Group intelligence multi-unmanned aerial vehicle air pollution detection method based on data security

Abstract

The invention discloses a data security-based swarm intelligence multi-unmanned aerial vehicle atmospheric pollution monitoring method, which comprises the following steps that a control center initializes identity information and security parameters of an unmanned aerial vehicle base station and an unmanned aerial vehicle to be distributed in a to-be-monitored area, and locks an unmanned aerial vehicle data acquisition function at the same time; the unmanned aerial vehicle gets in contact with the unmanned aerial vehicle base station in the to-be-monitored area to complete identity verification and unlocking of a data acquisition function; and the unmanned aerial vehicle executes a monitoring task according to path information dynamically planned by the control center and encrypts and sends monitoring data back to the control center. The monitoring method provided by the invention is based on a swarm intelligence optimization algorithm, the monitoring path points of the unmanned aerial vehicle are dynamically generated, and the optimization strategy of the unmanned aerial vehicle cooperative monitoring group is combined, so that efficient monitoring of point source and non-point source pollution is realized, and sensitive information is effectively protected through identity authentication and data encryption technologies.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle monitoring, and particularly relates to a swarm intelligence multi-unmanned aerial vehicle air pollution detection method based on data security. Background Art

[0002] Air pollution is a global environmental problem that poses a serious threat to human health and the ecosystem. Traditional air pollution monitoring methods rely on wireless sensor networks, monitoring vehicles, or handheld monitors, etc. These methods have problems such as limited monitoring range, high cost, and poor mobility.

[0003] In recent years, the rapid development of unmanned aerial vehicle technology has provided a new solution for air pollution monitoring. However, although a single unmanned aerial vehicle can improve flexibility, due to limited endurance, the monitoring task execution efficiency is low, making it difficult to meet the requirement of maintaining synchronization as much as possible in the monitoring time for multiple pollution sources. In contrast, multiple unmanned aerial vehicles can monitor multiple pollution sources simultaneously. This not only shortens the overall monitoring time but also ensures the synchronization of data collection.

[0004] In a complex industrial park environment, the air pollution monitoring task of multiple unmanned aerial vehicles faces multiple challenges. First, there are various types of pollution sources, including particulate matter (such as PM2.5, PM10), gaseous pollutants (such as SO 2 , NO x , O3, CO), and volatile organic compounds (VOCs). These pollution sources not only come from well-defined point sources (such as chimneys, exhaust stacks), but also widely exist in various area sources such as material stacking yards, construction sites, and fugitive emissions. In area source monitoring, due to the wide distribution of multiple sampling points, a single unmanned aerial vehicle often has difficulty completing all tasks within the specified time window. Therefore, a multi-unmanned aerial vehicle collaborative monitoring strategy is particularly important. Second, when conducting multi-unmanned aerial vehicle air monitoring, it is necessary to consider the impact of unstable flight altitude or frequent turning of the unmanned aerial vehicle on the quality of air data collection and the energy consumption of the unmanned aerial vehicle. In addition, the production processes, pollution discharge technologies, and specific pollution indices of enterprises in the industrial park are usually regarded as trade secrets and must be properly handled to avoid risks caused by information leakage. Therefore, it is necessary to authenticate and identify the detection unmanned aerial vehicle, and at the same time, it is also crucial to protect the privacy of the data collected by the unmanned aerial vehicle.

[0005] Currently, most research focuses more on optimizing the hardware performance and flight control algorithms of unmanned aerial vehicles, and less on data security and the collaboration of multiple unmanned aerial vehicles. Therefore, developing an air pollution detection system based on the collaboration of multiple unmanned aerial vehicles that can efficiently and safely detect pollutants in a mixed area of point sources and area sources, and at the same time ensure the consistency and privacy of data, has become a key problem to be solved urgently. Summary of the Invention

[0006] The present invention aims to solve the technical problems existing in the related art to a certain extent.

[0007] The purpose of the present invention is to provide an air pollution monitoring system based on multi-UAV collaboration, which can efficiently and safely detect pollutants in the mixed area of point sources and area sources, and at the same time ensure the consistency and privacy of data.

[0008] In order to achieve the above object, the present invention provides a swarm intelligence multi-UAV air pollution monitoring method based on data security, which comprises the following steps:

[0009] S1. The control center initializes the identity information and security parameters of the UAV base station in the area to be monitored and the UAVs to be dispatched, and at the same time locks the UAV data acquisition function;

[0010] S2. The UAV contacts the UAV base station in the area to be monitored to complete identity verification, and the UAV data acquisition function is unlocked;

[0011] S3. The UAV executes the monitoring task according to the path information dynamically planned by the control center, and at the same time encrypts and sends the monitoring data back to the control center; including:

[0012] S31. The control center obtains the position information and performance parameters of all dispatched UAVs, and at the same time analyzes the information of the pollution sources to be monitored according to historical data;

[0013] S32. The control center updates the current environmental information in real time, then calls the multi-agent reinforcement learning algorithm to generate the path points of each UAV, and schedules the UAVs to execute the monitoring tasks. The UAVs encrypt the collected monitoring data and send it back to the control center.

[0014] Preferably, in step S1, the control center initializes the identity information and security parameters of the UAV base station in the area to be monitored and the UAVs to be dispatched, and at the same time locks the UAV data acquisition function, specifically:

[0015] S11. The control center selects a large integer T, determines the number m of dispatched UAVs, and selects a hash function h();

[0016] S12. The control center initializes m UAVs to be dispatched, so that they have pre-distributed public keys PK i and private keys SK i , and saves the integer m, the hash function h(), and the unique identity serial number U i of the UAV. Based on the decomposition of the large integer T, each UAV randomly obtains an integer t i , where i = 1, 2,..., m; each UAV is equipped with a camera device, a sensing device, and a communication device, and all have encryption functions;

[0017] S13. The control center commands to lock the data acquisition function of the UAV, and sets that the data acquisition function is unlocked after the UAV receives the relevant data from the base station and obtains the large integer T after completing the identity authentication;

[0018] S14. Based on the (k, m) threshold mechanism, the control center sets that the UAV restores the large integer T after obtaining k t i , where the value of k is less than m;

[0019] S15. The control center initializes s UAV base stations in the area to be monitored. Each base station has a public-private key pair labeled as (PW j , SW j ) and a unique identity information B j , where j = 1, 2,..., s, and the B j serial number is pasted on the periphery of the base station; each base station stores the public key information PK of each UAV i , the hash function h(), and the total number m of dispatched UAVs, randomly allocated (k - 1) t i , and the large integer T;

[0020] S16. Each of the m UAVs stores the public key information PW of the s base stations j .

[0021] Preferably, in step S2, the UAV makes contact with the UAV base station in the area to be monitored to complete the identity authentication, and the data acquisition function of the UAV is unlocked. The specific steps are as follows:

[0022] S21. The UAV U i selects the base station B j as its identity authentication base station. The UAV U i scans through the imaging device to obtain the base station serial number B j ;

[0023] S22. The UAV U i operates the message packet {m||U i ||B j} with the hash function h() to obtain the message digest h(m||U i ||B j ), signs h(m||U i ||B j ) with the private key SK i to obtain and encrypts it together with U i , Timex with the public key PW of B j to obtain j and sends it to the base station B j ​, where || is the message concatenation symbol, U i is the unique identification serial number of the UAV, Timex is the current timestamp, m is the total number of UAVs dispatched this time, and Sign[] is the signature function;

[0024] S23. Base station B j Establishes a queue according to the received messages in sequence, and authenticates the identities of the UAVs in turn according to the principle of first come, first served; Base station B j After receiving the message packet of U i First, use its own private key SW j to decrypt and obtain Then select its public key PK according to the identity identification of Ui i to verify the signature whether it is consistent with the hash operation h(m||U i ||B j ) performed by the base station. If it is consistent, the verification passes and proceeds to step S24. If it is inconsistent, the verification fails and reports to the control center;

[0025] S24. Base station B j Encrypts (k - 1) t i using the public key PK of U i to obtain i and sends it to the UAV U ; i

[0026] S25. UAV U i After receiving uses the private key to decrypt and obtain (k - 1) t x , together with the t it holds i , to recover the large integer T to unlock the data acquisition function.

[0027] Preferably, in step S31, the control center obtains the position information and performance parameters of all dispatched UAVs, and analyzes the information of the pollution sources to be monitored based on historical data. Specifically:

[0028] S311. The control center obtains the current position of the base station B i where the UAV U is located j and collects the performance parameters of all UAVs at the same time. At the same time, it analyzes the information of the pollution sources to be monitored through historical monitoring data, including: the three-dimensional coordinates (x , y κ , z κ ) of the point source, and the three-dimensional coordinates (x κ ) of multiple monitoring sampling points of the area source, y κ,g , z κ,g , z κ,g ​​), and the monitoring duration of each pollution source For area sources, the monitoring duration is the same for each sampling point, and a time window is set for the monitoring of each pollution source of the area source. To ensure data consistency, where κ represents the number of the pollution source, κ = 1, 2, …, K; g represents the number of the sampling point of the area source, g = 1, 2, …, G;

[0029] S312. The control center calculates the geometric center and total monitoring duration of all monitoring sampling points of the area source pollution source, which is defined as the virtual point source of the area source.

[0030] Preferably, in step S32, the control center updates the current environmental information in real time, then calls the multi-agent reinforcement learning algorithm to generate the path points of each drone, and schedules the drones to perform the monitoring tasks. Specifically:

[0031] S321. The control center updates the current environmental information, calls the multi-agent reinforcement learning algorithm to generate the path points of each drone, and schedules the drones to perform the monitoring tasks;

[0032] When the drone flies to the target monitoring point according to the task instruction, it collects the gas characteristic data in the surrounding environment in real time. When the concentration of a certain type of pollutant exceeds the set normal threshold, and the position of the pollution source is a new position not recorded in the historical data, the drone will automatically determine it as a suspicious newly added pollution source. At the same time, privacy protection is implemented for the pollutant concentration data, and the encrypted pollutant concentration data, other gas characteristic data, and three-dimensional position information are sent to the control center in real time, and then it goes to step S323; if no newly added pollution source is detected, the original task is continued, and it goes to step S324;

[0033] S323. The control center processes the pollution source information through an artificial intelligence algorithm, analyzes the pollution source type, monitoring time, and monitoring sampling points, and adds them to the unallocated pollution source set Φ unmonitored , and at the same time obtains the states of all drones at this time, and then goes to step S322 to re-plan the drone paths;

[0034] S324. Drone U i Arrives at the monitoring sampling point and determines whether it is a point source or a virtual point source. If it is a virtual point source, taking the virtual point source position as the starting point and the next monitoring point of the drone as the end point, a sub-visit sequence is formed; call the local insertion heuristic algorithm to optimize the monitoring visit order of the sampling points within the area source pollution source of the drone, and evaluate whether the drone has the ability to complete the area source monitoring alone based on this order.

[0035] The ability evaluation formula is:

[0036]

[0037] wherein represents the total flight time of the UAV U i within the area source ; represents the hovering time of the UAV U i at the g-th monitoring and sampling point

[0038] If go to step S325 for execution

[0039] Otherwise, execute the existing monitoring tasks according to the generated optimal path, and execute step S326;

[0040] S325. The UAV U i will send a collaborative request signal to other nearby UAVs and attach the information of the area source pollution source that needs to be collaboratively monitored to form a UAV collaborative monitoring group group. The control center dynamically optimizes the task allocation within the collaborative group using the swarm intelligence optimization algorithm;

[0041] S326. After the UAV arrives at the monitoring point, it hovers and continuously senses the pollutant concentration for a sensing time not less than Encrypt the collected pollutant concentration data values according to step S322, and then send the processed concentration data together with the corresponding time stamps to the control center;

[0042] S327. After the task is completed, the UAV goes to the next monitoring point according to the predetermined path or returns to the nearest base station for charging. During charging, the UAV will turn off the sensing system and communication module, and resume the remaining monitoring tasks after charging is completed.

[0043] Preferably, in step S321, the multi-agent reinforcement learning algorithm is called to generate the path points of each UAV, specifically:

[0044] The multi-agent reinforcement learning algorithm outputs the next action according to the current state s t which is the next monitoring point p selected by the UAV i,α+1 or returns to the nearest base station for charging. The UAV obtains the corresponding reward or penalty, and at the same time updates the environmental state to s according to the selected action t+1 and iteratively generates the next action until all monitoring tasks are completed

[0045] Design the following reward function and penalty function to constrain the actions of the UAV

[0046] Reward function:

[0047]

[0048] Among them, d(a,b) represents the function of the straight-line distance between two points a and b; p i,α represents the hovering position of the α-th monitoring point of the UAV U i ; w 1 represents the path length incentive factor; L max represents the maximum endurance distance of the UAV;

[0049]

[0050] Among them, h i,α represents the hovering height of the UAV U i at the α-th monitoring point, |h i,α -h i,α+1 | represents the height that the UAV U i needs to rise or fall to the next position; w 2 represents the flight stability incentive factor; H max represents the maximum height difference of all pollution sources;

[0051]

[0052] Among them, θ i,α+1 represents the angle that the UAV needs to turn corresponding to the next position; w 3 represents the turning incentive factor; w 1 +w 2 +w 3 =1; Θ max represents the maximum turning angle of the UAV,

[0053] Penalty function:

[0054]

[0055] Among them, represents the minimum energy required for the UAV U i to return to the nearest base station at the current position; E i (t) represents the remaining energy of the UAV at this time, w 4 represents the energy penalty factor, w 4 >1, and increases with the number of iterations,

[0056] Therefore, the reward function for the next monitoring point selected by the UAV is:

[0057] R = R 1 +R 2 +R 3 +Penalty

[0058] Preferably, in step S322, the pollutant concentration data is encrypted, and the specific method is:

[0059] The drone will add noise to the concentration data of pollutants to achieve privacy protection by adding noise that follows a Laplace distribution to the data. The drone determines the privacy budget ε requirement and sensitivity △f in advance, where △f is defined as the difference between the current concentration data and the set normal threshold. Thus, the probability density function of the Laplace distribution is obtained as

[0060]

[0061] For the concentration data value x of a certain type of pollutant component collected, after adding the generated Laplace noise y, the new data value obtained is x + y for uploading, which will not affect subsequent processing.

[0062] Preferably, in step S325, the drone U i will send a collaborative request signal to other nearby drones and attach the information of the area source pollution source that needs to be collaboratively monitored to form a drone collaborative monitoring group group. The control center uses a swarm intelligence optimization algorithm to dynamically optimize the task allocation within the collaborative group. The specific steps are as follows:

[0063] S3251. Each drone that receives the collaborative signal uploads its status information to the control center, and the control center introduces a collaborative factor C i , which represents the importance of the drone U i in the collaborative monitoring. The design formula is as follows:

[0064]

[0065] Among them, d i,κ represents the distance between the drone U i and the virtual point source of the area source that needs to be collaboratively monitored ; γ i represents the remaining unmonitored pollution source quantity of the drone U i , and β i represents the current task completion progress. When 0 < β i < 1, it means that the drone is using the sensor to monitor the gas concentration of the current pollution source. The larger β i is, the closer the task is to completion. When β i = 1, it means that the drone has completed the task of the current monitoring point and is ready to fly to the next monitoring point, or is in a flight state;

[0066] S3252. The control center calculates the C i of all candidate drones, sorts the candidate drones according to the size of the collaborative factor, and gradually selects the drones with larger collaborative factors from the candidate drones to join the drone collaborative monitoring group group, removes the subsequent monitoring tasks of the drone, and adds them to the unassigned pollution source set Φ unmonitoredand dynamically evaluate the ability A of the entire collaborative monitoring group group until A group is not greater than the time window of the area source so as to determine the appropriate number B of drones,

[0067] Design an evaluation formula for the ability A of the collaborative monitoring group group as follows:

[0068]

[0069] where λ i is the task load coefficient, which is determined by the ratio of the collaborative factor of the drone U i to the total number within the group, and A i,k represents the ability of the drone U i to independently monitor the area source after using the insertion heuristic algorithm to plan the path;

[0070] S3253. The control center calls the swarm intelligence optimization algorithm in two stages to allocate monitoring points for the drones in the monitoring group group: In the first stage, the task of the monitoring group is the sampling points within the area source, and the drones start from the position where they receive the collaborative signal; In the second stage, the task of the monitoring group is the set Φ of unmonitored pollution sources unmonitored and the drones start from the position where they complete the area source monitoring.

[0071] Preferably, the insertion heuristic algorithm is as follows: By trying to insert each monitoring point within the area source into different positions of the current path, calculate the optimization objective F after insertion, and record the insertion position with the minimum optimization objective, update the path, and iterate continuously until all monitoring points are inserted to obtain the final monitoring access order. The optimization objective F and constraint conditions of the algorithm are as follows:

[0072] Objective function F:

[0073]

[0074] Constraint conditions:

[0075] Energy constraint, the remaining energy E i of each drone U i (t) at any time t cannot be negative;

[0076] Position constraint, the drone needs to return to the nearby base station after completing the task;

[0077] Full coverage constraint, all pollution source monitoring points within the industrial park must be monitored;

[0078] Non - repetition constraint, each pollution source monitoring point is only monitored once by one drone, that is, there is no intersection between the monitoring point access sequences of any two drones;

[0079] Data consistency constraint. Once the drone U i starts monitoring a sampling point of the area source pollution source, all monitoring sampling points need to be completed within the time window ;

[0080] Hovering constraint. During the monitoring execution, the hovering time of the drone is not less than the required monitoring time of the pollution source and cannot be interrupted during the monitoring, that is, a monitoring point is either monitored completely at one time or not monitored;

[0081] Among them, L i represents the total path length of the drone U i ; H i represents the flight altitude stability of the drone U i , which is set as the average value of the altitude change in the flight path; Θ i represents the average value of the angles that the drone U i needs to turn during the flight.

[0082] Beneficial effects:

[0083] 1. The present invention verifies the identity of the drone through the (k, m) threshold mechanism, which greatly improves the security of the system. This mechanism effectively prevents unauthenticated drones from invading the system, thus ensuring the integrity and authenticity of the monitoring data, reducing the risk of data collection interruption caused by single-point failure or unauthorized access, improving the reliability of the entire monitoring system, and ensuring the smooth progress of the monitoring task in a complex environment.

[0084] 2. The present invention combines differential privacy protection technology and adds Laplace noise to the collected pollution source concentration data, effectively protecting the privacy of sensitive data in the industrial park and ensuring that the privacy of enterprises will not be leaked during the collection process. This measure significantly improves data security in the big data environment and ensures the balance between effectiveness and privacy protection during data analysis.

[0085] 3. The present invention designs a path planning algorithm based on multi-agent reinforcement learning to dynamically plan the optimal monitoring path of the drone, thereby realizing real-time and accurate monitoring of multiple pollution sources. By automatically identifying and marking newly added suspicious pollution sources, it can dynamically adjust the detection task, effectively optimize the energy consumption of the drone, and improve the efficiency of the monitoring task. By introducing a multi-dimensional reward and punishment mechanism, the system effectively motivates the drone to optimize path selection and energy management during the monitoring task execution.

[0086] 4. The present invention introduces a multi-UAV collaborative monitoring group strategy and uses a swarm intelligence optimization algorithm to achieve the collaborative work of multiple UAVs in a complex environment. Through the dynamic information sharing and collaborative scheduling among UAVs, the monitoring efficiency of UAVs in the area source region is effectively improved. Compared with the traditional monitoring method, this system can quickly organize multiple UAVs to work collaboratively, ensure the timely collection and high quality of pollution source data, significantly shorten the overall monitoring time, and meet the complex environmental monitoring requirements in industrial parks. Description of the Drawings

[0087] Figure 1 Flowchart of a swarm intelligence multi-UAV air pollution detection method based on data security provided by the invention;

[0088] Figure 2 Global flowchart of the embodiment of the present invention;

[0089] Figure 3 Flowchart of the UAV collaborative monitoring strategy in the embodiment of the present invention;

[0090] Figure 4 Overview diagram of the swarm intelligence multi-UAV air pollution detection method based on data security provided by the embodiment of the present invention. Detailed Embodiments

[0091] To make the objectives, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below in conjunction with the drawings in the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. They should not be construed as limiting the present invention. Based on the embodiments in the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the scope of protection of the present invention. In the description of the present invention, it should be understood that the terms used are only for the purpose of description and cannot be construed as indicating or implying relative importance.

[0092] Embodiment: The following will describe a swarm intelligence multi-UAV air pollution monitoring method based on data security provided by the present invention in conjunction with Figures 1-4 Describe a swarm intelligence multi-UAV air pollution monitoring method based on data security provided by the present invention.

[0093] As Figure 1 、 Figure 2 、 Figure 4 shown, this monitoring method includes the following steps:

[0094] S1. The control center initializes the identity information and security parameters of the UAV base station and the UAVs to be dispatched in the area to be monitored, and at the same time locks the UAV data collection function;

[0095] Step 1-1: The pollution monitoring and supervision agency that dispatches the drones carefully selects a large integer T, determines the number of dispatched drones to be 6 according to the historical pollution source monitoring data of the industrial park, and selects a hash function h().

[0096] Step 1-2: The pollution monitoring and supervision agency initializes m drones for pollution monitoring in the industrial park. Each drone is equipped with a high-definition camera, a fusion gas sensor, a lidar, and communication equipment, etc. It has a pre-distributed public key PK i and a private key SK i , and saves the integer m, the hash function h(), and the unique identity serial number U i of the drone. Based on the factorization of the large integer T, each drone randomly obtains an integer t i , where i = 1, 2,..., m; each drone can add local differential privacy noise to the data.

[0097] Step 1-3: The pollution monitoring and supervision agency sets an instruction to lock the data collection function of the drones. The data collection function can only be unlocked after receiving relevant data from the base station and obtaining the large integer T after completing the identity verification.

[0098] Step 1-4: The pollution monitoring and supervision agency is based on the (k, m) threshold mechanism. As long as k t i are obtained, the large integer T can be restored, where the value of k is much smaller than m.

[0099] Step 1-5: s drone base stations are set up in the industrial park to authenticate the drones and charge for endurance; each base station has a public-private key pair labeled (PW j , SW j ) and a unique identity information B j , where j = 1, 2,..., s, and the B j serial number will be pasted on the periphery of the base station; each base station saves the public key information PK i of each drone, the hash function h(), and the total number m of dispatched drones, the randomly assigned (k - 1) t i , and the large integer T.

[0100] Step 1-6: The m drones respectively save the public key information PW j of the s base stations.

[0101] S2. The drone contacts the drone base station in the area to be monitored to complete the identity verification and unlock the data collection function;

[0102] Step 2-1: Without loss of generality, assume that a certain drone U i selects a nearby base station B j with the maximum transmission power.As its identity authentication base station, the drone U i scans the base station serial number B through the camera device j ;

[0103] Step 2-2: If multiple drones send information to the same base station at the same time, the base station establishes a queue according to the received messages in sequence and performs identity authentication in turn according to the principle of first come, first served;

[0104] Step 2-3: The drone U i operates the message packet {m||U i ||B j} with the hash function h() to obtain the message digest h(m||U i ||B j ), signs h(m||U i ||B j ) with the private key SK i to obtain together with U i , Timex is encrypted with the public key PW j of B j to obtain and sent to the base station B j , where || is the message concatenation symbol, U i is the unique identity serial number of the drone, Timex is the current timestamp, m is the total number of drones dispatched this time, and Sign[] is the signature function;

[0105] Step 2-4: After the base station B j receives the message packet of U i it first decrypts it with its own private key SW j to obtain Then, according to the identity identifier of Ui, its public key PK i is used to verify whether the signature is consistent with the hash operation h(m||U i ||B j ) performed by the base station. If it is consistent, the verification passes and step 2-5 is executed. If it is inconsistent, the verification fails and the park management center will be reported;

[0106] Step 2-5: The base station B j encrypts (k - 1) t i with the public key PK i of U i to obtain and sends it to the drone U i (To distinguish the t i held by the drone, t x is used here to represent the t i held by the base station)));

[0107] Step 2-6: Drone U i Receives and decrypts it with the private key to obtain (k - 1) t's x , together with the t it holds i , to recover the large integer T to unlock the data collection function.

[0108] S3. The drone performs monitoring tasks according to the path information dynamically planned by the control center, and at the same time encrypts and sends the monitoring data back to the control center;

[0109] Step 3-1: The pollution monitoring and supervision agency obtains the current positions of all drones at base stations B j through the GPS positioning system of the drones and collects the performance parameters of all drones (drone flight speed 5 m / s, drone endurance time 40 min). At the same time, through historical monitoring data, the information of the pollution sources to be monitored is analyzed , including: the three-dimensional coordinates (x κ , y κ , z κ ) of point sources, the three-dimensional coordinates (x κ,g , y κ,g , z κ,g ) of multiple monitoring sampling points of area sources, and the monitoring duration of each pollution source For area sources, the monitoring duration of each sampling point is the same, and a time window is set for the monitoring of each area source pollution source to ensure data consistency. Among them, κ represents the number of pollution sources, κ = 1, 2,..., K; g represents the number of sampling points of area sources, g = 1, 2,..., G. The number of pollution sources K is set to 15, and the specific information is shown in Table 1.

[0110] Table 1 Pollution source information

[0111]

[0112] Step 3-2: The pollution monitoring and supervision agency calculates the geometric center and total monitoring duration of all monitoring sampling points of area source pollution sources, which are defined as the virtual point sources of the area sources. All virtual point sources will be used as the benchmarks for subsequent local collaborative optimization to complete the monitoring tasks of area source pollution sources by multiple drones.

[0113] For example:

[0114] Step 3-3: The pollution monitoring and supervision agency updates the current environmental information s t, including the current positions of all drones, remaining battery levels, all unassigned pollution source monitoring locations and monitoring times, etc. Call the MADDPG algorithm to generate the path points for each drone and dispatch the drones to perform the monitoring tasks. In each round of path generation, the MADDPG algorithm outputs the next action based on the current state s t The next action is the next monitoring point p i,α+1 (including virtual point sources) or return to the nearest base station for charging. The drone obtains the corresponding rewards or punishments and updates the environmental state to s t+1 according to the selected action, and iteratively generates the next action until all monitoring tasks are completed.

[0115] To motivate the drones to choose the shortest monitoring paths, reduce flight altitude changes and turning angles to reduce energy consumption, the following reward function is designed to constrain the actions of the drones.

[0116]

[0117] Among them, d(a,b) represents the function of the straight-line distance between two points a and b; p i,α represents the hovering position of the α-th monitoring point of drone U i ; w 1 represents the path length incentive factor; L max represents the maximum endurance distance of the drone;

[0118]

[0119] Among them, h i,α represents the hovering altitude of drone U i at the α-th monitoring point, |h i,α -h i,α+1 | represents the height that drone U i needs to rise or fall to the next position; w 2 represents the flight stability incentive factor; H max represents the maximum height difference of all pollution sources;

[0120]

[0121] Among them, θ i,α+1 represents the turning angle required for the drone corresponding to the next position; w 3 represents the turning incentive factor; w 1 +w 2 +w 3 =1; Θ max represents the maximum turning angle of the drone.

[0122] According to the relationship between the remaining energy of the UAV and the energy required to return to the base station, a penalty function is designed. If the UAV ignores the low-energy state and continues to select monitoring tasks, it will be penalized according to the following formula:

[0123]

[0124] Among them, represents the minimum energy required for the UAV U i to return to the nearest base station at the current location; E i (t) represents the remaining energy of the UAV at this time. w 4 represents the energy penalty factor, w 4 > 1 and increases with the number of iterations.

[0125] Therefore, the reward function for the next monitoring point selected by the UAV is:

[0126] R = R 1 + R 2 + R 3 + Penalty

[0127] Step 3-4: When the UAV flies to the target monitoring point according to the mission instructions, it starts various devices such as a high-definition camera, a fused gas sensor, and a lidar to collect gas characteristic data (pollutant concentration, temperature, humidity, wind speed, etc.) in the surrounding environment in real time. Calculate and set the normal concentration thresholds of all pollutants according to the national environmental air quality standards. When the concentration of a certain type of pollutant (SO 2 , NOx, CO, VOCs, PM10, PM2.5, etc.) exceeds the set normal threshold, and the location of the pollution source is a new location not recorded in the historical data, the UAV will automatically determine it as a suspected newly added pollution source.

[0128] To prevent the collected data from being attacked and leaked during transmission, the UAV will add noise to the pollutant concentration data to achieve privacy protection by adding noise that follows a Laplace distribution to the data. The UAV determines the privacy budget ε requirement and the sensitivity △f in advance. △f can be defined as the difference between the current concentration data and the set normal threshold. Thus, the probability density function of the Laplace distribution is

[0129]

[0130] For the concentration data value x of a certain type of pollutant component collected, add the generated Laplace noise y

[0131] and then the new data value obtained is x + y for uploading, which will not affect subsequent processing.

[0132] The UAV transmits the polluted concentration data with added noise, other gas characteristic data (temperature, humidity, wind speed, etc.) and three-dimensional position information to the pollution monitoring and supervision agency in real time, and proceeds to step 3-5; if no new pollution source is detected, the original task is continued and step 3-6 is executed.

[0133] Step 3-5: The pollution monitoring and supervision agency processes the pollution source information through image recognition and pattern matching algorithms, analyzes the pollution source type, monitoring time and monitoring sampling points, and adds them to the unassigned pollution source set Φ. unmonitored Obtain the positions and remaining battery levels of all UAVs at this time, and add all unmonitored pollution sources to the unassigned pollution source set Φ. unmonitored Proceed to step 3-3 to re-plan the UAV path.

[0134] Step 3-6: UAV U i Arrives at the monitoring sampling point and determines whether it is a point source or a virtual point source. If it is a virtual point source, starting from the virtual point source position, the next monitoring point of UAV U i is the end point, forming a sub-visit sequence; call the local insertion heuristic algorithm to optimize the monitoring visit order of the sampling points within the UAV area source pollution source. Specifically, the algorithm tries to insert each monitoring point within the area source into different positions of the current path, calculates the optimization objective F after insertion, and records the insertion position with the minimum optimization objective, updates the path, and iterates continuously until all monitoring points are inserted to obtain the final monitoring visit order. Evaluate whether UAV U i has the ability to complete the area source monitoring alone.

[0135] The ability evaluation formula is:

[0136]

[0137] where represents the total flight time of UAV U i within the area source ; represents the hovering time of UAV U i at the gth monitoring sampling point.

[0138] If proceed to step 3-7 for execution.

[0139] Otherwise, execute the existing monitoring tasks according to the generated optimal path and execute step 3-8.

[0140] For example: For the sub-visit monitoring sequence obtain the final monitoring visit sequence calculate

[0141] Step 3-7: The UAV will send a cooperation request signal to other nearby UAVs and attach the information of the area source pollution source that needs to be cooperatively monitored to form a UAV cooperative monitoring group group. The genetic algorithm based on multi-chromosome encoding will be used to dynamically optimize the task allocation within the cooperative group.

[0142] The specific steps are as follows:

[0143] Step 3-7-1: Each UAV that receives the cooperation signal uploads its status information (including remaining energy, distance from the virtual point source, number of unfinished tasks, etc.) to the pollution supervision and monitoring agency. Introduce a cooperation factor C i , which represents the importance of UAV U i in the cooperative monitoring, and the design formula is as follows:

[0144]

[0145] where d i,κ represents the distance between UAV U i and the virtual point source of the area source that needs to be cooperatively monitored ; γ i represents the number of unmonitored pollution sources remaining for UAV U i . β i represents the current task completion progress. When 0 < β i < 1, it means that the UAV is using the sensor to monitor the gas concentration of the current pollution source. The larger β i , the closer the task is to completion. When β i = 1, it means that the UAV has completed the task of the current monitoring point and is ready to fly to the next monitoring point, or is in the flight state.

[0146] Step 3-7-2: The pollution supervision and monitoring agency calculates the C i of all candidate UAVs, and sorts the candidate UAVs according to the size of the cooperation factor. The UAV with a larger cooperation factor has a higher priority in the cooperative task. Gradually select the UAVs with larger cooperation factors from the candidate UAVs to join the UAV cooperative monitoring group group, remove the subsequent monitoring tasks of the UAV, add them to the unallocated pollution source set Φ unmonitored , and dynamically evaluate the ability A group of the entire cooperative monitoring group until A group is not greater than the time window of the area source to determine the appropriate number of UAVs B.

[0147] Design the formula for evaluating the ability A group of the cooperative monitoring group as follows:

[0148]

[0149] where λ i is the task load factor, which is determined by the ratio of the cooperation factor of the UAV U i to the total number within the group. A i,κ represents the ability of the UAV U i to independently monitor the area source after planning the path using the insertion heuristic algorithm.

[0150] Step 3-7-3: The pollution supervision and monitoring agency calls the genetic algorithm based on multi-chromosome encoding in two stages to allocate monitoring points for the UAVs in the monitoring group group, and sets each chromosome in the individual as the monitoring access sequence of each UAV in the monitoring group group. In the first stage, the task of the monitoring group is the sampling points within the area source, and each UAV starts from the position where it receives the cooperation signal. In the second stage, the task of the monitoring group is the set Φ unmonitored of unmonitored pollution sources, and each UAV starts from the position where it completes the area source monitoring.

[0151] The optimization objective F and constraints of the algorithm are as follows.

[0152] Objective function F:

[0153]

[0154] Constraints:

[0155] · Energy constraint, the remaining energy E i (t) of each UAV U at any time t cannot be negative. i

[0156] · Position constraint, the UAV needs to return to the nearby base station after completing the task.

[0157] · Full coverage constraint, all pollution source monitoring points in the industrial park must be monitored.

[0158] · Non-repetition constraint, each pollution source monitoring point is only monitored once by one UAV, that is, there is no intersection between the monitoring point access sequences of any two UAVs.

[0159] · Data consistency constraint, once the UAV U i starts to monitor a sampling point of the area source pollution source, all monitoring sampling points must be completed within the time window .

[0160] · Hovering constraint, during the monitoring, the hovering time of the UAV is not less than the required monitoring time of the pollution source and cannot be interrupted during the monitoring period, that is, a monitoring point is either monitored once or not monitored.

[0161] where L iDenote the total path length of the drone U i ; H i Denote the flight altitude stability of the drone U i , which is set as the average value of the altitude change in the flight path; Θ i Denote the average value of the angles that the drone U i needs to turn during the flight process.

[0162] Step 3-8: After the drone arrives at the monitoring point, it hovers and uses the equipped multi-sensor system to continuously sense the pollutant concentration for no less than Record the concentration data of gas components (such as SO 2 , NO x , PM2.5, etc.) and the current timestamp. For the collected pollutant concentration data values, Laplace noise y will be added according to Step 3-4 to generate the noise-added pollutant concentration data. Then, the processed concentration data x + y and the corresponding timestamp will be sent to the pollution supervision and monitoring agency together. After the task is completed, the drone will go to the next monitoring point according to the predetermined path or return to the nearest base station for charging. During charging, the drone will turn off the sensing system and the communication module, and resume the remaining monitoring tasks after charging is completed.

[0163] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A swarm intelligence multi-UAV air pollution monitoring method based on data security, characterized in that: The steps include: S1. The control center initializes the identity information and security parameters of the drone base station in the monitored area and the drone to be dispatched, and locks the drone data collection function; S2. The drone contacts the drone base station in the monitored area to complete identity authentication, and the drone data collection function is unlocked; S3. The drone performs the monitoring task according to the path information dynamically planned by the control center, and encrypts the monitoring data and sends it back to the control center; including: S31, the control center obtains the location information and performance parameters of all dispatched drones, and analyzes the information of the pollution sources to be monitored based on historical data; S32. The control center updates the current environmental information in real time, then calls the multi-agent reinforcement learning algorithm to generate the path points for each drone, and dispatches the drone to perform monitoring tasks. The drone encrypts the collected monitoring data and sends it back to the control center.

2. The method for monitoring air pollution by multiple UAVs using swarm intelligence based on data security according to claim 1 is characterized in that: Step S1: The control center initializes the identity information and security parameters of the drone base station in the monitored area and the drone to be dispatched, and locks the drone data collection function, specifically: S11, the control center selects a large integer T, determines the number m of dispatched drones, and selects a hash function h(); S12, the control center initializes m drones to be dispatched, so that they have the pre-distributed public key PK i and private key S Ki, and save the integer m, hash function h(), and the drone’s unique identification serial number U i , based on the decomposition of the large integer T, each drone randomly obtains an integer t i , where i = 1, 2, ..., m; each drone is equipped with camera equipment, sensor equipment and communication equipment, and all have encryption functions; S13, the control center instructs the drone's data collection function to be locked, and sets the drone to unlock the data collection function after receiving the relevant data from the base station after completing identity authentication and obtaining a large integer T; S14, the control center sets the UAV to obtain k t based on the (k, m) threshold mechanism. i Then the large integer T is recovered, where the value of k is less than m; S15. The control center initializes the s drone base stations in the monitoring area. Each base station has a public-private key pair (PW j , S Wj) and unique identity information B j , where j = 1, 2, ..., s, B j The serial number is pasted on the periphery of the base station; each base station stores the public key information PK of each drone i , hash function h(), total number of dispatched drones m, randomly assigned (k-1) t i , and a large integer T; S16, m drones all save the public key information PW of s base stations j .

3. The method for monitoring air pollution by multiple drones using swarm intelligence based on data security according to claim 2 is characterized in that: Step S2: The drone contacts the drone base station in the monitored area to complete identity authentication, and the drone data collection function is unlocked. The specific steps are as follows: S21, UAV i Select Base Station B j As its identity authentication base station, drone U i Obtain the base station serial number B by scanning with a camera j ; S22, UAV i The message package {m||U i ||B j }After using the hash function h() to calculate the message digest h(m||U i ||B j ), h(m||U i ||B j )Use private key SK i After signing Together with U i 、Timex uses B j The public key PW j After encryption, Send to base station B j , where || is the message connection symbol, U i is the unique identification serial number of the drone, Timex is the current timestamp, m is the total number of drones dispatched this time, and Sign[] is the signature function; S23, base station B j According to the received messages, queues are established and the identities of drones are authenticated in turn according to the principle of first come first served; Base station B j Receive U i Message Pack After that, first use your own private key SW j After decryption, we get Then select its public key PK according to Ui's identity i Verify the signature Whether to perform hash operation h(m||U with the base station i ||B j ) are consistent, if they are consistent, the verification is passed and the process goes to step S24 for execution; if they are inconsistent, the verification fails and the process is reported to the control center; S24, base station B j Use U i The public key PK i Encrypt (k-1) t i get And send it to the drone U i ; S25, UAV i receive Then use the private key to decrypt and get (k-1) t x , together with the t i , recover the large integer T to unlock the data acquisition function.

4. The method for monitoring air pollution by multiple drones using swarm intelligence based on data security according to claim 3 is characterized in that: Step S31: The control center obtains the location information and performance parameters of all dispatched drones, and analyzes the pollution source information to be monitored based on historical data, specifically: S311, the control center obtains the UAV i Base station B j Current location At the same time, the performance parameters of all drones are collected, and the pollution sources to be monitored are analyzed through historical monitoring data The information includes: the three-dimensional coordinates of the point source (x κ ,y κ , z κ ), the three-dimensional coordinates of multiple monitoring sampling points of the surface source (x κ，g ,y κ，g , z κ，g ), and the monitoring duration for each pollution source For non-point sources, each sampling point is monitored for the same duration, and a time window is set for monitoring each non-point source of pollution. Ensure the consistency of data, where κ represents the number of the pollution source, κ = 1, 2, ..., K; g represents the number of the sampling point of the surface source, g = 1, 2, ..., G; S312. The control center calculates the geometric center of all monitoring sampling points of the surface source pollution source and the total monitoring time, and defines it as a virtual point source of the surface source.

5. The method for monitoring air pollution by multiple drones using swarm intelligence based on data security according to claim 1 is characterized in that: Step S32: The control center updates the current environment information in real time, then calls the multi-agent reinforcement learning algorithm to generate the path points for each drone, and dispatches the drone to perform the monitoring task. The drone encrypts the collected monitoring data and sends it back to the control center. Specifically: S321, the control center updates the current environment information, calls the multi-agent reinforcement learning algorithm to generate the path points for each drone, and dispatches the drone to perform the monitoring task; S322, when the drone flies to the target monitoring point according to the task instruction, it collects gas characteristic data in the surrounding environment in real time. When the concentration of a certain type of pollutant exceeds the set normal threshold, and the location of the pollution source is a new location not recorded in the historical data, the drone will automatically determine it as a suspected new pollution source, and encrypt the pollutant concentration data at the same time, and send the encrypted pollutant concentration data, other gas characteristic data and three-dimensional location information to the control center in real time, and go to step S323; if no new pollution source is detected, continue to execute the original task, and go to step S324; S323, the control center processes the pollution source information through artificial intelligence algorithms, analyzes the pollution source type, monitoring time and monitoring sampling points, and adds it to the unassigned pollution source setΦ unmonitored , and obtain the status of all drones at this time, and go to step S322 to re-plan the drone path; S324, UAV i Arriving at the monitoring sampling point, it is determined whether it is a point source or a virtual point source. If it is a virtual point source, the virtual point source position is used as the starting point and the next monitoring point of the drone is used as the end point to form a sub-access sequence; the local insertion heuristic algorithm is called to optimize the monitoring access sequence of the sampling points within the drone surface source pollution source, and based on this sequence, it is evaluated whether the drone has the ability to complete surface source monitoring alone. The ability assessment formula is: in Indicates drone U i In the surface source Total flight time; Indicates drone U i The hovering time of the g-th monitoring sampling point, like Go to step S325 to execute, Otherwise, the existing monitoring task is executed according to the generated optimal path, and step S326 is executed; S325, UAV i A collaborative request signal will be sent to other nearby drones with the information of the non-point source pollution sources that need to be collaboratively monitored, forming a drone collaborative monitoring group. The control center will use the swarm intelligence optimization algorithm to dynamically optimize the task allocation within the collaborative group. S326. After arriving at the monitoring point, the drone will hover and continuously sense the concentration of pollutants for a period of not less than The collected pollutant concentration data values ​​are encrypted according to step S322, and then the processed concentration data is sent to the control center together with the corresponding timestamp; S327. After the mission is completed, the drone goes to the next monitoring point according to the predetermined path, or returns to the nearest base station for charging. During the charging period, the drone will turn off the perception system and communication module, and resume the remaining monitoring mission after charging is completed.

6. The method for monitoring air pollution by multiple drones using swarm intelligence based on data security according to claim 5 is characterized in that: In step S321, the multi-agent reinforcement learning algorithm is called to generate the path points of each drone, specifically: The multi-agent reinforcement learning algorithm is based on the current state s t Output the next action, which is the next monitoring point p selected by the drone i，α+1 Or return to the nearest base station to charge, the drone receives corresponding rewards or penalties, and updates the environment state to s according to the selected action. t+1 , and iteratively generate the next action until all monitoring tasks are completed. Design the following reward function and penalty function to constrain the drone's actions: Reward function: Where d(a, b) represents the function of the straight-line distance between points a and b; p i，α Indicates drone U i The hovering position of the αth monitoring point; w1 represents the path length excitation factor; L max Indicates the maximum endurance distance of the drone; Among them, h i，α Indicates drone U i The hovering height at the αth monitoring point, |h i，α -h i，α+1 |Indicates droneU i The height required to ascend or descend to the next position; w2 represents the flight stability incentive factor; H max Indicates the maximum height difference of all pollution sources; Among them, θ i，α+1 Indicates the angle that the drone needs to turn to for the next position; w3 represents the steering excitation factor; w1+w2+w3=1; Θ max Indicates the maximum turning angle of the drone, Penalty function: in, Indicates drone U i The minimum energy required to return to the nearest base station at the current location; E i (t) represents the remaining energy of the drone at this time, w4 represents the energy penalty factor, w4>1, and as the number of iterations increases, Therefore, the reward function for the next monitoring point selected by the drone is: R=R1+R2+R3+Penalty.

7. The method for monitoring air pollution by multiple drones using swarm intelligence based on data security according to claim 5 is characterized in that: In step S322, the pollutant concentration data is encrypted, and the specific method is as follows: The drone will add noise to the pollutant concentration data to achieve privacy protection by adding noise that follows the Laplace distribution to the data. The drone will add noise according to the pre-determined privacy budget ε and sensitivity Δf, where Δf is defined as the difference between the current concentration data and the set normal threshold. The probability density function of the Laplace distribution is: For the collected data value x of the concentration of a certain type of pollutant component, the generated Laplace noise y is added to obtain a new data value x+y for uploading, which will not affect subsequent processing.

8. The method for monitoring air pollution by multiple drones using swarm intelligence based on data security according to claim 5 is characterized in that: Step S325 Drone U i A collaborative request signal will be sent to other nearby drones with the information of the surface pollution source that needs to be collaboratively monitored, forming a drone collaborative monitoring group. The control center uses the swarm intelligence optimization algorithm to dynamically optimize the task allocation within the collaborative group. The specific steps are as follows: S3251. Each drone that receives the coordination signal uploads its status information to the control center, and the control center introduces a coordination factor C. i , this factor represents the UAV U i The importance of collaborative monitoring, the design formula is as follows: Among them, d i，κ Indicates drone U i Virtual point source with area source that needs to be monitored in coordination The distance; γ i Indicates drone U i The number of remaining unmonitored pollution sources, β i Indicates the current task completion progress. When 0<β i When <1, it means that the drone is using sensors to monitor the gas concentration of the current pollution source, β i The larger the value, the closer the task is to completion. i =1, indicating that the UAV has completed the task of the current monitoring point and is ready to fly to the next monitoring point, or is in flight; S3252, the control center calculates the C of all candidate drones i , and sort the candidate drones according to the size of the synergy factor, gradually select the drones with larger synergy factors from the candidate drones to join the drone collaborative monitoring group group, remove the subsequent monitoring tasks of the drone, and add it to the unassigned pollution source set Φ unmonitored , and dynamically assess the capabilities of the entire collaborative monitoring group A group , until A group The time window is no larger than the surface source Thus, the appropriate number of drones B is determined. Design and Assessment of the Collaborative Monitoring Team Capability A group The formula is as follows: where λ i is the mission load factor, which is determined by the UAV U i The ratio of the synergistic factors to the total number of the group is determined by A i，κ Indicates drone U i Ability to monitor area sources independently after planning a path using an insertion heuristic algorithm; S3253, the control center calls the group intelligent optimization algorithm in two stages to assign monitoring points to the monitoring group group drones: in the first stage, the task of the monitoring group is to take sampling points within the surface source, and the drone starts from the position where the coordinated signal is received; in the second stage, the task of the monitoring group is to take sampling points within the surface source, and the drone starts from the position where the coordinated signal is received; unmonitored , the UAV departs from the location where the surface source monitoring is completed.

9. A swarm intelligence multi-UAV air pollution monitoring method based on data security according to claim 5 or 8, characterized in that: The insertion heuristic algorithm is as follows: by trying to insert each monitoring point in the surface source into different positions of the current path, calculating the optimization target F after insertion, and recording the insertion position of the minimum optimization target, updating the path, and iterating continuously until all monitoring points are inserted to obtain the final monitoring access sequence. The optimization target F and constraints of the algorithm are as follows: Objective function F: Constraints: Energy constraints, each UAV U i The remaining energy E at any time t i (t) cannot be negative; Location constraints: the drone needs to return to a nearby base station after completing the mission; Full coverage constraint: all pollution source monitoring points in the industrial park must be monitored; No duplication constraint, each pollution source monitoring point is monitored by only one drone once, that is, there is no intersection between the monitoring point visit sequences of any two drones; Data consistency constraints, once the drone U i Start monitoring a sampling point of non-point source pollution, and all monitoring sampling points must be within the time window Complete monitoring within Hover constraint: the drone’s hovering time during monitoring Not less than the monitoring time required for the pollution source The monitoring period cannot be interrupted, that is, a monitoring point is either monitored once or not; Among them, L i Indicates drone U i The total path length; H i Indicates drone U i The flight altitude stability is set as the average value of the altitude change in the flight path; Θ i Indicates drone U i The average value of the required turning angle during flight.

Images