Ambient air monitoring method and system based on unmanned aerial vehicle mobile networking

Abstract

The invention discloses an ambient air monitoring method and system based on unmanned aerial vehicle mobile networking, and the method comprises the steps: collecting the observation data of each unmanned aerial vehicle, and actively receiving and fusing a neighbor unmanned aerial vehicle model state set, thereby forming a rich first unmanned aerial vehicle local information set; based on the local information set and the current sensor calibration parameters, the system performs joint posterior probability modeling on the prior field model to obtain the local posterior belief of the unmanned aerial vehicle to the current environment; on the basis, gradient optimization is further carried out on the calibration parameters of the sensor, self-adaptive calibration of the sensor is achieved, and finally calibrated observation data with high accuracy are output. Through the mode, each unmanned aerial vehicle can autonomously optimize the sensing ability in a dynamic environment, and effectively cope with complex environmental changes, so that the intelligent level of environmental air quality monitoring is comprehensively improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent monitoring, and more particularly, to an environment air monitoring method and system based on unmanned aerial vehicle mobile networking. BACKGROUND

[0002] It is particularly important to construct an efficient, accurate and real-time environment air monitoring scheme in order to timely find pollution sources, assess pollution levels and provide scientific basis for environmental governance. Although the traditional fixed monitoring station has accurate data, its coverage is limited, which is difficult to meet the dynamic monitoring needs of large areas or sudden pollution events, especially in scenarios requiring rapid response and flexible deployment, its limitations are particularly prominent.

[0003] In order to overcome the limitations of traditional monitoring methods, environment air monitoring technology based on unmanned aerial vehicles has emerged as the times require, and has attracted much attention due to its strong mobility and flexible deployment. However, the existing unmanned aerial vehicle mobile networking monitoring scheme still faces many technical challenges. One of the core contradictions is the conflict between network connectivity and detection coverage in resource allocation. In order to maintain strong network connectivity between unmanned aerial vehicles and guarantee bandwidth, low delay and high reliability of data transmission, unmanned aerial vehicles usually need to keep a close distance; but this will limit the spatial dispersion of unmanned aerial vehicles, thus sacrificing the wide detection coverage; on the contrary, if you want to achieve wider detection coverage, unmanned aerial vehicles need to fly dispersedly, which is easy to cause the attenuation or even interruption of communication link, forming an "island" effect, so that critical data cannot be transmitted and shared in real time. In addition, in complex environments such as urban canyons and mountainous areas, unmanned aerial vehicles also need to deal with the problems of autonomous obstacle avoidance and communication signal shielding. The communication link is easy to be shielded and interrupted, which threatens flight safety and task continuity. Existing technologies often cannot effectively solve this dynamic contradiction, resulting in the inability to accurately integrate and calibrate monitoring data, thereby affecting the efficiency and data reliability of the entire monitoring system. Especially in the multi-unmanned aerial vehicle cooperation scene, there is a lack of a mechanism that can effectively integrate neighbor unmanned aerial vehicle state information and adaptively calibrate its own sensor data, making the monitoring result vulnerable to environmental changes and sensor drift, further reducing data accuracy and reliability.

[0004] Therefore, an optimized environment air monitoring scheme based on unmanned aerial vehicle mobile networking is expected. SUMMARY

[0005] To solve the above technical problems, the present application is proposed. Embodiments of the present application provide an environment air monitoring method and system based on unmanned aerial vehicle mobile networking, wherein each unmanned aerial vehicle not only collects its own observation data, but also actively receives and fuses a neighbor unmanned aerial vehicle model state set, thereby forming a rich first unmanned aerial vehicle local information set; based on the local information set and a current sensor calibration parameter, the system models a joint posterior probability of a prior field model to obtain a local posterior belief of the unmanned aerial vehicle on the current environment; on this basis, the sensor calibration parameter is further optimized by gradient, realizing adaptive calibration of the sensor, and finally outputting highly accurate calibrated observation data. In this way, each unmanned aerial vehicle can autonomously optimize its perception ability in a dynamic environment, effectively coping with complex environmental changes, thereby comprehensively improving the intelligent level of environment air quality monitoring.

[0006] According to one aspect of the present application, an environment air monitoring method based on unmanned aerial vehicle mobile networking is provided, which comprises: collecting first unmanned aerial vehicle observation data and receiving a neighbor unmanned aerial vehicle model state set; combining the first unmanned aerial vehicle observation data and the neighbor unmanned aerial vehicle model state set to obtain a first unmanned aerial vehicle local information set; based on the first unmanned aerial vehicle local information set and a first unmanned aerial vehicle sensor calibration parameter, modeling a joint posterior probability of a prior field model to obtain a first unmanned aerial vehicle local posterior belief; based on the first unmanned aerial vehicle local information set and the first unmanned aerial vehicle local posterior belief, performing calibration parameter gradient optimization on the first unmanned aerial vehicle sensor calibration parameter to obtain an updated first unmanned aerial vehicle sensor calibration parameter; based on the updated first unmanned aerial vehicle sensor calibration parameter, calibrating the first unmanned aerial vehicle observation data to obtain calibrated first unmanned aerial vehicle observation data.

[0007] According to another aspect of the present application, an environment air monitoring system based on unmanned aerial vehicle mobile networking is provided, which comprises: an information collection module for collecting first unmanned aerial vehicle observation data and receiving a neighbor unmanned aerial vehicle model state set; an unmanned aerial vehicle local information acquisition module for combining the first unmanned aerial vehicle observation data and the neighbor unmanned aerial vehicle model state set to obtain a first unmanned aerial vehicle local information set; a joint posterior probability modeling module for modeling a joint posterior probability of a prior field model based on the first unmanned aerial vehicle local information set and a first unmanned aerial vehicle sensor calibration parameter to obtain a first unmanned aerial vehicle local posterior belief; a calibration parameter gradient optimization module, configured to perform calibration parameter gradient optimization on the first unmanned aerial vehicle sensor calibration parameter based on the first unmanned aerial vehicle local information set and the first unmanned aerial vehicle local posterior belief to obtain an updated first unmanned aerial vehicle sensor calibration parameter; a first unmanned aerial vehicle observation data calibration module, configured to calibrate the first unmanned aerial vehicle observation data based on the updated first unmanned aerial vehicle sensor calibration parameter to obtain calibrated first unmanned aerial vehicle observation data.

[0008] Compared with the prior art, the method and system for monitoring environment air based on unmanned aerial vehicle mobile networking provided by the application, wherein each unmanned aerial vehicle not only collects its own observation data, but also actively receives and fuses the neighbor unmanned aerial vehicle model state set, thereby forming a rich first unmanned aerial vehicle local information set; based on the local information set and the current sensor calibration parameter, the system performs joint posterior probability modeling on the prior field model to obtain the local posterior belief of the unmanned aerial vehicle on the current environment; on this basis, the sensor calibration parameter is further gradient optimized to realize adaptive calibration of the sensor, and finally highly accurate calibrated observation data is output. In this way, each unmanned aerial vehicle can autonomously optimize its perception ability in a dynamic environment, effectively cope with complex environmental changes, and thus comprehensively improve the intelligent level of environment air quality monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0009] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description of embodiments of the present application taken in conjunction with the accompanying drawings. The drawings provided in the specification and the embodiments of the present application are only for further illustrating the present application and are not intended to limit the scope of the present application. In the drawings, the same reference numerals generally indicate the same components or steps throughout the specification and the drawings.

[0010] Figure 1 a flow chart of the method for monitoring environment air based on unmanned aerial vehicle mobile networking according to the embodiments of the present application; Figure 2 a data flow schematic diagram of the method for monitoring environment air based on unmanned aerial vehicle mobile networking according to the embodiments of the present application; Figure 3 a block diagram of the system for monitoring environment air based on unmanned aerial vehicle mobile networking according to the embodiments of the present application. DETAILED DESCRIPTION

[0011] Hereinafter, example embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, and not all the embodiments of the present application, and it should be understood that the present application is not limited to the example embodiments described herein.

[0012] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0013] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0014] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0015] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0016] The technical solution of this application proposes an ambient air monitoring method based on UAV mobile networking. Figure 1 This is a flowchart of an ambient air monitoring method based on UAV mobile networking according to an embodiment of this application. Figure 2 This is a system architecture diagram of an ambient air monitoring method based on UAV mobile networking according to an embodiment of this application. Figure 1 and Figure 2 As shown, the ambient air monitoring method based on UAV mobile networking according to an embodiment of this application includes the following steps: S1, collecting observation data from a first UAV and receiving a set of neighboring UAV model states; S2, combining the first UAV observation data and the set of neighboring UAV model states to obtain a local information set of the first UAV; S3, performing joint posterior probability modeling on the prior field model based on the local information set of the first UAV and the sensor calibration parameters of the first UAV to obtain the local posterior belief of the first UAV; S4, performing calibration parameter gradient optimization on the sensor calibration parameters of the first UAV based on the local information set of the first UAV and the local posterior belief of the first UAV to obtain updated sensor calibration parameters of the first UAV; S5, calibrating the first UAV observation data based on the updated sensor calibration parameters of the first UAV to obtain calibrated first UAV observation data.

[0017] In particular, the S1 collects first unmanned aerial vehicle observation data and receives a neighbor unmanned aerial vehicle model state set. It should be understood that conventional environmental air monitoring solutions are difficult to provide continuous and high-precision monitoring data in a large range or a complex dynamic environment. By collecting the observation data of the first unmanned aerial vehicle itself, detailed environmental information at a specific location can be obtained; at the same time, receiving the neighbor unmanned aerial vehicle model state set can realize distributed sensing, make up for the limitations of the single unmanned aerial vehicle perspective, and effectively cope with technical challenges such as resource allocation conflicts between network connectivity and detection coverage. The fusion of such multi-source data is a key prerequisite for realizing collaborative monitoring and subsequent intelligent calibration, which can significantly improve the environmental sensing capability and data reliability of the entire system.

[0018] Among them, the first unmanned aerial vehicle observation data refers to the environmental air quality parameters and related spatio-temporal information obtained by the sensors (such as PM2.5, SO2, temperature, etc.) carried by the first unmanned aerial vehicle in the flight process; each neighbor unmanned aerial vehicle model state in the neighbor unmanned aerial vehicle model state set includes a timestamp, a spatial coordinate, an original sensor reading, and a current sensor calibration parameter. Specifically, the set contains the data generation time point (timestamp) of each neighbor unmanned aerial vehicle, its accurate geographic position in three-dimensional space (spatial coordinate), the initial measurement data of the sensor without calibration (original sensor reading), and the calibration gain and offset used by the current sensor of the neighbor unmanned aerial vehicle (current sensor calibration parameter).

[0019] In specific implementation, first, the first unmanned aerial vehicle observation data is collected: the first unmanned aerial vehicle is equipped with various environmental air sensors, and when performing a monitoring task, these sensors will continuously work, convert the detected environmental air quality parameters (such as PM2.5 concentration) into electrical signals, and process them through the on-board data acquisition module, finally generating digitized observation data containing a timestamp and its own spatial coordinate; second, the neighbor unmanned aerial vehicle model state set is received: the first unmanned aerial vehicle establishes a communication connection with the neighbor unmanned aerial vehicles within the communication range through the wireless communication module carried by it. The neighbor unmanned aerial vehicles periodically broadcast or send their own model state information according to the request. After receiving these data packets, the first unmanned aerial vehicle will parse them and extract the key state information of each neighbor unmanned aerial vehicle, thereby forming the neighbor unmanned aerial vehicle model state set. In this way, the first unmanned aerial vehicle can dynamically obtain the local sensing data of the surrounding environment and the calibration information of the neighbor unmanned aerial vehicles, providing comprehensive input for subsequent joint data processing.

[0020] In particular, S2 combines the first UAV observation data and the neighbor UAV model state set to obtain the first UAV local information set. That is, by combining the first UAV observation data and the neighbor UAV model state set, the real-time perception capability of the first UAV itself and the distributed monitoring information of the neighboring UAVs are integrated to form a more comprehensive and multi-dimensional local environment view, thereby providing a solid data foundation for subsequent joint posterior probability modeling and sensor adaptive calibration. Through this combination, the limitations of limited monitoring coverage and possible local bias of single UAV monitoring data can be effectively overcome. The obtained local information set essentially integrates the observation data of the first UAV itself with the data of all neighbor UAVs in the similar time period after their own preliminary processing and calibration, forming a comprehensive description of the local environment state. This rich information set can support more intelligent decision-making and improve the accuracy and reliability of overall monitoring.

[0021] In specific implementation, after the first UAV completes the collection of its own observation data and receives the model state set from the neighbor UAVs, the system starts the data combination module to integrate these heterogeneous but interrelated data into a unified first UAV local information set. In this process, first, the observation data of the first UAV itself, including time stamp, spatial coordinates, and original sensor readings, etc., are taken as the basic part of the information set; second, for the received neighbor UAV model state set, the module iterates through each neighbor UAV in the set; finally, by logically merging or set operation of the first UAV's own data and the state data of all neighbor UAVs, the first UAV local information set is obtained. This combination ensures the integrity and consistency of the information set, which integrates all relevant observation points and their associated calibration information into a data structure for subsequent processing.

[0022] In particular, S3 models the joint posterior probability of the prior field model based on the first UAV local information set and the first UAV sensor calibration parameters to obtain the first UAV local posterior belief. That is, by combining the latest observation data obtained by the first UAV and its neighbor UAVs with the known sensor characteristics of the first UAV, a more accurate and less uncertain local environment model is generated from the preset environmental prior knowledge. This posterior belief not only provides the best estimate of the current environment field state, but also provides an important probability basis for subsequent optimization of the first UAV sensor calibration parameters, effectively solving the problem that single observation data or fixed model cannot accurately reflect the dynamic changes of complex environment.

[0023] The prior field model refers to an initial assumption or rough estimate of the environmental air quality distribution, which can be based on historical data, geographic information or physical laws, and represents the environmental field cognition before any new observation data. The first unmanned aerial vehicle local posterior belief refers to the updated probability distribution of the first unmanned aerial vehicle on the local environmental field state after the latest observation data (i.e. the first unmanned aerial vehicle local information set) and the first unmanned aerial vehicle sensor calibration parameter correction. This belief contains the estimated value (mean) and uncertainty (variance) of the environmental field at each spatial point, which integrates prior knowledge and actual observation, thereby providing more reliable environmental perception.

[0024] In specific implementation, first, based on the first unmanned aerial vehicle local information set and the first unmanned aerial vehicle sensor calibration parameter, a training input matrix and a training output vector are constructed. The training input matrix contains the spatial coordinates of all observation points in the first unmanned aerial vehicle local information set. For each data point in the local information set, whether it comes from the first unmanned aerial vehicle itself or a neighbor unmanned aerial vehicle, its spatial coordinates will be a row. The training output vector corresponds to the calibrated sensor readings of these observation points. Specifically, for each original sensor reading in the local information set, the sensor calibration parameter of the unmanned aerial vehicle to which the observation point belongs is used for preliminary calibration. If the observation point comes from the first unmanned aerial vehicle, the current calibration parameter of the first unmanned aerial vehicle is used for calibration; if it comes from a neighbor unmanned aerial vehicle, the calibration parameter reported by the neighbor unmanned aerial vehicle is used for calibration. The calibrated readings constitute the corresponding elements of the training output vector. This process ensures that all observation data input into the model are comparable and reflect the current calibration state of the sensor; Then, based on the training input matrix and the prior field model, an observation covariance matrix is constructed. The observation covariance matrix is a square matrix, whose diagonal elements represent the total variance (including signal variance and measurement noise variance) of each observation data point, and the non-diagonal elements represent the covariance between different observation data points, reflecting their spatial correlation. The observation covariance matrix can accurately quantify the correlation (i.e. spatial correlation) between each observation data point, and at the same time consider the inherent random error in the sensor measurement process. By constructing this matrix, the structure and uncertainty of the observation data can be fully understood, which is crucial for subsequent accurate estimation of the environmental field state, optimization of the sensor calibration parameter and reliable probability inference. It ensures that the model can effectively distinguish between real environmental changes and measurement noise, thereby avoiding model bias caused by data uncertainty.

[0025] In this process, firstly, the kernel function and measurement noise variance are extracted from the prior field model. Before performing joint posterior probability modeling, the system pre-defines a prior field model, which embeds assumptions about the statistical properties of the environmental field. In this sub-step, two key parameters are obtained from this prior field model: the kernel function and the measurement noise variance. The kernel function is used to describe the correlation between any two points in space. In the embodiments of this application, the kernel function is expressed by the formula: , in, Two coordinate points and The square of the Euclidean distance between them For signal variance, For length scale; Next, based on the training input matrix and the kernel function, the kernel matrix is ​​calculated. Once the kernel function is determined, the system uses the coordinates of all observation points contained in the training input matrix to calculate the kernel matrix. Assume the training input matrix X contains N observation points, i.e. Then the kernel matrix K will be a A symmetric matrix, where each element Represents the i-th observation point and the j-th observation point The kernel matrix captures the spatial correlation between all pairs of observation points. Points that are closer together typically have more similar environmental characteristics, resulting in a larger covariance value; points that are farther apart have a weaker correlation, resulting in a smaller covariance value. Furthermore, the measurement noise variance is added to the diagonal of the kernel matrix to obtain the observation covariance matrix. That is, after obtaining the kernel matrix K, which purely reflects spatial correlation, it is also necessary to consider the inherent noise of each sensor measurement process, independent of environmental field variations. This noise is typically modeled as independent, identically distributed Gaussian noise with a variance of... To incorporate it into the covariance matrix, the measurement noise variance needs to be added to the diagonal of the kernel matrix K. The final matrix obtained is the observation covariance matrix, and this process can be expressed by the formula: , in, It is The identity matrix is ​​obtained by adding the measurement noise variance to the diagonal, and the observation covariance matrix is ​​obtained. It not only includes the spatial correlation between observation points, but also considers the measurement uncertainty of each observation data point itself, thus providing a more comprehensive and realistic description of the covariance of the observation data.

[0026] Further, based on the training input matrix, the training output vector and the observation covariance matrix, a first local posterior belief of the unmanned aerial vehicle is determined. That is, in the technical solution of the present application, the prior environmental field model is updated and corrected by fusing all the acquired observation data (training input matrix and training output vector) and its inherent uncertainty (observation covariance matrix), so that the first unmanned aerial vehicle obtains the most accurate and most comprehensive probabilistic understanding of the local environmental state. Such local posterior belief not only provides the best estimated value of the environmental parameter, but more importantly, quantifies the uncertainty of the estimated value, which has irreplaceable value for guiding the subsequent intelligent decision of the unmanned aerial vehicle, optimizing the sensor calibration parameter and realizing more refined environmental perception, effectively solving the problem that single observation data or rough model cannot accurately reflect the dynamic change of complex environment.

[0027] wherein the first local posterior belief of the unmanned aerial vehicle refers to an updated probability distribution of the local environmental field state formed by the first unmanned aerial vehicle through Bayesian inference. Such belief is usually represented in the form of a mean function and a covariance function. In specific implementation, it can be realized based on a Gaussian process regression (GPR) mathematical framework. After obtaining the training input matrix X, the training output vector y and the observation covariance matrix , the system will use this information to calculate the posterior mean and the posterior variance of the environmental parameter of any interested unknown prediction point . This process can be regarded as a process of updating the belief of the environmental field function (such as the air pollutant concentration distribution) by combining the prior knowledge (defined by the kernel function and the measurement noise variance) with the actual observation data under the Bayesian framework. Specifically, for any spatial point to be predicted, the posterior mean of the environmental parameter of the prediction point is calculated according to the following formula: , wherein is the inverse matrix of the observation covariance matrix , is a row vector, and the jth element of the row vector is the covariance between the prediction point and the jth observation point in the training input matrix X. The formula propagates the influence of the training data to the prediction point by weighted average, and the weight is determined by the spatial correlation; Meanwhile, the calculation formula of the posterior variance of the prediction point is as follows: , wherein is the covariance of the prediction point with itself, reflecting the inherent variance of the prediction point without any observation data, is a column vector. The formula shows that the uncertainty of the prediction point is the amount by which the uncertainty of the prediction point is reduced on the basis of its prior uncertainty through the information provided by the observation data. By calculating the two values, the first unmanned vehicle obtains a complete probability description of the state of the local environment field, i.e., the local posterior belief of the first unmanned vehicle.

[0028] In particular, the S4 calibrates the sensor calibration parameters of the first unmanned vehicle based on the local information set of the first unmanned vehicle and the local posterior belief of the first unmanned vehicle to obtain updated sensor calibration parameters of the first unmanned vehicle. That is, in the technical solution of the present application, the sensor calibration parameters of the first unmanned vehicle are dynamically adjusted and optimized by using all currently available local observation data and the latest probabilistic understanding of the environment field. In this way, the system can effectively correct the drift and deviation of the sensors due to aging, environmental changes, or initial calibration errors, thereby ensuring that the subsequent observation data output by the first unmanned vehicle is closer to the true environment state, significantly improving the reliability and accuracy of the monitoring results, and coping with complex and variable environmental monitoring challenges.

[0029] wherein the sensor calibration parameters of the first unmanned vehicle refer to gain and offset parameters used to correct the original readings of the sensors of the first unmanned vehicle, which convert the original signals into more accurate physical quantities; the calibration parameter gradient optimization is an iterative algorithm, and the goal is to find a set of optimal calibration parameters so that the observation data has the maximum possibility under the given observation data and environment field posterior belief.

[0030] In specific implementation, first, a log marginal likelihood function is calculated based on the training output vector and the observation covariance matrix to obtain a log marginal likelihood value. The log marginal likelihood function is an index for measuring the goodness of fit of a model to data in statistics, and in the specific examples of the present application, the higher the value, the stronger the explanatory ability of the model under the current calibration parameters to the observation data. In this process, the system calculates the log marginal likelihood value of the environment field model under the current calibration parameters by using the constructed training output vector y (containing the observation data after preliminary calibration) and the observation covariance matrix (reflecting the spatial correlation of the observation data and the measurement noise), and the log marginal likelihood value of the environment field model under the current calibration parameters is calculated. The log marginal likelihood function is a commonly used model selection and parameter optimization objective function in Gaussian process regression, and its expression is usually as follows: , wherein, is the number of observation points, represents all hyperparameters in the model, including the sensor calibration parameters of the current first unmanned vehicle. The log marginal likelihood value quantifies the consistency of the observation data with the Gaussian process model assumption under the current sensor calibration parameters; Then, based on the training output vector, the observation covariance matrix and the first UAV local information set, the gradient of the likelihood function with respect to the calibration parameters of the first UAV sensor calibration parameters is calculated to obtain the first UAV gain gradient and the first UAV gain offset. The first UAV gain offset refers to the partial derivative of the log marginal likelihood function with respect to the first UAV gain and offset parameters, indicating the direction and amplitude of adjusting these parameters to increase the likelihood value. Since the sensor calibration parameters of the first UAV directly affect the calibration results of its own observation data, and further affect part of the training output vector , and may indirectly affect the structure of the observation covariance matrix (e.g., through measurement noise variance). Therefore, the partial derivatives of the log marginal likelihood function with respect to and need to be calculated; wherein the partial derivatives, i.e., the gain gradient and the gain offset, indicate how to adjust and to make the log marginal likelihood value increase the fastest. For example, if the original observation data of the first UAV is , then the term of its calibrated contribution to the training output vector is expressed by the formula: , Here, the calculation of the gradient involves the chain rule, combining the derivative of the log likelihood with respect to y with the derivative of y with respect to and ; Further, based on the first UAV gain gradient and the first UAV gain offset, the calibration parameter update is performed on the first UAV sensor calibration parameters to obtain the updated first UAV sensor calibration parameters. Specifically, the calibration parameter update is performed on the first UAV sensor calibration parameters by the following formula: , , wherein and are the first UAV gain gradient and the first UAV gain offset, and are the first UAV sensor calibration parameters, and are the updated first UAV sensor calibration parameters. By continuously adjusting the parameters in the direction of the gradient, the sensor calibration parameters will gradually converge to the optimal value that can make the observation data and the environment field posterior belief most consistent. This process will continue until the log marginal likelihood value converges or reaches a preset number of iterations.

[0031] In particular, the S5 calibrates the first UAV observation data based on the updated first UAV sensor calibration parameter to obtain calibrated first UAV observation data. That is, the optimal sensor calibration parameter obtained by gradient optimization in the previous step is directly applied to the original observation data newly collected by the first UAV. By performing this step, the inherent systematic errors of the sensor (such as gain drift and zero offset) can be effectively eliminated or weakened, the original measurement signal that may have deviations is converted into data that is closer to the true value of the environment and has clear physical meaning, thereby ensuring the accuracy, reliability and consistency of the output results of the entire monitoring system.

[0032] wherein the updated first UAV sensor calibration parameter refers to the newly obtained gain parameter and offset parameter obtained by iterative calculation in the previous calibration parameter gradient optimization step, which represent the best estimation of the sensor characteristics under the current environment; the calibrated first UAV observation data is the final output of this step, which is obtained by applying the updated calibration parameter to correct the original observation data, and is considered to be the most accurate measurement value of the environmental air quality parameter at the current time.

[0033] In a specific implementation, the first UAV observation data is calibrated based on the updated first UAV sensor calibration parameter according to the following formula: wherein, is the measurement noise. is the first UAV observation data. By performing this calculation, the original sensor reading can be converted into accurately calibrated and reliable environmental monitoring data.

[0034] In summary, the environmental air monitoring method based on UAV mobile networking according to the embodiments of the present application is illustrated, wherein each UAV not only collects its own observation data, but also actively receives and fuses the neighbor UAV model state set, thereby forming a rich first UAV local information set; based on the local information set and the current sensor calibration parameter, the system performs joint posterior probability modeling on the prior field model to obtain the local posterior belief of the UAV on the current environment; on this basis, the sensor calibration parameter is further gradient-optimized to realize adaptive calibration of the sensor, and finally highly accurate calibrated observation data is output. In this way, each UAV can autonomously optimize its perception ability in a dynamic environment, effectively cope with complex environmental changes, and thus comprehensively improve the intelligent level of environmental air quality monitoring.

[0035] Further, an environmental air monitoring system based on UAV mobile networking is also provided.

[0036] Figure 3 ​A block diagram of an environment air monitoring system based on mobile networking of UAVs according to an embodiment of the present application. As shown in Figure 3 The environment air monitoring system based on mobile networking of UAVs 300 according to an embodiment of the present application comprises: an information collection module 310, configured to collect first UAV observation data and receive a neighbor UAV model state set; a UAV local information acquisition module 320, configured to combine the first UAV observation data and the neighbor UAV model state set to obtain a first UAV local information set; a joint posterior probability modeling module 330, configured to perform joint posterior probability modeling on a prior field model based on the first UAV local information set and a first UAV sensor calibration parameter to obtain a first UAV local posterior belief; a calibration parameter gradient optimization module 340, configured to perform calibration parameter gradient optimization on the first UAV sensor calibration parameter based on the first UAV local information set and the first UAV local posterior belief to obtain an updated first UAV sensor calibration parameter; and a UAV observation data calibration module 350, configured to calibrate the first UAV observation data based on the updated first UAV sensor calibration parameter to obtain calibrated first UAV observation data.

[0037] As described above, the environment air monitoring system based on mobile networking of UAVs 300 according to an embodiment of the present application can be implemented in various wireless terminals, such as a server with an environment air monitoring algorithm based on mobile networking of UAVs, and the like. In one possible implementation, the environment air monitoring system based on mobile networking of UAVs 300 according to an embodiment of the present application can be integrated into a wireless terminal as a software module and / or a hardware module. For example, the environment air monitoring system based on mobile networking of UAVs 300 can be a software module in an operating system of the wireless terminal, or can be an application program developed for the wireless terminal; of course, the environment air monitoring system based on mobile networking of UAVs 300 can also be one of many hardware modules of the wireless terminal.

[0038] Alternatively, in another example, the environment air monitoring system based on mobile networking of UAVs 300 and the wireless terminal can also be separate devices, and the environment air monitoring system based on mobile networking of UAVs 300 can be connected to the wireless terminal through a wired and / or wireless network, and transmit interactive information in an agreed data format.

[0039] Having described above several embodiments of the disclosure, any modifications and variations that fall within the scope of the described embodiments are also intended to be within the scope of the disclosure. As will be apparent to those skilled in the art, some modifications and variations to the embodiments described above can be practiced while staying within the scope and spirit of the described embodiments. The foregoing description of the described embodiments has been presented for the purposes of illustration and description. It is not intended to be exhaustive or to limit the described embodiments to the precise form disclosed. Many modifications and variations are possible in light of the above teachings. It is intended that the disclosed embodiments be limited only by the claims.

Claims

1.A method for environmental air monitoring based on unmanned aerial vehicle mobile networking, characterized in that, The method comprises: collecting first unmanned aerial vehicle observation data and receiving a set of neighbor unmanned aerial vehicle model states; combining the first unmanned aerial vehicle observation data and the set of neighbor unmanned aerial vehicle model states to obtain a first unmanned aerial vehicle local information set; based on the first unmanned aerial vehicle local information set and first unmanned aerial vehicle sensor calibration parameters, modeling a joint posterior probability of a prior field model to obtain a first unmanned aerial vehicle local posterior belief; based on the first unmanned aerial vehicle local information set and the first unmanned aerial vehicle local posterior belief, performing calibration parameter gradient optimization on the first unmanned aerial vehicle sensor calibration parameters to obtain updated first unmanned aerial vehicle sensor calibration parameters; based on the updated first unmanned aerial vehicle sensor calibration parameters, calibrating the first unmanned aerial vehicle observation data to obtain calibrated first unmanned aerial vehicle observation data. 2.The method of claim 1, wherein, Each neighbor unmanned aerial vehicle model state in the set of neighbor unmanned aerial vehicle model states comprises a timestamp, a spatial coordinate, an original sensor reading, and a current sensor calibration parameter. 3.The method of claim 2, wherein, Based on the first unmanned aerial vehicle local information set and the first unmanned aerial vehicle sensor calibration parameters, modeling a joint posterior probability of a prior field model to obtain a first unmanned aerial vehicle local posterior belief, comprising: based on the first unmanned aerial vehicle local information set and the first unmanned aerial vehicle sensor calibration parameters, constructing a training input matrix and a training output vector; based on the training input matrix and the prior field model, constructing an observation covariance matrix; based on the training input matrix, the training output vector, and the observation covariance matrix, determining the first unmanned aerial vehicle local posterior belief. 4.The method of claim 3, wherein, Based on the training input matrix and the prior field model, constructing an observation covariance matrix, comprising: extracting a kernel function and a measurement noise variance from the prior field model; based on the training input matrix and the kernel function, calculating a kernel matrix; adding the measurement noise variance on the diagonal of the kernel matrix to obtain the observation covariance matrix. 5.The method of claim 4, wherein, The kernel function is: where is the squared Euclidean distance between two coordinate points and , is the signal variance, is the length scale. 6.The method of claim 3, wherein, Based on the first unmanned aerial vehicle local information set and the first unmanned aerial vehicle local posterior belief, performing calibration parameter gradient optimization on the first unmanned aerial vehicle sensor calibration parameters to obtain updated first unmanned aerial vehicle sensor calibration parameters, comprising: based on the training output vector and the observation covariance matrix, performing log marginal likelihood function calculation to obtain a log marginal likelihood value; based on the training output vector, the observation covariance matrix, and the first unmanned aerial vehicle local information set, performing gradient calculation of a likelihood function with respect to the calibration parameters on the first unmanned aerial vehicle sensor calibration parameters to obtain a first unmanned aerial vehicle gain gradient and a first unmanned aerial vehicle gain offset; based on the first unmanned aerial vehicle gain gradient and the first unmanned aerial vehicle gain offset, performing calibration parameter update on the first unmanned aerial vehicle sensor calibration parameters to obtain updated first unmanned aerial vehicle sensor calibration parameters. 7.The method of claim 6, wherein, Based on the first unmanned aerial vehicle gain gradient and the first unmanned aerial vehicle gain offset, performing calibration parameter update on the first unmanned aerial vehicle sensor calibration parameters to obtain updated first unmanned aerial vehicle sensor calibration parameters, comprising: performing calibration parameter update on the first unmanned aerial vehicle sensor calibration parameters according to the following formula: wherein, and are a first drone gain gradient and a first drone gain offset, and are first drone sensor calibration parameters, and are updated first drone sensor calibration parameters. 8.The method of claim 1, wherein, Based on the updated first unmanned aerial vehicle sensor calibration parameters, calibrating the first unmanned aerial vehicle observation data to obtain calibrated first unmanned aerial vehicle observation data, comprising: calibrating the first unmanned aerial vehicle observation data according to the following formula: wherein, to measure noise. is first drone observation data. 9.A system for environmental air monitoring based on mobile networking of unmanned aerial vehicles, characterized in that, Comprise: An information collection module for collecting first unmanned aerial vehicle observation data and receiving a neighbor unmanned aerial vehicle model state set; An unmanned aerial vehicle local information acquisition module for combining the first unmanned aerial vehicle observation data and the neighbor unmanned aerial vehicle model state set to obtain a first unmanned aerial vehicle local information set; A joint posterior probability modeling module for modeling a prior field model based on the first unmanned aerial vehicle local information set and the first unmanned aerial vehicle sensor calibration parameter to obtain a first unmanned aerial vehicle local posterior belief; A calibration parameter gradient optimization module for optimizing the first unmanned aerial vehicle sensor calibration parameter based on the first unmanned aerial vehicle local information set and the first unmanned aerial vehicle local posterior belief to obtain updated first unmanned aerial vehicle sensor calibration parameter; An unmanned aerial vehicle observation data calibration module for calibrating the first unmanned aerial vehicle observation data based on the updated first unmanned aerial vehicle sensor calibration parameter to obtain calibrated first unmanned aerial vehicle observation data.

Images