Multi-unmanned aerial vehicle cooperative data collection method, readable storage medium and equipment

Abstract

The invention discloses a multi-unmanned aerial vehicle cooperative data collection method, a readable storage medium and equipment, and belongs to the technical field of wireless communication. Based on the environment model, determining a local network model taking maximization of the total data collection amount of the multiple unmanned aerial vehicles and the shortest data collection path length as targets, training network parameters of the local network model by using an improved QMIX algorithm, and periodically performing aggregation processing on the network parameters to obtain a trained unmanned aerial vehicle global model; the optimal path length of the maximum data collection amount of the multiple unmanned aerial vehicles is obtained through the unmanned aerial vehicle global model according to the position information of the target urban area Internet of Things equipment and the data transmission channel feature data, and the multiple unmanned aerial vehicles are controlled to carry out data collection on the target Internet of Things equipment. The robustness of the system is improved while the data collection efficiency is improved and the flight path length is reduced.

Description

Technical Field

[0001] The present invention relates to a multi-UAV collaborative data collection method, a readable storage medium and a device, and belongs to the technical field of wireless communications. Background Art

[0002] With the rapid development of science and technology and the continuous advancement of drone technology, the application of multi-UAV systems in various fields is increasing. The flexibility, high efficiency and low cost of drones make them show significant advantages in data collection, monitoring and analysis. Especially in the fields of environmental monitoring, agricultural management, disaster assessment, urban planning and military reconnaissance, drones have become an indispensable tool. In terms of data collection, multi-UAV systems can significantly improve the efficiency and accuracy of data collection by working together, expand coverage, and reduce time and labor costs. However, trajectory planning in a multi-UAV system is a challenging task, especially in dynamic and complex environments, and it is necessary to consider many factors such as energy consumption, task allocation, environmental perception and coordination between drones. Summary of the invention

[0003] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a multi-UAV collaborative data collection method, a readable storage medium and a device, which improves data collection efficiency, reduces flight path length and improves system robustness, thereby solving the current problem of difficulty in collaborative data collection by multiple UAVs in dynamic and complex environments.

[0004] In order to achieve the above purpose / solve the above technical problems, the present invention is implemented by adopting the following technical solutions: The present invention provides a method for collaborative data collection of multiple unmanned aerial vehicles, comprising: Obtain location information of IoT devices and data transmission channel characteristic data in the target city area; Input the IoT device location information and the channel feature data into a pre-trained drone global model to obtain an optimal path with the shortest path length for multiple drones to collect the maximum amount of data; According to the optimal path, control multiple drones to collect data from target IoT devices; Among them, the training of the pre-trained drone global model includes: Establish an environmental model for data collection based on the number of drones and the location distribution of IoT devices; Based on the environment model, a local network model is determined with the goal of maximizing the total data collection amount of multiple UAVs and the shortest data collection path length; The network parameters of the local network model are trained and the network parameters are aggregated periodically to obtain a trained UAV global model.

[0005] Furthermore, the environmental model for data collection is established according to the number of drones and the location distribution of IoT devices, including: The environmental model includes a multi-UAV network model, a UAV channel model and an energy consumption model; The multi-UAV network model includes: in: i Represents a collection of drones The drone number in , I represents the total number of drones; Indicates i A drone in t+ 1 moment position, Indicates i A drone in t The location at the moment, , , represents the total duration of the drone data collection mission, Indicates i Drones at the moment t of x coordinate, Indicates i A drone at a time t of y coordinate, Indicates i The height of the drone; The position of the UAV at the beginning of the data collection mission is denoted as , the position of the UAV at the end of the data collection mission is expressed as ; Indicates at time t Drones i Select the action that changes its position a ,action a Including movement and hovering, the drone can only choose to land at the last moment; the position of the drone at the end of the data collection mission is expressed as ; Indicates at time t Drones i Select the action that changes its position a ,action a Including moving and hovering, the drone can only choose to land at the last moment; The drone channel model includes: in: Indicated in t Moment, i The drone and kchannels between IoT devices; k Represents a collection of IoT devices The IoT device number, K Indicates the number of IoT devices. The IoT device set includes devices with known location information and devices with unknown location information. Indicated in t Moment i The drone and k The distance between IoT devices, , Indicates k The location of IoT devices, , Indicates k IoT devices x coordinate, Indicates k IoT devices y coordinate; represents the shadow component with a Gaussian random variable, Indicates the reference distance d =1m logarithmic mean channel gain, represents the path loss constant, z Indicates that the channel is affected by line-of-sight (LOS) or non-line-of-sight (NLOS) conditions. ; Indicated in t Moment i The drone and k The actual channel throughput between IoT devices; Indicated in t Moment k The amount of data stored in each IoT device is , Indicates that at the beginning of the data collection task k The amount of data stored by each IoT device; Indicated in t Moment i The drone and k The actual channel throughput between IoT devices; Indicated in t Moment k The amount of data stored in each IoT device is , Indicates that at the beginning of the data collection task k The amount of data stored by each IoT device; Indicates the total duration of the drone data collection mission divided into consecutive time intervals; Energy consumption models include: in: Indicates i A drone in t The remaining energy at the moment; Indicates i A drone in t- The remaining energy at 1 moment; represents the energy consumed by the drone calculation, , Indicates the drone calculation power; represents the energy consumed in data transmission, P k ; Indicates the energy consumed by the drone during flight. , Indicates the flight power, V Indicates the speed of the drone; , represents the blade profile power of the UAV, Indicates the induced power of the drone in hovering. represents the average rotor induced speed of the drone when it is hovering, represents the tip speed of the UAV rotor blade, represents the drag ratio of the drone body, represents the air density, Indicates the solidity of the rotor, Represents the rotor disk area.

[0006] Furthermore, based on the environment model, determining a local network model with the goal of maximizing the total data collection amount of multiple UAVs and the shortest data collection path length includes: Based on the environmental model, the optimization objective is constructed with the goal of maximizing the total data collection volume of multiple UAVs and the shortest data collection path length; Based on Markov decision making, the optimization objective is mapped to the observable virtual environment space; The neural network model and particle swarm optimization algorithm are used to learn the UAV channel knowledge, determine the location information of unknown IoT devices in the virtual environment space, and obtain the local network model.

[0007] Furthermore, based on the environment model, the optimization objective is constructed with the goal of maximizing the total data collection amount of multiple UAVs and the shortest data collection path length, where the optimization objective is expressed as:

[0008] in: i Represents a collection of drones The drone number in , I represents the total number of drones; Indicates i A drone in t+ 1 moment position, Indicates i A drone in t The location at the moment, , , represents the total duration of the drone data collection mission, Indicates i A drone at a time t of x coordinate, Indicates i A drone at a time t of y coordinate, Indicates i The height of the UAV; the position of the UAV at the beginning of the data collection mission is expressed as , the position of the UAV at the end of the data collection mission is expressed as ; Indicates at time t Drones i Select the action that changes its position a ,action a Including moving and hovering, the drone can only choose to land at the last moment; Indicates i The total energy of each drone; Indicates i A drone in t The remaining energy at the moment, ; Indicates i A drone in t- The remaining energy at 1 moment; It represents the energy consumed by the drone calculation; Indicates the energy consumed by data transmission; Indicates the energy consumed by the drone during flight; It means no one is there at the moment t joint action; , indicating the status of whether the drone collects data from IoT devices; Indicates j The height of the drone; Furthermore, the Markov decision-making method maps the optimization target to an observable virtual environment space, including: Convert the optimization objective into a tuple: ;in: Represents the state space of the virtual environment, which is composed of the drone state s composition; Indicates the actions of all drones Combined joint action space; Represents the state of the environment, which is based on the state transition function: Make the conversion; Indicates the next drone state; represents the reward function, , Represents the reward after taking an action in the current state; represents the UAV joint observation space; represents the discount factor; Represents the observation probability function; the drone is based on the observation probability function O ( , , )= Determine the local observation value.

[0009] Furthermore, the neural network model and particle swarm optimization algorithm are used to learn the UAV channel knowledge and determine the location information of unknown IoT devices in the virtual environment space, including: The channel gain between the drone and the unknown IoT device is determined by the following formula: in: It means that the drone measures the channel gain between itself and the unknown IoT device. represents the function to be sought, represents the channel parameters, Indicated in t Moment i The drone and k The distance between IoT devices, represents the elevation angle, Indicates that the channel is affected by line-of-sight (LOS) or non-line-of-sight (NLOS) conditions. , represents the shadow component with a Gaussian random variable; Using IoT devices with known location information to learn channel characteristics, using parameters equivalent to channel gain The neural network determines the channel model of unknown IoT devices; Based on the channel model of unknown IoT devices, the particle swarm optimization algorithm is used to index the best particle position, and the best particle position is used as the location information of the unknown IoT devices.

[0010] Furthermore, the network parameters of the training local network model are aggregated periodically to obtain a trained UAV global model, including: According to the individual action of each UAV in the virtual environment space, the joint action of multiple UAVs in the virtual environment space is obtained; Construct a reward function with the goal of maximizing the total amount of data collected and minimizing the total path length, and evaluate the flight strategy of multiple drones to reach the mission endpoint each time with joint actions; Based on the flight strategy, the maximum entropy QMIX algorithm is used to integrate the strategy entropy into the optimization target; Based on the reward function, train the network parameters of the local network model; The model parameters of the local network model during the training process are periodically sent to the aggregator for aggregation processing to obtain the trained UAV global model.

[0011] Furthermore, the flight strategy-based QMIX algorithm using maximum entropy is used to integrate the strategy entropy into the optimization target, including: calculating the strategy entropy by the following formula: in: i Represents a collection of drones The drone number in , I represents the total number of drones; Represents the weight coefficient of strategy entropy, which is used to adjust the influence of different agents in strategy entropy; Indicates drone i In observation Next select action The probability of strategy; Calculate the state by the following formula s In the future, drones will adopt strategies The expected value of the cumulative rewards that can be obtained: in: Indicates in status s Take action when the drone is down a The expected reward in the future, , Indicates immediate reward, represents the discount factor, Indicates the expected state value for the next state; Based on the strategy entropy and the expected value of the accumulated reward, the strategy optimization goal is determined: in: , Indicates the total duration of the drone data collection mission; Representation strategy The objective function is represents the discount factor at timet The power of Indicates at time t The state of Indicates at time t The action of time, Indicates i Drone strategy.

[0012] On the other hand, the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the multi-UAV collaborative data collection method described above is implemented.

[0013] The present invention also provides a computer device, comprising: A memory for storing instructions; The processor is used to execute the instructions so that the device executes the above-mentioned multi-UAV collaborative data collection method.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention establishes an environmental model for data collection according to the number of multiple drones and the location distribution of Internet of Things devices, and determines a local network model with the goal of maximizing the total data collection amount of multiple drones and the shortest data collection path length based on the environmental model. The network parameters of the local network model are trained using an improved QMIX algorithm and the network parameters are periodically aggregated to obtain a trained drone global model. The drone global model is used to obtain the optimal path length for the maximum data collection amount of multiple drones according to the location information of Internet of Things devices in a target urban area and characteristic data of data transmission channels, and multiple drones are controlled to collect data from target Internet of Things devices. This improves the data collection efficiency and reduces the flight path length while improving the robustness of the system, and solves the problem of difficulty in collaborative data collection by multiple drones in dynamic and complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a flowchart of a method for collaborative data collection of multiple unmanned aerial vehicles, a readable storage medium, and an embodiment of the present invention; Figure 2 It is a training flow chart of a pre-trained drone global model of an embodiment of a multi-drone collaborative data collection method, a readable storage medium, and a device of the present invention; Figure 3 1. It is a schematic diagram comparing the data collection rate of a multi-UAV collaborative data collection method of the present invention and the existing method; Figure 4 It is a schematic diagram comparing the data collection path lengths of a multi-UAV collaborative data collection method of the present invention and the existing method. DETAILED DESCRIPTION

[0016] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention. Example

[0017] like Figure 1 As shown, a multi-UAV collaborative data collection method includes: Obtain location information of IoT devices and data transmission channel characteristic data in the target city area; The location information and channel characteristic data of IoT devices are input into the pre-trained drone global model to obtain the optimal path with the shortest path length for multiple drones to collect the maximum amount of data; like Figure 2 As shown, the training of the pre-trained drone global model includes: Establish an environmental model for data collection based on the number of drones and the location distribution of IoT devices. Specifically, the environmental model includes a multi-drone network model, a drone channel model, and an energy consumption model. Multi-UAV Network Model:

[0018] in: i Represents a collection of drones The drone number in , I represents the total number of drones; Indicates i A drone in t +1 position at the moment, Indicates i A drone in t The location at the moment, , , represents the total duration of the drone data collection mission, Indicates i A drone at a time t of x coordinate, Indicates i A drone at a time t of y coordinate, Indicates i The height of the UAV; the position of the UAV at the beginning of the data collection mission is expressed as , the position of the UAV at the end of the data collection mission is expressed as ; Indicates at time t Drones i Select the action that changes its position a ,actiona Including moving and hovering, the drone can only choose to land at the last moment; Drone channel model:

[0019] in: Indicated in t Moment, i The drone and k channels between IoT devices; k Represents a collection of IoT devices The IoT device number, K Indicates the number of IoT devices. The IoT device set includes devices with known location information and devices with unknown location information. , Indicates a device whose location information is known. Indicates a device whose location information is unknown; Indicated in t Moment i The drone and k The distance between IoT devices, , Indicates k The location of IoT devices, , Indicates k IoT devices x coordinate, Indicates k IoT devices y coordinate; represents the shadow component with a Gaussian random variable, Indicates the reference distance d =1m logarithmic mean channel gain, represents the path loss constant, z Indicates that the channel is affected by line-of-sight (LOS) or non-line-of-sight (NLOS) conditions. ; Indicated in t Moment i The drone and k The actual channel throughput between IoT devices; Indicated in t Moment k The amount of data stored in each IoT device is , Indicates that at the beginning of the data collection task k The amount of data stored by each IoT device; Indicates i The drone and kThe channel transmission rate between IoT devices, , B represents the channel bandwidth, represents the signal-to-noise ratio, , P k Indicates the signal transmission power of IoT devices. represents Gaussian white noise; Indicates the total duration of the drone data collection mission divided into consecutive time intervals; Energy consumption model:

[0020] in: Indicates i A drone in t The remaining energy at the moment; Indicates i A drone in t- The remaining energy at 1 moment; represents the energy consumed by the drone calculation, , Indicates the drone calculation power; represents the energy consumed in data transmission, ; Indicates the energy consumed by the drone during flight. , Indicates the flight power, , represents the blade profile power of the UAV, Indicates the induced power of the drone in hovering. represents the average rotor induced speed of the drone when it is hovering, represents the tip speed of the UAV rotor blade, represents the drag ratio of the drone body, represents the air density, Indicates the solidity of the rotor, Represents the rotor disk area.

[0021] Based on the environment model, determine the local network model with the goal of maximizing the total data collection volume of multiple drones and the shortest data collection path length, including: Based on the environmental model, the optimization objectives are constructed with the goal of maximizing the total data collection volume of multiple drones and the shortest data collection path length. Specifically:

[0022] in: Indicates i The total energy of the drones, ; It means no one is there at the moment t joint action; , indicating the status of whether the drone collects data from IoT devices; Indicates j The height of the drone; Based on Markov decision making, the optimization objective is mapped to the observable virtual environment space. Specifically: Convert the optimization objective into a tuple: ; in: Represents the state space of the virtual environment, which is composed of the drone state s composition, s= (s a ,s u ) ; s a Indicates drone information. , done Indicates whether the drone has completed the mission. done is a binary variable, ; s u Indicates IoT device information. ; The safety controller determines the minimum amount of power required for the drone to reach the destination. A safety controller is set in the drone's action space to ensure that the drone has the remaining power. Less than or equal to the minimum power required to reach the destination When ; Indicates the actions of all drones Combined joint action space; Represents the state of the environment, which is based on the state transition function: Make the conversion; Indicates the next drone state; represents the reward function, , Represents the reward after taking an action in the current state; represents the UAV joint observation space; represents the discount factor; Represents the observation probability function; the drone is based on the observation probability function O ( , , )= Determine local observations; The neural network model and particle swarm optimization algorithm are used to learn the UAV channel knowledge and determine the location information of unknown IoT devices in the virtual environment space. Specifically: In each actual deployment process, in addition to data collection, the UAV will also measure the channel gain between itself and the device. The channel gain between the UAV and the unknown IoT device is determined by the following formula:

[0023] in: Indicates drone i exist t Constantly measure the channel gain between itself and unknown IoT devices. represents the function to be sought, represents the channel parameters, Indicated in t Moment i The drone and k The distance between IoT devices, represents the elevation angle, , Indicates that the channel is affected by line-of-sight (LOS) or non-line-of-sight (NLOS) conditions. , represents the shadow component with a Gaussian random variable; Drones need to learn to determine channel parameters and function , the probability distribution of a single measurement value of a UAV can be expressed as , represents the normal distribution function, represents Gaussian white noise under line-of-sight conditions or non-line-of-sight conditions, and the measured negative log-likelihood function is:

[0024] The optimization problem It is decomposed into two sub-problems: channel learning and location device location estimation. Since the channel is only affected by urban structure and environmental obstacles, the channel characteristics are learned using the known device location. The parameters used are A neural network to approximate the function , the input vector of the neural network , output the estimated channel gain ; Based on the channel model of unknown IoT devices, the particle swarm optimization algorithm is used to index the best particle position, and the best particle position is used as the location information of the unknown IoT device. Specifically: Define the particle as , , Represents the set of real numbers ，M Indicates the number of unknown IoT devices; The negative log-likelihood function can be rewritten as:

[0025]

[0026] in: Indicates that the particle swarm algorithm i The first drone iteration m particle, represents the position of the assumed node 1 and the particle m A set of time sets at the same time, if passed If there is no obstacle in the straight line between the drone and the measured value Considered LOS; The optimal particle is obtained by minimizing the objective Get, where: represents the index of the best particle in minimizing the objective function; The best particle position of the last iteration is used as the estimate of the device node position; According to the location information of unknown IoT devices, a local network model is obtained.

[0027] The improved QMIX algorithm is used to train the network parameters of the local network model and periodically aggregate the network parameters to obtain a trained UAV global model, including: According to the individual action of each UAV in the virtual environment space, the joint action of multiple UAVs in the virtual environment space is obtained; Construct a reward function with the goal of maximizing the total amount of data collected and minimizing the total path length, and evaluate the flight strategy of multiple drones to reach the mission endpoint with joint actions each time. Specifically: The reward function is calculated as follows: ; in: The total amount of data collected is maximized to optimize the goal:

[0028] Express the flight path minimization optimization objective: ; Rewards for drones to explore IoT devices in a virtual environment. Representation and The associated priority weight, Representation and the associated priority weight; Based on the flight strategy, the maximum entropy QMIX algorithm is used to integrate the strategy entropy into the optimization target. Specifically, the strategy entropy is calculated by the following formula:

[0029] in: Represents the weight coefficient of strategy entropy, which is used to adjust the influence of different agents in strategy entropy; Indicates drone i In observation Next select action The probability of strategy; Calculate the state by the following formula s In the future, drones will adopt strategies The expected value of the cumulative rewards that can be obtained:

[0030] in: Indicates in status s Take action when the drone is down a The expected reward in the future, , Indicates immediate reward, represents the discount factor, Indicates the expected state value for the next state; Based on the strategy entropy and the expected value of the accumulated reward, the strategy optimization goal is determined:

[0031] in: Representation strategy The objective function is represents the discount factor at time t The power of Indicates at time t The state of Indicates at time t The movement of the drone, Indicates i strategy for each drone; Based on the reward function, train the network parameters of the local network model; The model parameters of the local network model during the training process are periodically sent to the aggregator for aggregation processing to obtain the trained drone global model. Specifically: in: represents the global model parameters of the UAV, Indicates i The model parameters of the local network model of the UAV, Indicates i The weights of the local network model of each drone.

[0032] Based on the optimal path length, multiple drones are controlled to collect data from target IoT devices.

[0033] Combination Figure 3 and Figure 4 In the figure, Ours represents the method of this embodiment. The method of this embodiment has a significant improvement in data collection efficiency, and can also effectively reduce the length of the flight path, while also improving the ability to explore the environment, which shows that the method is efficient in data collection in complex environments and the system has good robustness. Example

[0034] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the following method: Obtain location information of IoT devices and data transmission channel characteristic data in the target city area; Input the IoT device location information and the channel feature data into a pre-trained drone global model to obtain an optimal path with the shortest path length for multiple drones to collect the maximum amount of data; According to the optimal path, control multiple drones to collect data from target IoT devices; Among them, the training of the pre-trained drone global model includes: Establish an environmental model for data collection based on the number of drones and the location distribution of IoT devices; Based on the environment model, a local network model is determined with the goal of maximizing the total data collection amount of multiple UAVs and the shortest data collection path length; The network parameters of the local network model are trained and the network parameters are aggregated periodically to obtain a trained UAV global model. Example

[0035] The present invention also provides a computer device, comprising: A memory for storing instructions; The processor is used to execute instructions so that the device performs the steps of implementing the following method: Obtain location information of IoT devices and data transmission channel characteristic data in the target city area; Input the IoT device location information and the channel feature data into a pre-trained drone global model to obtain an optimal path with the shortest path length for multiple drones to collect the maximum amount of data; According to the optimal path, control multiple drones to collect data from target IoT devices; Among them, the training of the pre-trained drone global model includes: Establish an environmental model for data collection based on the number of drones and the location distribution of IoT devices; Based on the environment model, a local network model is determined with the goal of maximizing the total data collection amount of multiple UAVs and the shortest data collection path length; The network parameters of the local network model are trained and the network parameters are aggregated periodically to obtain a trained UAV global model.

[0036] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0037] The present application is described with reference to the flowcharts of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process in the flowchart can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the process in the flowchart. Figure 1 A process or multiple processes or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0038] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A function specified in a process or multiple processes.

[0039] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 The steps of a specified function in a process or multiple processes.

[0040] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. A multi-UAV collaborative data collection method, characterized in that: include: Obtain location information of IoT devices and data transmission channel characteristic data in the target city area; Input the IoT device location information and the channel feature data into a pre-trained drone global model to obtain an optimal path with the shortest path length for multiple drones to collect the maximum amount of data; According to the optimal path, control multiple drones to collect data from target IoT devices; Among them, the training of the pre-trained drone global model includes: Establish an environmental model for data collection based on the number of drones and the location distribution of IoT devices; Based on the environment model, a local network model is determined with the goal of maximizing the total data collection amount of multiple UAVs and the shortest data collection path length; The network parameters of the local network model are trained and the network parameters are aggregated periodically to obtain a trained UAV global model.

2. The multi-UAV collaborative data collection method according to claim 1, characterized in that: The environmental model for data collection is established according to the number of drones and the location distribution of IoT devices, including: The environmental model includes a multi-UAV network model, a UAV channel model and an energy consumption model; The multi-UAV network model includes: in: i Represents a collection of drones The drone number in , I represents the total number of drones; Indicates i A drone in t+ 1 moment position, Indicates i A drone in t The location at the moment, , , represents the total duration of the drone data collection mission, Indicates i A drone at a time t of x coordinate, Indicates i A drone at a time t of y coordinate, Indicates i The height of the drone; The position of the UAV at the beginning of the data collection mission is represented as , the position of the UAV at the end of the data collection mission is expressed as ; Indicates at time t Drones i Select the action that changes its position a ,action a Including moving and hovering, the drone can only choose to land at the last moment; The drone channel model includes: in: Indicated in t Moment, i The drone and k channels between IoT devices; k Represents a collection of IoT devices The IoT device number, K Indicates the number of IoT devices. The IoT device set includes devices with known location information and devices with unknown location information. Indicated in t Moment i The drone and k The distance between IoT devices, , Indicates k The location of IoT devices, , Indicates k IoT devices x coordinate, Indicates k IoT devices y coordinate; represents the shadow component with a Gaussian random variable, Indicates the reference distance d =1m logarithmic mean channel gain, represents the path loss constant, z Indicates that the channel is affected by line-of-sight (LOS) or non-line-of-sight (NLOS) conditions. ; Indicated in t Moment i The drone and k The actual channel throughput between IoT devices; Indicated in t Moment k The amount of data stored in each IoT device is , Indicates that at the beginning of the data collection task k The amount of data stored by each IoT device; Indicates i The drone and k The channel transmission rate between IoT devices, , B represents the channel bandwidth, represents the signal-to-noise ratio, , P k Indicates the signal transmission power of IoT devices. represents Gaussian white noise; Indicates the total duration of the drone data collection mission divided into consecutive time intervals; Energy consumption models include: in: Indicates i A drone in t The remaining energy at the moment; Indicates i A drone in t- The remaining energy at 1 moment; represents the energy consumed by the drone calculation, , Indicates the drone calculation power; represents the energy consumed in data transmission, P k ; Indicates the energy consumed by the drone during flight. , Indicates the flight power, V Indicates the speed of the drone; , represents the blade profile power of the UAV, Indicates the induced power of the drone in hovering. represents the average rotor induced speed of the drone when it is hovering, represents the tip speed of the UAV rotor blade, represents the drag ratio of the drone body, represents the air density, Indicates the solidity of the rotor, Represents the rotor disk area.

3. The multi-UAV collaborative data collection method according to claim 1, characterized in that: The determining, based on the environment model, a local network model with the goal of maximizing the total data collection amount of multiple UAVs and the shortest data collection path length comprises: constructing an optimization target based on the environment model with the goal of maximizing the total data collection amount of multiple UAVs and the shortest data collection path length; Based on Markov decision making, the optimization objective is mapped to the observable virtual environment space; The neural network model and particle swarm optimization algorithm are used to learn the UAV channel knowledge, determine the location information of unknown IoT devices in the virtual environment space, and obtain the local network model.

4. The multi-UAV collaborative data collection method according to claim 3, characterized in that: Based on the environmental model, the optimization objective is constructed with the goal of maximizing the total data collection amount of multiple UAVs and the shortest data collection path length, where the optimization objective is expressed as: Where: i represents the drone set The drone number in , I represents the total number of drones; Indicates i A drone in t+ 1 moment position, Indicates i A drone in t The location at the moment, , , represents the total duration of the drone data collection mission, Indicates i A drone at a time t of x coordinate, Indicates i A drone at a time t of y coordinate, Indicates i The height of the UAV; the position of the UAV at the beginning of the data collection mission is expressed as , the position of the UAV at the end of the data collection mission is expressed as ; Indicates at time t Drones i Select the action that changes its position a ,action a Including moving and hovering, the drone can only choose to land at the last moment; Indicates i The total energy of each drone; Indicates i A drone in t The remaining energy at the moment, ; Indicates i A drone in t- The remaining energy at 1 moment; It represents the energy consumed by the drone calculation; Indicates the energy consumed in data transmission; Indicates the energy consumed by the drone during flight; It means no one is there at the moment t joint action; Indicates whether the drone is collecting data from IoT devices. ; Indicates j The height of the drone.

5. The multi-UAV collaborative data collection method according to claim 3, characterized in that: The Markov decision-making method maps the optimization target to an observable virtual environment space, including: Convert the optimization objective into a tuple: ;in: Represents the state space of the virtual environment, which is composed of the drone state s composition; Indicates the actions of all drones Combined joint action space; Represents the state of the environment, which is based on the state transition function: Make the conversion; Indicates the next drone state; represents the reward function, , Represents the reward after taking an action in the current state; represents the UAV joint observation space; represents the discount factor; Represents the observation probability function; the drone is based on the observation probability function O ( , , )= Determine the local observation value.

6. The multi-UAV collaborative data collection method according to claim 3, characterized in that: The method adopts a neural network model and a particle swarm optimization algorithm to learn the UAV channel knowledge and determine the location information of the unknown IoT device in the virtual environment space, including: determining the channel gain between the UAV and the unknown IoT device by the following formula: in: It means that the drone measures the channel gain between itself and the unknown IoT device. represents the function to be sought, represents the channel parameters, Indicated in t Moment i The drone and k The distance between IoT devices, represents the elevation angle, Indicates that the channel is affected by line-of-sight (LOS) or non-line-of-sight (NLOS) conditions. , Represents the shadow component with Gaussian random variables; uses IoT devices with known location information to learn channel characteristics, using parameters equivalent to channel gain The neural network determines the channel model of the unknown IoT device; based on the channel model of the unknown IoT device, the particle swarm optimization algorithm is used to index the best particle position, and the best particle position is used as the location information of the unknown IoT device.

7. The multi-UAV collaborative data collection method according to claim 3, characterized in that: The network parameters of the local network model are trained and the network parameters are periodically aggregated to obtain a trained UAV global model, including: According to the individual action of each UAV in the virtual environment space, the joint action of multiple UAVs in the virtual environment space is obtained; Construct a reward function with the goal of maximizing the total amount of data collected and minimizing the total path length, and evaluate the flight strategy of multiple drones to reach the mission endpoint each time with joint actions; Based on the flight strategy, the maximum entropy QMIX algorithm is used to integrate the strategy entropy into the optimization target; Based on the reward function, train the network parameters of the local network model; The model parameters of the local network model during the training process are periodically sent to the aggregator for aggregation processing to obtain the trained UAV global model.

8. The multi-UAV collaborative data collection method according to claim 7, characterized in that: The flight strategy-based QMIX algorithm uses maximum entropy to integrate strategy entropy into the optimization goal, including: The policy entropy is calculated by the following formula: in: i Represents a collection of drones The drone number in , I represents the total number of drones; Represents the weight coefficient of strategy entropy, which is used to adjust the influence of different agents in strategy entropy; Indicates drone i In observation Next select action The probability of strategy; Calculate the state by the following formula s In the future, drones will adopt strategies The expected value of the cumulative rewards that can be obtained: in: Indicates in status s Take action when the drone is down a The expected reward in the future, , Indicates immediate reward, represents the discount factor, Indicates the expected state value for the next state; Based on the strategy entropy and the expected value of the accumulated reward, the strategy optimization goal is determined: in: , Indicates the total duration of the drone data collection mission; Representation strategy The objective function is represents the discount factor at time t The power of Indicates at time t The state of Indicates at time t The action of time, Indicates i Drone strategy.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the multi-UAV collaborative data collection method as described in any one of claims 1 to 8 is implemented.

10. A computer device, characterized in that: include: A memory for storing instructions; A processor is used to execute the instruction so that the device implements the multi-UAV collaborative data collection method as described in any one of claims 1 to 8.

Images