drone can adjust the flight path and charging action according to the real-time task, thereby ensuring the continuous and efficient operation of the computing drone in a dynamic environment.

[0028] The following is a detailed explanation of an intelligent scheduling method for low-altitude economic activities of the present invention in conjunction with an embodiment: An intelligent scheduling method for low-altitude economic activities, comprising the steps of: S1. Construct a task environment model, obtain a task requirement set, a computing drone set, and a charging drone set; the task environment model includes a task area and several equipment locations; the computing drone set includes several computing drones; the charging drone set includes several charging drones; the task requirement set includes ground equipment and resource requirements corresponding to several tasks; The device position is represented by ; Wherein, h represents the ground equipment number; The task requirement set is expressed as ,in, Indicates the task number. represents the resource requirement of the qth task of the hth ground equipment; The computing drone set is represented as ; In one embodiment of the present invention, the charging drone and the computing drone are equipped with wireless charging components, such as magnetic resonance or inductive charging systems, which can transmit power to the computing drone through an electromagnetic field. After the charging drone approaches the computing drone, it maintains a relatively stable and appropriate distance and angle to ensure the efficiency and effect of the wireless charging process. Usually, this distance is between a few centimeters and tens of centimeters. In the implementation of the present invention, the charging drone and the computing drone during the charging process can be regarded as being in the same position.

[0029] After ensuring a relatively stable and appropriate distance and angle, the charging drone activates the wireless charging system to generate an electromagnetic field; the receiving device (such as a receiving coil) on the computing drone converts the electromagnetic field into electrical energy, thereby charging its battery.

[0030] During the charging process, the charging drone will continuously monitor the charging status, including voltage, current and distance. If it is calculated that the drone is moving or the environment has changed, the charging drone will adjust the position and charging parameters in real time to optimize charging efficiency and safety.

[0031] S2. Divide the mission sub-areas according to the mission area, equipment location, and ground equipment and resource requirements corresponding to the mission; For further information, see Figure 2 The step of dividing the task sub-areas according to the task area, equipment location, ground equipment and resource requirements corresponding to the task includes the following steps: S21. Obtain the total resource requirement of each ground device according to the task and its corresponding resource requirement; The total resource demand of the ground equipment is expressed as: ; in, represents the total resource demand of the hth ground equipment; represents the resource requirement of the qth task of the hth ground equipment; Represents the total number of tasks of the hth ground equipment.

[0032] This step first calculates the total resources required by each ground device in order to understand the needs of each ground device in the entire mission area and provide basic data for the subsequent division of mission sub-areas.

[0033] S22, determine the number of task sub-regions, initialize the position and speed of each particle in the particle swarm optimization algorithm; the position of the particle is used to represent the division scheme of the task sub-region; the position of the particle is expressed as: , in, is the number of task sub-areas; represents the cluster center of the k-th task sub-region, , and represent the longitude and latitude of the cluster center of the k-th task sub-region respectively; represents the clustering radius of the k-th task sub-region.

[0034] In this step, the number of task sub-areas can be reasonably determined based on factors such as the size of the task area, the number of charging drones, the number and distribution of ground equipment, task frequency, etc. In particular, in one embodiment of the present invention, the number of task sub-areas is set to the number of charging drones, that is, each task sub-area is assigned a charging drone to provide charging services for the computing drone therein.

[0035] The position of each particle represents a task sub-region division scheme, including the coordinates (longitude, latitude) and cluster radius of the cluster center of each task sub-region. The speed of the particle is a small value generated randomly.

[0036] S23, calculate the fitness values ​​of all particles, and update the group's best position and each particle's personal best position; The fitness value is expressed as: , in, represents the distance fitness function; represents the resource balance fitness function; and Respectively represent the distance fitness function and the weight parameter of the distance fitness function; represents the distance between the hth ground device and the kth cluster center; represents the total number of tasks in the kth task sub-area; The fitness value consists of two parts: distance fitness function and resource balance fitness function. In one embodiment, the optimization goal of the particle is to minimize the fitness value. , In the distance fitness function, by constraining The distances between all ground devices and the cluster center in each cluster area are summed to evaluate the clustering results by geographical location. The calculation formula of the resource balance fitness function is based on Jain's Fairness Index equation, and its value range is , The closer it is to 1, the more balanced the resource demands of each cluster area are, that is, the difference in the sum of the resource demands of the ground equipment in each cluster area is smaller.

[0037] S24, updating the speed and position of each particle according to the individual best position and the group best position; In the particle swarm optimization algorithm, the speed and position of each particle are updated according to the individual optimal position and the group optimal position.

[0038] First, the particle speed update takes three factors into consideration. The current speed of the particle is attenuated by an inertia weight to maintain a certain inertia; then, the particle adjusts its speed according to its best position and moves towards this personal best position, while combining a random factor to increase the diversity of exploration; finally, the particle will also move closer to the best position found by the group to understand the collective optimal solution. This speed adjustment ensures that the particle can comprehensively consider the advantages of individuals and groups to improve search efficiency.

[0039] The particle position is updated by adding the updated velocity to the previous position. This means that the particle will move to a new position under the influence of the new velocity, thereby exploring the possible optimal solution in the solution space.

[0040] S25. Repeat steps S23 to S24 until the maximum number of iterations is reached or the fitness value is less than a set threshold, and use the optimal position of the group as a division scheme for the task sub-areas.

[0041] By repeatedly executing steps S22 to S23 to update the speed and position until the number of iterations reaches a preset value or the fitness value is less than a set threshold, the particle group gradually converges to the optimal solution of the objective function in this process. Finally, the optimal position of the group is used as the final task sub-area division scheme.

[0042] This embodiment divides several task sub-areas based on the particle swarm optimization algorithm according to the task area, equipment location, and ground equipment and resource requirements corresponding to the task; evaluates the concentration of the location of ground equipment in each task sub-area based on the distance fitness function; and can effectively evaluate and optimize the equilibrium state of resource requirements in each task sub-area based on the resource balance fitness function; sets the fitness function of the particle swarm optimization algorithm as the weighted sum of the distance fitness function and the resource balance fitness function, and continuously optimizes the task area division by updating the speed and position of the particles, so as to comprehensively consider the geographical location and resource requirements, make the distribution of equipment and tasks in the cluster area more reasonable, reduce the extra consumption of drones caused by distance problems and the drone scheduling conflicts caused by uneven resource allocation, and improve the overall task execution efficiency.

[0043] S3, allocating computing drones to each task sub-area; allocating tasks within the task sub-area to computing drones through a market bidding method; After completing the division of task sub-areas, the device location and total resource demand in each task sub-area are planned to allocate computing resources, namely computing drones, to the task sub-areas. In this solution, computing drones act as edge computing nodes. Their advantage is that by sinking computing power to the edge of the network, close to the source of data, real-time data processing and analysis can be achieved, thereby reducing latency and improving response time. As mobile edge nodes, computing drones are equipped with computing and storage capabilities and can provide computing services within the task sub-areas.

[0044] During the implementation process, the computing drone will move to a suitable location according to the division of the task sub-area and resource requirements to ensure that sufficient computing resources are provided in the area that needs to be calculated. According to the resource requirements corresponding to the tasks in the task sub-area and the equipment location of the ground equipment corresponding to each task, the benefit evaluation index of the execution of the task can be generated, and the task can be assigned to the appropriate computing drone according to the benefit evaluation index; after the task is assigned, the computing drone will use its own computing resources to directly process data on site and perform computing tasks.

[0045] In one embodiment, see Figure 3 , the tasks in the task sub-area are allocated to the computing drones by the market bidding method, specifically: the following steps are repeated until all tasks in the task requirement set are allocated: S31, obtaining the task with the largest unallocated resource demand in the task sub-region in the task demand set, and recording it as the current task to be allocated; The unassigned task with the largest resource demand in the task sub-region is obtained from the task demand set and recorded as the current task to be assigned. In this way, tasks with larger resource demand are prioritized, thereby improving resource utilization with lower computational complexity.

[0046] S32, calculating a quotation index according to the difference between the expected total revenue of the current task to be assigned after being added to the task sequence and the current total revenue of the calculated drone, and generating a quotation request according to the quotation index; The quotation index is expressed as:

[0047] Among them, i represents the number of the calculated drone, represents the current tasks to be assigned in the kth task sub-area; represents the task sequence of the i-th computing drone; Indicates that the current task to be assigned is added to the task sequence The expected total return after Represents the current total benefit of the i-th calculated drone.

[0048] S33, each computing drone broadcasts the quotation request to other computing drones in all task sub-areas; In this step, each computing drone will broadcast the generated quotation request to all other computing drones in the task sub-area, so that other drones can synchronize their information and participate in the subsequent bidding evaluation to increase transparency and competitiveness, ensure that each task can be undertaken by the most suitable drone, and thus optimize the overall task allocation results.

[0049] S34, after receiving the quotation request, each computing drone generates an evaluation result according to its own quotation index and transmits it to the dispatch center; After receiving the quotation request, each computing drone generates an evaluation result based on its own quotation index, and then transmits the evaluation result to the dispatch center. The evaluation results of each drone are combined to ensure that each task can be reasonably assigned to the most suitable drone without affecting its own operating efficiency.

[0050] S35. The dispatch center assigns the current tasks to be assigned to the corresponding computing drones according to all evaluation results, and updates the task requirement set and the task sequence of the computing drones.

[0051] After receiving all the evaluation results, the dispatch center selects the computing drone with the best evaluation result to assign the current task to be assigned, and updates the task requirement set and the task sequence of the computing drone at the same time.

[0052] This embodiment ensures that tasks can be efficiently and reasonably allocated to computing drones through a market auction method, so as to improve resource utilization, enhance the fairness of task allocation, and improve the overall operating efficiency of the system, thereby achieving optimal configuration and execution of tasks within task sub-areas.

[0053] In one embodiment, the method for obtaining the expected total revenue after the current task to be assigned is added to the task sequence is as follows: S321, determining the optimal position of the current task to be assigned in the task sequence according to the device position corresponding to the current task to be assigned, and generating a reference task sequence; The optimal position of the current task to be assigned in the task sequence is determined based on the device position corresponding to the current task to be assigned, with the goal of finding an optimal position in the task sequence to insert the current task to be assigned so that the execution efficiency of the entire task sequence is optimal (for example, minimizing the total moving time). The optimal position can be obtained by calculating the distance between the device position corresponding to the current task to be assigned and the device positions corresponding to the tasks in all task sequences; and inserting the current task to be assigned into the task corresponding to the device position with the smallest distance.

[0054] S322, calculating the expected total benefit according to the scheduling duration between adjacent tasks in the reference task sequence, the task benefit of each task and the calculation task execution time.

[0055] After generating the reference task sequence, the scheduling duration of each pair of adjacent tasks in the reference task sequence, the task benefit of each task, and the calculation task execution time are obtained.

[0056] Among them, the scheduling duration is the movement duration caused by the different device locations corresponding to the task, which can be calculated based on the distance between the device locations and the mobility performance of the drone.

[0057] In the edge computing environment, task benefit is a measure that represents the benefit obtained by the system after the task is executed. This benefit can be in various forms. In one embodiment, the evaluation factors of task benefit include response time, resource utilization, user satisfaction and quality of task execution. Specifically, for response time, edge computing deploys computing resources close to the data source, such as user devices, IoT devices, etc. This can greatly shorten the delay of data transmission, thereby improving the response time of the system; for resource utilization, edge computing can make full use of the computing resources of edge devices instead of sending all tasks to remote servers for processing, thereby improving the resource utilization of the entire system; for user satisfaction, due to fast response time and high reliability, user experience is significantly improved; for the quality of task execution, edge computing can efficiently execute tasks while improving the quality and reliability of task completion under low latency and energy consumption. When generating task information, the task benefit is calculated by quantitatively evaluating the influencing factors, namely response time, resource utilization, user satisfaction and quality of task execution, and stored in association with the resource demand of the task, and is directly used to measure and calculate the benefits of edge computing tasks when this step is implemented.

[0058] S4. During the calculation of the UAV’s mission execution, the real-time scheduling strategy of the charging UAV is generated through the DDPG algorithm.

[0059] The DDPG (Deep Deterministic Policy Gradient) algorithm is a deep reinforcement learning algorithm suitable for continuous action space. It combines the advantages of policy gradient and Q learning and is suitable for real-time decision-making in dynamic environments.

[0060] For further information, see Figure 4 The real-time scheduling strategy for charging drones generated by the DDPG algorithm includes the following steps: S41. Construct a Markov decision process for charging the UAV; the Markov decision process consists of a state space, an action space, a transition probability, a reward function, and a discount factor; Markov Decision Process (MDP) is a mathematical framework for modeling decision problems, especially decision-making processes in a random environment. MDP is widely used in fields such as reinforcement learning, control theory, and decision analysis.

[0061] In the present invention, the state space is composed of states corresponding to several time slots; the states include the position, power and task progress of the calculation drone, and the position, power reserve, charging state and movement state of the charging drone; The action space is composed of actions corresponding to the charging drone in several time slots; the actions include moving direction, charging action and charging target; The transition probability is used to represent the probability that the current state of the charging drone transfers to the next possible state after executing an action in the action set; The reward function is used to represent the reward obtained by the charging drone after the current state is transferred to the next state after the action is performed; The discount factor is used to determine the impact function of future rewards on current rewards.

[0062] Specifically, the reward function is expressed as: ; in, Used to indicate Rewards for charging drones during time slots; express The amount of charge provided by the charging drone in the time slot; It is a preset constant used to encourage charging drones to perform more charging actions instead of performing meaningless movements or hovering; represents the charging fairness indicator, , represents the number of computing drones in the kth task sub-area. The charging fairness index is used to encourage the charging drones to charge each computing drone fairly. In one embodiment, it is based on Jain's Fairness Index equation; When The charging drone did not perform charging action during the time slot. When The charging drone performs charging actions within the time slot.

[0063] In one embodiment, the number of task sub-areas is set to the number of charging drones, that is, each task sub-area is assigned a charging drone to provide charging services for the computing drone therein. By allowing each charging drone to focus on one sub-area, the complexity of the state and action space that the DDPG algorithm needs to process is reduced, so that the learning efficiency and convergence speed of the algorithm are improved. The algorithm only needs to optimize the strategy for the state of the current sub-area to avoid dealing with the complex deduction of the global state. Each charging drone performs tasks in a fixed sub-area, and can collect empirical data more stably. These data play an effective role in continuously strengthening the same strategy, reducing the frequency of strategy adjustments caused by environmental changes, thereby improving the stability and convergence of the algorithm. Under the premise of independent optimization of each sub-area, the charging strategy of each drone tends to the local optimal solution. This ensures that the computing drones in each area can obtain charging services in a timely manner. At the same time, these local optimal solutions jointly promote the improvement of global performance, so that the overall task completion time is minimized.

[0064] S42, setting an optimization objective function for the implementation scheduling strategy of the charging drone; The optimization objective function is set to minimize the task completion time of all computing drones, expressed as: ; in, represents the task completion time of the i-th computing drone, represents the number of drones calculated in the kth task sub-area; Represents the numbered set of computing drones in the k-th task sub-area.

[0065] S43. Generate a real-time scheduling strategy for charging drones using a DDPG algorithm based on the Markov decision process and the optimization objective function.

[0066] Further, step S43 includes: S431, initialize DDPG algorithm parameters: initialize the parameters of the policy network and Q network, and generate the corresponding target network; initialize the experience replay pool to store the state transition of the charging drone after performing actions in the environment.

[0067] S432, calculating the reward function: by evaluating the effect of the charging drone action in real time, calculating the immediate reward.

[0068] The instant reward function consists of four parts: the amount of charge provided by the charging drone in the time slot, the charging fairness index, the task completion time of the drone calculated in the task area, and the penalty / reward coefficient for whether to perform the charging action.

[0069] S433, strategy network generates actions: Use the strategy network to input the current state and output the action of the charging drone in the current time slot, including the moving direction and charging target. In the exploration phase, appropriate noise is added to increase the exploratory nature of the strategy.

[0070] S434, execute actions and observe results: adjust the behavior of the charging drone according to the actions output by the policy network, and execute actions in the environment; observe the new state after executing the action, and store the state transition (current state, action, reward and next state) in the experience replay pool.

[0071] S436, update strategy and value network: randomly extract samples (state transition data) from the experience replay pool to update the strategy and value network; calculate the target Q value, and update the Q network by minimizing the error of the Q value.

[0072] The policy network is optimized by policy gradient ascent so that the actions output by the policy network can maximize the output of the Q network.

[0073] This embodiment can continuously optimize the scheduling strategy of charging drones through the DDPG algorithm, so that it can provide charging services for computing drones more efficiently in complex dynamic environments, thereby minimizing the overall task completion time. The generation process of this real-time scheduling strategy not only takes into account the current status and task requirements of each drone, but also effectively balances charging fairness and mobility, ensuring the achievement of the global optimization goal of the charging process.

[0074] In the several embodiments provided in the present application, it should be understood that the disclosed system can be implemented in other ways. For example, the system embodiments described above are only schematic. For example, the division of the modules is only a logical function division. There may be other division methods in actual implementation, such as multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the modules can be electrical, mechanical or other forms.

[0075] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0076] In addition, each functional module in each embodiment of the present application can be integrated into a processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The above integrated modules can be implemented in the form of hardware or software functional modules.

[0077] If the integrated module is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), disk or optical disk and other media that can store program code.

[0078] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. An intelligent dispatching method for low-altitude economic activities, characterized by: Includes steps: Construct a task environment model, and obtain a task requirement set, a computing drone set, and a charging drone set; the task environment model includes a task area and several equipment locations; the computing drone set includes several computing drones; the charging drone set includes several charging drones; the task requirement set includes ground equipment and resource requirements corresponding to several tasks; Divide the mission into several sub-areas according to the mission area, equipment location, and ground equipment and resource requirements corresponding to the mission; Allocate computing drones in each task sub-area; allocate tasks in the task sub-area to computing drones through market bidding; During the calculation of the UAV's mission execution, the DDPG algorithm is used to generate the real-time scheduling strategy for the charging UAV.

2. The intelligent dispatching method for low-altitude economic activities according to claim 1 is characterized by: The method of dividing the task sub-areas according to the task area, the equipment location, and the ground equipment and resource requirements corresponding to the task includes the following steps: S21. Obtain the total resource requirement of each ground device according to the task and its corresponding resource requirement; S22, determining the number of task sub-regions, initializing the position and velocity of each particle in the particle swarm optimization algorithm; the position of the particle is used to represent the division scheme of the task sub-region; S23, calculate the fitness values ​​of all particles, update the group's best position and each particle's personal best position; S24, updating the speed and position of each particle according to the individual best position and the group best position; S25. Repeat steps S23 to S24 until the maximum number of iterations is reached or the fitness value is less than a set threshold, and use the optimal position of the group as a division scheme for the task sub-areas.

3. The intelligent dispatching method for low-altitude economic activities according to claim 2 is characterized by: The total resource demand of the ground equipment is expressed as: ， in, represents the total resource demand of the hth ground equipment; represents the resource requirement of the qth task of the hth ground equipment; Represents the total number of tasks of the hth ground equipment.

4. The intelligent dispatching method for low-altitude economic activities according to claim 3 is characterized by: The position of the particle is expressed as: , in, is the number of task sub-areas; represents the cluster center of the k-th task sub-region, , and represent the longitude and latitude of the cluster center of the k-th task sub-region respectively; represents the clustering radius of the k-th task sub-region.

5. The intelligent dispatching method for low-altitude economic activities according to claim 4 is characterized by: The fitness value is expressed as: , in, represents the distance fitness function; represents the resource balance fitness function; and Respectively represent the distance fitness function and the weight parameter of the distance fitness function; represents the distance between the hth ground device and the kth cluster center; Represents the total number of tasks in the kth task sub-region.

6. The intelligent dispatching method for low-altitude economic activities according to claim 1 is characterized by: The method of allocating tasks in the task sub-area to computing drones by market bidding is as follows: Repeat the following steps until all tasks in the task requirement set are allocated: Obtain the task with the largest unallocated resource demand in the task sub-region in the task demand set, and record it as the current task to be allocated; Calculate the quotation index based on the difference between the expected total revenue of the current task to be assigned after it is added to the task sequence and the current total revenue of the calculated drone, and generate a quotation request based on the quotation index; Each computing drone broadcasts the quotation request to other computing drones in all task sub-areas; After receiving the quotation request, each computing drone generates an evaluation result based on its own quotation index and transmits it to the dispatch center; The dispatch center assigns the current tasks to be assigned to the corresponding computing drones based on all evaluation results, and updates the task requirement set and the task sequence of the computing drones.

7. The intelligent dispatching method for low-altitude economic activities according to claim 6 is characterized by: The quotation index is expressed as: ， Among them, i represents the number of the calculated drone, represents the current tasks to be assigned in the kth task sub-area; represents the task sequence of the i-th computing drone; Indicates that the current task to be assigned is added to the task sequence The expected total return after Represents the current total benefit of the i-th calculated drone.

8. The intelligent dispatching method for low-altitude economic activities according to claim 6 is characterized by: The method for obtaining the expected total benefit after the current task to be assigned is added to the task sequence is as follows: Determine the optimal position of the current task to be assigned in the task sequence according to the device position corresponding to the current task to be assigned, and generate a reference task sequence; The expected total benefit is calculated based on the scheduling duration between adjacent tasks in the reference task sequence, the task benefit of each task, and the calculation task execution time.

9. The intelligent dispatching method for low-altitude economic activities according to claim 1 is characterized by: The real-time scheduling strategy for charging drones generated by the DDPG algorithm includes the following steps: Construct a Markov decision process for charging the drone; the Markov decision process consists of a state space, an action space, a transition probability, a reward function, and a discount factor; Set the optimization objective function of the charging drone implementation scheduling strategy; A real-time scheduling strategy for charging UAVs is generated using the DDPG algorithm based on the Markov decision process and the optimization objective function.

10. The intelligent dispatching method for low-altitude economic activities according to claim 9 is characterized by: The optimization objective function is set to minimize the task completion time of all computing drones; The reward function is expressed as: ， in, express Rewards for charging drones during time slots; express The amount of charge provided by the charging drone in the time slot; is a preset constant; represents the charging fairness indicator, , represents the number of drones calculated in the kth task sub-area; When The charging drone did not perform charging action during the time slot. When The charging drone performs charging actions within the time slot.

Images