Unmanned aerial vehicle flight control method based on multi-sensor data fusion

Abstract

The invention provides an unmanned aerial vehicle flight control method and device based on multi-sensor data fusion, and the method comprises the following steps: carrying out the denoising and clustering weight fusion processing of flight control initial data collected by multiple sensors, and generating fusion data; performing real-time state estimation on the fused data by adopting an extended state observer to obtain real-time flight state parameters; performing deviation compensation based on a deviation value between a preset target state and a real-time flight state parameter, optimizing a fuzzy adaptive PID control algorithm of parameters through a genetic algorithm, and dynamically adjusting the parameters of a control machine in a global optimization mode; and outputting the optimal control quantity to realize accurate flight control of the unmanned aerial vehicle. According to the method, the data confidence is improved through a clustering weighting mechanism, parameter global optimization is carried out in combination with disturbance observation of an extended state observer and a fuzzy adaptive PID control algorithm, and the problem that the flight control precision is insufficient due to the fact that a current unmanned aerial vehicle cannot realize control parameter global optimization is effectively solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a UAV flight control method based on multi-sensor data fusion. BACKGROUND

[0002] The flight control performance of a UAV directly affects the accuracy, stability and reliability of its task execution, and is the core part of UAV technology. In order to obtain more comprehensive and accurate flight state information, multi-sensor fusion technology is widely used in UAV flight control systems, which integrates data collected by multiple sensors to improve the perception ability of UAV flight state.

[0003] However, at the level of flight control algorithm, fixed control parameters are difficult to adapt to the dynamic changes of UAVs under different flight conditions and task requirements, and the control precision and response speed are limited. Even if a fuzzy adaptive PID control algorithm is used, its parameter adjustment range and optimization ability are insufficient, and it is difficult to achieve global optimal control. Moreover, the traditional method lacks an effective global optimization mechanism and cannot quickly find the optimal control parameters in complex flight environments, resulting in insufficient flight control accuracy of the UAV and difficulty in meeting high-precision and high-stability flight requirements. SUMMARY

[0004] The main purpose of the present application is to provide a UAV flight control method based on multi-sensor data fusion, which aims to overcome the current UAV's inability to achieve global optimization of control parameters, resulting in insufficient flight control accuracy.

[0005] To achieve the above purpose, the present application provides a UAV flight control method based on multi-sensor data fusion, comprising the following steps:

[0006] Denoising the UAV flight data collected by multiple sensors to obtain pure data; clustering and weighting fusion processing is performed on the pure data to obtain fusion data;

[0007] Performing state estimation processing on the fusion data based on an extended state observer to obtain real-time UAV flight state parameters;

[0008] Based on the preset target flight state parameters, the real-time UAV flight state parameters are compensated for deviation to obtain a compensated deviation value;

[0009] According to the compensated deviation value, a fuzzy adaptive PID control algorithm optimized by a genetic algorithm is used to globally optimize the parameters of the fuzzy adaptive PID to obtain the optimal UAV flight control amount;

[0010] Based on the optimal UAV flight control amount, the UAV flight is controlled.

[0011] Further, the unmanned aerial vehicle flight data collected by the multiple sensors is denoised to obtain pure data, including:

[0012] The unmanned aerial vehicle flight data collected by the multiple sensors is analyzed for correlation based on mutual information entropy to obtain a data correlation degree matrix.

[0013] According to the data correlation degree matrix, the flight data is reduced in dimension and denoised by a sparse representation reconstruction algorithm to obtain the pure data.

[0014] Further, the pure data is processed for clustering and weighted fusion to obtain fused data, including:

[0015] The pure data is input into a long short-term memory network model containing an attention mechanism; the attention mechanism adaptively assigns weights according to the data correlation degree matrix, performs multi-scale feature extraction and weighted processing, and obtains a weighted feature vector.

[0016] The weighted feature vector is analyzed for clustering to obtain multiple data feature clusters.

[0017] For each data feature cluster, an adaptive particle swarm optimization algorithm is used to adjust an inertia weight and a learning factor according to dynamic changes of the feature cluster, perform weight parameter optimization processing, and obtain optimal fusion weights of each cluster.

[0018] According to the optimal fusion weights of each cluster, the pure data is processed for clustering and weighted fusion to obtain fused data.

[0019] Further, the fused data is processed for state estimation based on an extended state observer to obtain real-time flight state parameters of the unmanned aerial vehicle, including:

[0020] The fused data is divided into state variables to obtain a state variable group.

[0021] According to the state variable group, a state space equation containing dynamic characteristics of the unmanned aerial vehicle is constructed to obtain an unmanned aerial vehicle state model.

[0022] An extended state variable is introduced into the unmanned aerial vehicle state model, unmodeled dynamics and external disturbance terms are processed as new state variables, and an extended state space equation is obtained.

[0023] According to the extended state space equation, the fused data is calculated in real time by a preset feedback gain matrix to obtain an observation error.

[0024] Based on the observation error, the feedback gain matrix is updated in parameters to obtain an optimized feedback gain matrix.

[0025] Iterative calculation is performed on the extended state space equation by using the optimized feedback gain matrix, so as to obtain the real-time flight state parameter of the unmanned aerial vehicle.

[0026] Further, the state variable group comprises position, speed, attitude angle and angular velocity.

[0027] Further, deviation compensation is performed on the real-time flight state parameter of the unmanned aerial vehicle based on preset target flight state parameters, so as to obtain a compensated deviation value, comprising:

[0028] State variable extraction is performed on the preset target flight state parameter and the real-time flight state parameter of the unmanned aerial vehicle, so as to obtain a target state vector and a real-time state vector;

[0029] According to the target state vector and the real-time state vector, an H-infinity robust control model is constructed;

[0030] In the H-infinity robust control model, an optimal robust control gain matrix is obtained by solving a minimax optimization problem;

[0031] According to the optimal robust control gain matrix, linear transformation is performed on the deviation between the real-time state vector and the target state vector, so as to obtain a preliminary compensated deviation;

[0032] Stability verification is performed on the preliminary compensated deviation, and when a stability condition is met, the preliminary compensated deviation is taken as a final compensated deviation value.

[0033] Further, according to the compensated deviation value, a fuzzy adaptive PID control algorithm is used to globally optimize parameters of the fuzzy adaptive PID by combining a genetic algorithm, so as to obtain an optimal unmanned aerial vehicle flight control amount, comprising:

[0034] The proportional coefficient, the integral coefficient and the differential coefficient of the fuzzy adaptive PID control algorithm are encoded;

[0035] According to the compensated deviation value at adjacent time points, a deviation change amount is calculated, and a deviation change speed is obtained in combination with a time interval;

[0036] An adaptability evaluation function is constructed by using the compensated deviation value and the deviation change speed, and an initial population is generated;

[0037] Based on the fuzzy adaptive PID control algorithm, selection, crossover and mutation operations are performed on the initial population, and a coding sequence with the highest evaluation value is screened out;

[0038] The coding sequence is decoded, so as to obtain an optimized PID parameter;

[0039] According to the optimized PID parameter and the compensated deviation value, global optimization calculation of the PID parameter is performed to obtain optimal UAV flight control quantity.

[0040] The application further provides a UAV flight control device based on multi-sensor data fusion, comprising:

[0041] A fusion module is configured to denoise UAV flight data collected by the multi-sensor to obtain pure data, and perform clustering and weighted fusion processing on the pure data to obtain fusion data.

[0042] An estimation module is configured to perform state estimation processing on the fusion data based on an extended state observer to obtain real-time UAV flight state parameters.

[0043] A compensation module is configured to compensate the real-time UAV flight state parameters based on preset target flight state parameters to obtain a compensated deviation value.

[0044] An optimization module is configured to perform global optimization on parameters of a fuzzy adaptive PID by combining a fuzzy adaptive PID control algorithm with a genetic algorithm optimization parameter according to the compensated deviation value to obtain optimal UAV flight control quantity.

[0045] A control module is configured to control UAV flight based on the optimal UAV flight control quantity.

[0046] The UAV flight control method based on multi-sensor data fusion provided by the application comprises the following steps: denoising UAV flight data collected by the multi-sensor to obtain pure data; performing clustering and weighted fusion processing on the pure data to obtain fusion data; performing state estimation processing on the fusion data based on an extended state observer to obtain real-time UAV flight state parameters; compensating the real-time UAV flight state parameters based on preset target flight state parameters to obtain a compensated deviation value; performing global optimization on parameters of a fuzzy adaptive PID by combining a fuzzy adaptive PID control algorithm with a genetic algorithm optimization parameter according to the compensated deviation value to obtain optimal UAV flight control quantity; and controlling UAV flight based on the optimal UAV flight control quantity. In the application, the real-time UAV flight state parameters are compensated, and global optimization is performed on parameters of a fuzzy adaptive PID by combining a fuzzy adaptive PID control algorithm with a genetic algorithm optimization parameter to obtain optimal UAV flight control quantity, thereby overcoming the defect that current UAVs cannot realize global optimization of control parameters, resulting in insufficient flight control precision. BRIEF DESCRIPTION OF DRAWINGS

[0047] Figure 1 is a step schematic diagram of the UAV flight control method based on multi-sensor data fusion in an embodiment of the application.

[0048] Figure 2 is a structure block diagram of a UAV flight control device based on multi-sensor data fusion in an embodiment of the present application;

[0049] Figure 3 is a structure schematic block diagram of a UAV in an embodiment of the present application.

[0050] The implementation, functional features and advantages of the present application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0052] Referring to Figure 1 , an embodiment of the present application provides a UAV flight control method based on multi-sensor data fusion, comprising the following steps:

[0053] Step S1, denoising the UAV flight data collected by the multi-sensor to obtain pure data; clustering and weighting fusion processing is performed on the pure data to obtain fusion data;

[0054] Step S2, performing state estimation processing based on an extended state observer on the fusion data to obtain real-time flight state parameters of the UAV;

[0055] Step S3, performing deviation compensation on the real-time flight state parameters of the UAV based on a preset target flight state parameter to obtain a compensated deviation value;

[0056] Step S4, according to the compensated deviation value, performing global optimization on the parameters of the fuzzy adaptive PID control algorithm by combining a fuzzy adaptive PID control algorithm optimized by a genetic algorithm to obtain optimal UAV flight control quantity;

[0057] Step S5, controlling the UAV flight based on the optimal UAV flight control quantity.

[0058] In this embodiment, as described in step S1 above, the original flight data collected by multiple sensors is inevitably disturbed by environmental noise, sensor errors and other factors. If the above noise is not processed, it will seriously affect the accuracy of subsequent data analysis. Through denoising processing, the interference components in the data can be effectively removed, and relatively pure data can be obtained, laying a foundation for subsequent processing. The cluster-weighted fusion processing is based on the fact that the importance and reliability of different sensor data in different flight states differ. By clustering the pure data, data with similar characteristics are classified, and different weights are assigned according to the characteristics and importance of each cluster data, which can more reasonably fuse multi-source data, fully utilize the advantages of each sensor, improve the quality and reliability of the fused data, and make the fused data more accurately reflect the actual flight state of the UAV.

[0059] As described in step S2 above, the flight state of the UAV is affected by many factors, and its dynamic model has complexity and uncertainty, so traditional state estimation methods cannot accurately obtain the real-time flight state of the UAV. The extended state observer can estimate the unmodeled dynamics and external disturbance terms in the system as new state variables, and through analysis and processing of the fused data, the uncertainty of the system can be observed and compensated in real time. Based on the extended state observer, the state estimation process can effectively overcome the problems of inaccurate model and disturbance in the flight process of the UAV, and quickly and accurately obtain the real-time flight state parameters of the UAV such as position, speed, attitude angle, etc., providing reliable state basis for subsequent flight control.

[0060] As described in step S3 above, in actual flight, due to various factors, the real-time flight state of the UAV often deviates from the preset target flight state. Based on the preset target flight state parameters, the deviation of the real-time flight state parameters can be compensated, which can correct the deviation through specific algorithms and strategies according to the uncertainty and disturbance in the flight process of the UAV. This step compensates for the deviation, making the flight state of the UAV closer to the target state, improving the accuracy and stability of the flight control, and providing protection for subsequent precise control.

[0061] As described in the above step S4, the traditional PID control algorithm is not good in control effect when facing complex and changeable flight environment due to fixed parameters. The fuzzy self-adaptive PID control algorithm can adjust the PID parameters according to the deviation and the deviation change rate, but it still has deficiencies in the comprehensiveness and accuracy of parameter optimization. The genetic algorithm is a global optimization algorithm simulating the natural evolution process. In this step, the genetic algorithm is combined with the fuzzy self-adaptive PID control algorithm to search the optimal fuzzy self-adaptive PID parameters in the global range according to the compensation of the post deviation value. Through genetic operations such as selection, crossover and mutation, the PID parameters are continuously optimized, so that the fuzzy self-adaptive PID control algorithm can better adapt to different flight conditions, thereby obtaining the optimal unmanned aerial vehicle flight control quantity and significantly improving the control performance and response speed of the unmanned aerial vehicle.

[0062] As described in the above step S5, the optimal unmanned aerial vehicle flight control quantity obtained through the previous steps is derived by comprehensively considering multi-sensor data fusion, accurate state estimation, effective deviation compensation and optimized control algorithm. Outputting the control quantity to the actuators of the unmanned aerial vehicle, such as motors and rudders, can accurately control the flight attitude, speed and trajectory of the unmanned aerial vehicle, so that the unmanned aerial vehicle can fly stably according to the preset target flight state, and ensure that the unmanned aerial vehicle can efficiently and reliably complete various flight tasks in complex environments.

[0063] In this embodiment, the beneficial effects of the technical solutions in this embodiment are illustrated by comparison from the following aspects.

[0064] Flight stability comparison:

[0065] In the experiment of simulating strong wind interference environment, the flight control method based on single sensor or simple data fusion has obvious fluctuations in the flight attitude of the unmanned aerial vehicle, and the mean value of the attitude angle deviation is as high as ± 15°, and the maximum distance of the flight trajectory deviating from the preset path is 5 meters. By using the method based on multi-sensor data fusion of the present application, through accurate denoising and reasonable clustering and weighted fusion of multi-source data, the interference influence is effectively suppressed, the mean value of the attitude angle deviation is controlled within ± 5°, and the maximum distance of the flight trajectory deviation is only 1.5 meters, which significantly improves the flight stability.

[0066] Control precision comparison:

[0067] In actual flight test, for the fixed point hovering task, the traditional PID control algorithm is difficult to adapt to the complex environment change, and the position deviation of the unmanned aerial vehicle is large when hovering, the horizontal direction deviation can reach ± 0.8 meters, and the vertical direction deviation can reach ± 0.5 meters. The fuzzy adaptive PID control algorithm combining the genetic algorithm optimization parameter is used in the application, the global optimization is carried out according to the deviation value after compensation, the horizontal direction deviation of the unmanned aerial vehicle when hovering is reduced to ± 0.2 meters, the vertical direction deviation is reduced to ± 0.1 meters, and the control precision is greatly improved.

[0068] Data processing efficiency comparison:

[0069] When processing a large amount of flight data collected by multiple sensors, the traditional data processing method has long data processing time due to the lack of effective denoising and fusion strategy, and the average time for processing 1GB data is 30 minutes, and the accuracy of the fused data is low, and the effective data ratio is only 60%. The application carries out correlation analysis and sparse representation denoising algorithm denoising based on mutual information entropy on the multi-sensor data, and carries out clustering and weighted fusion processing, the average time for processing 1GB data is shortened to 10 minutes, and the effective data ratio of the fused data is improved to 90%, which greatly improves the data processing efficiency and quality.

[0070] Complex environment adaptability comparison:

[0071] In the flight test in complex environments such as simulated mountainous areas and urban high-rise areas, the traditional method has insufficient processing ability for unmodeled dynamics and external disturbances, and the unmanned aerial vehicle frequently appears flight anomalies, such as signal loss 5 times per hour in mountainous environments, and high risk of flight interruption. The application carries out state estimation processing based on an extended state observer, and processes unmodeled dynamics and external disturbance terms as new state variables, and in the same complex environment, the signal loss frequency is reduced to 1 time per hour, which effectively enhances the adaptability of the unmanned aerial vehicle to complex environments and ensures flight safety.

[0072] In summary, whether it is flight stability, control accuracy, data processing efficiency or complex environment adaptability, the unmanned aerial vehicle flight control method based on multi-sensor data fusion proposed in the application has significant advantages compared with the existing method.

[0073] In the embodiment, the deviation of the real-time flight state parameters of the unmanned aerial vehicle is compensated, and the fuzzy adaptive PID control algorithm combining the genetic algorithm optimization parameter is used to globally optimize the parameters of the fuzzy adaptive PID, so as to obtain the optimal unmanned aerial vehicle flight control quantity, and overcome the defect that the current unmanned aerial vehicle cannot realize global optimization of control parameters, resulting in insufficient flight control precision.

[0074] In an embodiment, the unmanned aerial vehicle flight data collected by multiple sensors is denoised to obtain pure data, including:

[0075] The UAV flight data collected by multiple sensors is subjected to correlation analysis based on mutual information entropy to obtain the data correlation matrix;

[0076] According to the data correlation matrix, the flight data is subjected to dimensionality reduction and denoising by a sparse representation reconstruction algorithm to obtain the pure data.

[0077] In this embodiment, the UAV flight data collected by multiple sensors contains information from different types of sensors (such as accelerometers, gyroscopes, barometers, etc.), and there are complex correlations between the above data. Mutual information entropy is a measurement method used to measure the degree of dependence between variables. By calculating the mutual information entropy between different sensor data, the correlation between them can be quantified. Performing a correlation analysis based on mutual information entropy on the flight data collected by multiple sensors can uncover potential dependencies between the data. For example, the changing trends of certain sensor data under specific flight postures of the UAV have strong correlations. Through this analysis, the degree of correlation between each pair of data is expressed in numerical form, and finally a data correlation matrix is ​​constructed. This matrix comprehensively displays the correlation network between multi-sensor data, providing a key basis for subsequent data processing, and helping to clarify which data is redundant and which data is more important for accurately reflecting the flight status of the UAV.

[0078] After obtaining the data association matrix, direct processing of the data collected by multiple sensors is not only computationally complex but may also introduce interference due to its high dimensionality and redundancy. The core idea of ​​the sparse representation reconstruction algorithm is to find a sparse representation of the data within a certain dictionary, allowing the data to be represented using a smaller number of basis vector combinations. Based on the data association matrix, the sparse representation reconstruction algorithm can identify redundant and noisy data that contributes less to the description of the drone's flight status. By mapping the original high-dimensional flight data into a low-dimensional space, components with small coefficients in the sparse representation, namely noise and redundant information, are removed, achieving data dimensionality reduction and denoising. After processing by this algorithm, the resulting clean data not only reduces the data dimension and computational complexity, but also more accurately retains effective information reflecting the drone's true flight status, providing a high-quality data foundation for subsequent steps such as clustering and weighted fusion processing, thereby improving the data processing efficiency and accuracy of the entire drone flight control system.

[0079] In one embodiment, performing clustering and weighted fusion processing on the clean data to obtain fused data includes:

[0080] Inputting the clean data into a long short-term memory network model including an attention mechanism; the attention mechanism adaptively assigns weights according to the data association matrix, performs multi-scale feature extraction and weighted processing, and obtains a weighted feature vector;

[0081] perform clustering analysis on the weighted feature vector to obtain a plurality of data feature clusters;

[0082] For each data feature cluster, an inertia weight and a learning factor are adjusted according to dynamic changes of the feature cluster by using an adaptive particle swarm optimization algorithm, and a weight parameter optimization process is performed to obtain an optimal fusion weight of each cluster.

[0083] According to the optimal fusion weight of each cluster, the pure data is subjected to a cluster-weighted fusion process to obtain fused data.

[0084] In this embodiment, the pure data collected by the multi-sensor of the unmanned aerial vehicle contains time sequence information and features of different dimensions, and the correlation between the data is complex. The long short-term memory network (LSTM) is good at processing time sequence data, can effectively capture the long-term dependence of data in the time dimension, solve the problems of gradient disappearance and gradient explosion existing in the traditional recurrent neural network, and is suitable for analyzing the change law of the unmanned aerial vehicle flight data over time. The introduction of the attention mechanism can automatically identify the more important data part for the unmanned aerial vehicle control decision under different flight states based on the quantization result of the correlation degree matrix of the data. For example, when the unmanned aerial vehicle performs a rapid turning operation, the gyroscope data may be more critical for flight state judgment, and the attention mechanism will allocate a higher weight to this part of data. Through this adaptive weight allocation method, combined with the LSTM for multi-scale feature extraction and weighted processing, the features of different levels, different scales and containing important correlation information can be accurately extracted from the pure data. After the features are weighted and integrated, a weighted feature vector is obtained, which lays a foundation for subsequent data clustering and fusion, and effectively improves the pertinence and effectiveness of data feature extraction.

[0085] The weighted feature vector contains multi-sensor data features that have been screened and weighted, but these features are still relatively scattered. Clustering analysis is a method of grouping data objects into multiple classes or clusters, so that the data objects in the same cluster have a high similarity, and the data objects between different clusters have a low similarity. Clustering analysis on the weighted feature vector can divide vectors with similar features into the same cluster according to the similarity of data features. For example, data features reflecting the stable flight state of the unmanned aerial vehicle may be grouped into one cluster, and data features reflecting the maneuvering flight state of the unmanned aerial vehicle may be grouped into other clusters. Through clustering analysis, the complex weighted feature vector is divided into a plurality of data feature clusters, so that the data structure is more clear, and subsequent differential processing of different types of data features is facilitated, which provides a classification basis for determining the fusion weight of each part of data, and helps to achieve more reasonable data fusion.

[0086] After obtaining multiple data feature clusters, each cluster plays a different role and has a different degree of importance in the flight control of the UAV, and needs to be assigned a suitable fusion weight. The adaptive particle swarm optimization algorithm is an intelligent optimization algorithm that simulates the foraging behavior of a bird swarm, and finds the optimal solution through the movement of particles in the solution space. In this step, the fusion weight of each data feature cluster is taken as a solution in the particle swarm optimization algorithm, and each particle represents a set of possible weight parameters. The algorithm adaptively adjusts the inertia weight and the learning factor according to the dynamic changes of the feature clusters, such as real-time fluctuations of the data, changes in the feature distribution, etc. The inertia weight affects the degree of inheritance of the particle's own historical speed, and the learning factor determines the degree of learning of the particle to the global optimal and individual optimal positions. When the feature cluster data changes less, the inertia weight is reduced, so that the particle pays more attention to local search and optimizes the weight parameters accurately; when the feature cluster data changes more, the inertia weight is increased, so that the search speed of the particle in the solution space is accelerated to adapt to the dynamic changes of the data. Through continuous iterative search, the fusion weight of each cluster that makes the data fusion effect optimal is found, ensuring that in different flight states, the sensor data can be fused in the most appropriate proportion, improving the accuracy and reliability of the fused data.

[0087] Based on the optimal fusion weight of each cluster obtained in the previous step, the pure data is clustered and weighted fused. According to the division of the data feature clusters, the pure data in each cluster is weighted and calculated according to its corresponding optimal fusion weight. For example, for a feature cluster with a higher degree of importance, a higher weight is assigned, and this part of data has a greater impact on the final fusion result in the fusion process; while for a relatively secondary feature cluster, a lower weight is assigned. Through this way of clustered and weighted fusion, the importance differences of different sensor data in different flight scenarios are fully considered, and the multi-source data is organically integrated, and the fused data can more comprehensively and accurately reflect the actual flight state of the UAV, providing high-quality data support for subsequent state estimation based on an extended state observer, etc., thereby improving the overall performance and stability of the UAV flight control.

[0088] In an embodiment, a state estimation process based on an extended state observer is performed on the fused data to obtain real-time flight state parameters of the UAV, including:

[0089] The fused data is divided into state variables to obtain a state variable group;

[0090] According to the state variable group, a state space equation containing the dynamic characteristics of the UAV is constructed to obtain a UAV state model;

[0091] In the UAV state model, an extended state variable is introduced, and unmodeled dynamics and external disturbance terms are processed as new state variables to obtain an extended state space equation;

[0092] According to the extended state space equation, the fusion data is calculated in real time by a preset feedback gain matrix to obtain an observation error;

[0093] Based on the observation error, the feedback gain matrix is updated in parameters to obtain an optimized feedback gain matrix;

[0094] The optimized feedback gain matrix is used to iteratively calculate the extended state space equation to obtain real-time flight state parameters of the unmanned aerial vehicle.

[0095] In this embodiment, the fusion data contains various types of information collected by multiple sensors and processed, which is rich and complex. In order to analyze and utilize these data more systematically, it is necessary to divide the state variables. State variables are key parameters that describe the state of the system. For unmanned aerial vehicles, they usually include position, velocity, attitude angle (such as pitch angle, roll angle, yaw angle) and angular velocity, etc. Through detailed analysis of the fusion data, according to the physical characteristics and control requirements of the flight state of the unmanned aerial vehicle, the related data are classified into different state variables to obtain a state variable group. This division makes the information in the fusion data more clear and orderly, providing a basis for subsequent construction of an accurate unmanned aerial vehicle state model, ensuring that the model can accurately reflect the actual flight state of the unmanned aerial vehicle.

[0096] The state space equation is a mathematical model used to describe the dynamic behavior of a system in modern control theory. Based on the divided state variable group, combined with the dynamic characteristics of the unmanned aerial vehicle, i.e. the relationship between the forces and moments received by the unmanned aerial vehicle during flight and the motion state, the state space equation is constructed. This equation connects the rate of change of state variables with input control quantities and system internal parameters through matrix form, and can accurately describe the state evolution process of the unmanned aerial vehicle under different control inputs. For example, the equation will consider the influence of the mass, moment of inertia, air force and other factors of the unmanned aerial vehicle on the flight state. By constructing such a state space equation, the state model of the unmanned aerial vehicle is obtained, which provides a mathematical framework for subsequent analysis and estimation of the flight state of the unmanned aerial vehicle, and is an important prerequisite for realizing accurate state estimation.

[0097] In actual flight, the flight state of the UAV will be affected by many unmodeled dynamic factors (such as complex airflow, mechanical wear, etc.) and external disturbances (such as electromagnetic interference, gust, etc.), which are difficult to accurately describe in the traditional UAV state model. In order to effectively handle these uncertainties, an extended state variable is introduced into the UAV state model, and the unmodeled dynamics and external disturbance terms are regarded as new state variables. In this way, the disturbances and uncertainties that are difficult to handle are included in the model and analyzed and estimated together with other state variables. The extended state space equation obtained in this way can more comprehensively reflect the real state of the UAV in a complex environment, enhance the adaptability and robustness of the model to the actual flight environment, and provide a more perfect model basis for accurately estimating the real-time flight state parameters of the UAV.

[0098] The feedback gain matrix is a key parameter matrix for adjusting the performance of the observer. According to the extended state space equation, the fusion data is taken as input and real-time feedback calculation is performed using the preset feedback gain matrix. During the calculation process, the predicted state based on the extended state space equation is compared with the actual fusion data to obtain the observation error. This observation error reflects the difference between the model predicted state and the actual state, and it contains the deviation caused by model inaccuracy, disturbance, and data noise. By calculating the observation error in real time, the inconsistency between the actual flight state of the UAV and the model predicted state can be found in time, providing a basis for subsequent adjustment of model parameters and optimization of state estimation.

[0099] The existence of the observation error indicates that the preset feedback gain matrix may not enable the state observer to achieve optimal performance. Based on the observation error, a specific algorithm or strategy is used to update the parameters of the feedback gain matrix. For example, an adaptive control algorithm can be used to dynamically adjust the parameters of the feedback gain matrix according to the size and trend of the observation error. By continuously optimizing the feedback gain matrix, the observer can better adapt to the actual flight state of the UAV, reduce the observation error, and improve the accuracy of state estimation. The optimized feedback gain matrix obtained after parameter updating can enable the state observer to more accurately track the actual state changes of the UAV, providing more effective support for accurately obtaining the real-time flight state parameters of the UAV.

[0100] After obtaining the optimized feedback gain matrix, it is substituted into the extended state space equation for iterative calculation. In each iteration process, according to the state estimation result of the last time and the current fusion data, combined with the optimized feedback gain matrix, the state of the unmanned aerial vehicle is re-estimated. Through continuous iteration, the deviation between the model prediction state and the actual state is gradually corrected, so that the state estimation result is more and more close to the real flight state of the unmanned aerial vehicle. After multiple iterative calculations, the accurate real-time flight state parameters of the unmanned aerial vehicle are finally obtained, such as accurate position, speed, attitude angle, etc. These parameters provide a reliable basis for subsequent unmanned aerial vehicle flight control, which can effectively improve the flight control accuracy and stability of the unmanned aerial vehicle, and ensure that the unmanned aerial vehicle can fly safely and stably in complex environment.

[0101] In the embodiment, the state variable group includes position, speed, attitude angle and angular velocity.

[0102] In an embodiment, the deviation of the real-time flight state parameter of the unmanned aerial vehicle is compensated based on the preset target flight state parameter to obtain a compensated deviation value, including:

[0103] The state variable extraction is performed on the preset target flight state parameter and the real-time flight state parameter of the unmanned aerial vehicle to obtain a target state vector and a real-time state vector;

[0104] According to the target state vector and the real-time state vector, an H-infinity robust control model is constructed;

[0105] In the H-infinity robust control model, the optimal robust control gain matrix is obtained by solving a minimax optimization problem;

[0106] According to the optimal robust control gain matrix, the deviation between the real-time state vector and the target state vector is linearly transformed to obtain a preliminary compensation deviation;

[0107] The stability of the preliminary compensation deviation is verified, and when the stability condition is met, it is taken as the final compensated deviation value.

[0108] In this embodiment, the preset target flight state parameters and the real-time flight state parameters of the UAV cover a wealth of information. In order to facilitate subsequent analysis and processing based on control theory, targeted state variable extraction is needed. State variable extraction is to select variables that are crucial to describing the flight state of the UAV from a large number of parameters, such as position coordinates, flight speed, attitude angle, etc. These extracted key state variables are integrated into target state vectors and real-time state vectors, respectively, where the target state vectors represent the flight state that the UAV is expected to achieve, and the real-time state vectors reflect the actual flight state of the UAV. This vector form of representation structurally processes complex parameter information, making subsequent control model construction and deviation analysis based on vector operations more concise and efficient, and laying a data foundation for achieving precise deviation compensation.

[0109] H-infinity robust control theory aims to design a controller that enables the system to maintain good performance and stability when facing various uncertain factors such as model parameter changes and external disturbances. Based on the target state vector and the real-time state vector, combined with the dynamic characteristics and uncertainty sources of the UAV flight control system, an H-infinity robust control model is constructed. This model, through mathematical equations, links the system dynamics, control inputs, and uncertainty factors of the UAV, and can comprehensively describe the operation of the UAV in complex environments. During model construction, interference factors such as airflow disturbance and sensor measurement error, as well as possible changes in UAV parameters during flight, are fully considered to ensure that the model can accurately reflect various situations in actual flight scenarios, providing a theoretical framework for subsequent deviation compensation and control optimization through the model.

[0110] In the H-infinity robust control model, there is a minimax optimization problem, which aims to minimize the error between the system output and the expected output under the most adverse interference conditions. By solving this minimax optimization problem, a set of optimal robust control gain matrix parameters can be determined. The robust control gain matrix plays a key role in the control process, determining the response of the system to input signals and disturbance signals. The solving process usually uses mathematical optimization algorithms such as the linear matrix inequality (LMI) method, which adjusts the elements of the gain matrix to find the optimal matrix form that achieves the best system performance index. The optimal robust control gain matrix obtained can enable the UAV to maintain stable operation to the greatest extent when facing various uncertainties and disturbances, achieving effective control of the flight state.

[0111] After the optimal robust control gain matrix is obtained, it is applied to the deviation processing of the real-time state vector and the target state vector. Through matrix operation, the deviation is linearly transformed to convert the original deviation vector into a preliminary compensation deviation. This linear transformation is based on the theoretical framework of the H-infinity robust control model and can reasonably adjust and compensate the deviation according to the dynamic characteristics and interference of the system. The preliminary compensation deviation is calculated based on the theoretical model and the optimized gain matrix, which preliminarily considers how to make the actual flight state of the UAV approach the target state, but it still needs to be verified for stability to ensure that the compensated system can operate stably.

[0112] Stability is one of the key performance indicators of the UAV flight control system. The stability of the preliminary compensation deviation is usually verified by using Lyapunov stability theory and other methods to analyze whether the state of the UAV system can remain stable after the preliminary compensation deviation is introduced, and whether unstable phenomena such as divergence or oscillation occur. If the system satisfies the stability condition after verification, it means that the preliminary compensation deviation can effectively reduce the difference between the actual flight state and the target state of the UAV while ensuring the stability of the system, and at this time it is determined as the final compensated deviation value. This final compensated deviation value will be used for subsequent control quantity calculation to enable the UAV to accurately adjust according to the compensated deviation and achieve stable and accurate flight control.

[0113] In an embodiment, according to the compensated deviation value, a fuzzy adaptive PID control algorithm with optimized parameters is used to globally optimize the parameters of the fuzzy adaptive PID to obtain the optimal UAV flight control quantity, including:

[0114] The proportional coefficient, integral coefficient, and derivative coefficient of the fuzzy adaptive PID control algorithm are encoded;

[0115] The deviation change quantity is calculated according to the compensated deviation values of adjacent time points, and the deviation change speed is obtained in combination with the time interval;

[0116] An adaptability evaluation function is constructed based on the compensated deviation value and the deviation change speed, and an initial population is generated;

[0117] Based on the fuzzy adaptive PID control algorithm, selection, crossover, and mutation operations are performed on the initial population to select the coding sequence with the highest evaluation value;

[0118] The coding sequence is decoded to obtain the optimized PID parameters;

[0119] Global optimization calculation of the PID parameters is performed according to the optimized PID parameters and the compensated deviation value to obtain the optimal UAV flight control quantity.

[0120] In this embodiment, the proportional coefficient, integral coefficient and derivative coefficient in the fuzzy adaptive PID control algorithm are the key parameters that determine the control effect. In practical applications, in order to use the genetic algorithm to perform global optimization on these parameters, it is necessary to convert them into a form that can be processed by the genetic algorithm, i.e., coding. The coding process is like giving these parameters a specific gene sequence. Common coding methods include binary coding, Gray coding or floating-point coding. Through coding, the continuous parameter value range is mapped to a discrete coding space, so that the proportional coefficient, integral coefficient and derivative coefficient can participate in the selection, crossover and mutation operations of the genetic algorithm in the form of coding sequences, laying the foundation for subsequent parameter optimization based on the evolutionary idea and realizing the transition from traditional fixed-parameter PID control to parameter adaptive optimization control.

[0121] In the flight control of unmanned aerial vehicles, the deviation change rate is an important indicator reflecting the dynamic characteristics of the system, which can help the control system to perceive the speed of the change in the flight state of the unmanned aerial vehicle. By obtaining the compensated deviation values at adjacent time points, the difference between them is calculated to obtain the deviation change. Dividing the deviation change by the time interval can accurately obtain the deviation change rate. For example, when the unmanned aerial vehicle suddenly encounters airflow interference causing the flight attitude to deviate from the target state, the quickly calculated deviation change rate can enable the control system to respond quickly. This data, together with the compensated deviation value, provides key dynamic information for subsequent construction of the fitness evaluation function, which helps to more comprehensively evaluate the control effect under different PID parameter combinations.

[0122] The fitness evaluation function is a standard for evaluating the quality of individuals in the genetic algorithm, which directly affects the direction and effect of parameter optimization. Based on the compensated deviation value and the deviation change rate, a fitness evaluation function is constructed, which aims to quantify the performance of the unmanned aerial vehicle flight control under different PID parameter combinations. For example, the function can consider factors such as the size of the compensated deviation value, the stability of the deviation change rate, and the energy consumption in the control process. The smaller the deviation, the more stable the change rate, and the lower the energy consumption, the higher the fitness value of the parameter combination. By setting appropriate function forms and weight coefficients, the advantages and disadvantages of different parameter combinations can be accurately evaluated. Generating an initial population is to randomly generate a set of individuals containing different PID parameter coding sequences. These individuals serve as the starting point for the genetic algorithm optimization, and through continuous evolution, better parameter combinations will be gradually selected.

[0123] The selection operation is to select individuals with higher fitness from the initial population according to the results of the fitness evaluation function, so that the excellent PID parameter code sequence has a greater chance of being retained to the next generation, just like the survival of the fittest in nature. The crossover operation simulates the biological gene recombination process, randomly selects part of the coding fragments of two individuals to exchange, thereby generating new parameter combinations, increasing the diversity of the population, and exploring a wider parameter space. The mutation operation randomly changes some gene positions in the individual code with a certain probability, avoiding the algorithm from falling into a local optimal solution, and providing more possibilities for finding the global optimal PID parameters. By repeatedly selecting, crossing, and mutating, the individuals in the population are constantly evolving, and finally the coding sequence with the highest fitness evaluation value is selected. The PID parameter combination corresponding to the sequence performs optimally in the current optimization process.

[0124] The coding sequence selected by the selection, crossover, and mutation operations of the genetic algorithm is the optimal solution found in the coding space. In the actual fuzzy adaptive PID control algorithm, specific proportional, integral, and derivative coefficients are needed for control. Therefore, the coding sequence is decoded to restore it from the coded form to the actual parameter values, obtaining the optimized PID parameters. The decoding process is the inverse operation of coding, which accurately converts the discrete coding sequence into continuous parameter values through a specific mapping relationship. These optimized PID parameters will be used for subsequent unmanned aerial vehicle flight control, providing parameter support for more accurate control.

[0125] After obtaining the optimized PID parameters, they are substituted into the fuzzy adaptive PID control algorithm and combined with the compensated error value for calculation. The fuzzy adaptive PID control algorithm adjusts the PID parameters in real time through fuzzy inference rules according to the error value and error change speed to adapt to different flight conditions. On this basis, the optimized PID parameters are used as initial values for further global optimization calculation. By continuously adjusting the parameters and evaluating the control effect, the optimal control quantity is finally obtained, which can make the unmanned aerial vehicle achieve the optimal control performance in the current flight state. This optimal control quantity will be output to the actuators of the unmanned aerial vehicle, such as motors and rudders, to drive the unmanned aerial vehicle to adjust the flight attitude and trajectory, achieving stable and precise flight.

[0126] In an embodiment, based on the optimal unmanned aerial vehicle flight control quantity, the unmanned aerial vehicle is controlled to fly, comprising:

[0127] Obtaining path information, flight information, and collected task information in the flight task of the unmanned aerial vehicle; the flight information includes flight time, average flight speed, and total flight distance;

[0128] Based on the path information, a two-dimensional closed curve is fitted; based on the flight information, a digital character array is generated;

[0129] obtaining a preset character table, wherein the character table comprises a plurality of non-numeric characters arranged in sequence and without repetition;

[0130] locally changing characters in the character table based on the array of numeric characters to obtain a changed character table, wherein the number of changed characters in the character table is consistent with the number of array elements in the array of numeric characters;

[0131] rearranging characters in the changed character table based on the two-dimensional closed curve to obtain a rearranged character table;

[0132] adding characters in each cell of the rearranged character table to a conversion character column of a conversion table in sequence to obtain a rearranged conversion table, wherein the conversion table comprises an original data column and the conversion character column in one-to-one mapping;

[0133] converting a data stream corresponding to the task information based on the rearranged conversion table and sending the data stream to a management terminal.

[0134] In this embodiment, during the process of the UAV performing the flight task, real-time acquisition of relevant information of the UAV is crucial for subsequent processing and analysis. The path information records the trajectory of the UAV flight, which helps to understand whether the flight route of the UAV is in accordance with the predetermined plan, and whether there is deviation during the flight process. The flight time, average flight speed, and total flight distance in the flight information can reflect the performance indicators of the flight efficiency and energy consumption of the UAV. The collected task information is the data obtained by the UAV in performing a specific task, such as air quality data collected in an environmental monitoring task, images taken in a photography task, etc. The acquisition of this information provides a raw data basis for subsequent information processing and transmission.

[0135] Fitting the path information into a two-dimensional closed curve is to represent the flight path of the UAV in a more intuitive and concise manner. By connecting each point on the path into a closed curve through mathematical fitting method, subsequent analysis and processing of path characteristics are facilitated. Generating a numeric character array based on flight information is to convert numerical information such as flight time, average flight speed, and total flight distance into character form array. The purpose of this is to be able to use these numerical information to operate and transform other character tables in the future, combining flight data with character processing to provide a new way for information encryption and conversion.

[0136] The preset character table is an ordered set containing a plurality of non-numeric characters, which play a key role in subsequent information processing. The existence of the character table provides a basic framework for character operation and transformation, and its sequential arrangement and non-repetition ensure the uniqueness and certainty in character changing and sorting operations.

[0137] The local change of the character table by the digital character array is an operation mode of associating the flight information with the character table. The elements in the digital character array replace or modify the corresponding characters in the character table. This has the advantage of associating the content of the character table with the flight information of the UAV, increasing the confidentiality and uniqueness of the information. And the number of characters changed in the character table is consistent with the number of array elements in the digital character array, ensuring the standardization and consistency of the operation. For example, the preset character table sequentially arranges 10 non-repeating non-numeric characters, which are numbered as character 1, character 2, character 3, …, character 10. Based on the flight information of the UAV, a digital character array is generated, which has three numbers, namely number 3, number 7 and number 2. The first number 3 in the digital character array represents the change operation on the character in the 3rd position of the character table. The set change rule is to move the characters in the character table to the right by one position (here is just an example rule, the actual rule can be various), so the character originally in the 3rd position becomes the character originally in the 4th position after the change. The second number 7 means to change the character in the 7th position of the character table. According to the same change rule, this character becomes the character originally in the 8th position after the change. The third number 2 means to change the character in the 2nd position of the character table, which becomes the character originally in the 3rd position after the operation. After a series of change operations on the corresponding characters in the character table based on the digital character array, the original character table becomes a new changed character table. In practical applications, the change rule can be designed more complex and diversified according to specific needs and algorithms to achieve different information processing effects and security requirements. For example, the digital characters in the digital array can be added to the corresponding characters in the character table to form a new character combination.

[0138] According to the shape and characteristics of the two-dimensional closed curve, the characters in the changed character table are reordered. This ordering method can determine the new order of characters according to the direction and angle of the curve. In this way, the character table is further transformed, so that the arrangement order of the character table is associated with the flight path of the UAV, increasing the complexity and security of information processing. The rearranged character table provides a new character order for subsequent data conversion.

[0139] The characters in the rearranged character table are sequentially added to the conversion character column of the conversion table to construct a new rearranged conversion table. The conversion table has an original data column and a conversion character column, and the elements in the two columns are one-to-one mapped. This structure provides a clear mapping relationship for data conversion, so that subsequent data streams corresponding to task information can be accurately converted according to this conversion table.

[0140] Each data element in the data stream corresponding to the task information is converted according to the mapping relationship between the original data column and the conversion character column by using the rearranged conversion table. Through the conversion, the data of the task information is represented in a new form, and the security and safety of data transmission are increased. The converted data stream is sent to the management terminal, and the management terminal can restore and analyze the data according to the same conversion rule, so as to obtain the original task information collected by the unmanned aerial vehicle. Such a process realizes encrypted transmission and effective management of the task information.

[0141] In summary, the above technical solutions realize encrypted transmission of the task information and associated processing with the flight information by acquiring, processing and converting various information in the unmanned aerial vehicle flight task, and improve the security and management efficiency of the data.

[0142] In an embodiment, the characters in the change character table are reordered based on the two-dimensional closed curve to obtain a rearranged character table, including:

[0143] The two-dimensional closed curve is divided into a plurality of fan-shaped sub-regions by equal-angle fan-shaped region division;

[0144] Each character in the change character table is assigned a unique identifier, and the characters are arranged in an initial character sequence in the order of the identifiers;

[0145] The characters in the initial character sequence are assigned to the fan-shaped sub-regions;

[0146] A polar coordinate system is established with the geometric center of the two-dimensional closed curve as the pole, and the angle of each character in each fan-shaped sub-region in the polar coordinate system is calculated;

[0147] The characters in each fan-shaped sub-region are sorted in ascending order according to the angle of each character in the region;

[0148] The sorted characters in each region are connected in turn according to the clockwise arrangement order of the fan-shaped sub-regions on the two-dimensional closed curve to obtain a rearranged character table.

[0149] In this embodiment, the two-dimensional closed curve is fitted from the unmanned aerial vehicle flight path information and carries the spatial characteristics of the unmanned aerial vehicle flight trajectory. Dividing it into equal-angle fan-shaped regions is to convert the complex curve space into a plurality of regular and easy-to-handle sub-regions. This division method is based on geometric principles and ensures that each fan-shaped sub-region is uniform in the angle dimension, making the subsequent processing of each sub-region uniform and comparable. Each fan-shaped sub-region corresponds to a specific part of the unmanned aerial vehicle flight path, and through division, the overall path information can be disassembled into local features, laying a foundation for subsequent processing combined with character information, and facilitating the association of unmanned aerial vehicle flight states and character operations from the spatial dimension.

[0150] Changing the characters in the character table is an important element of subsequent data conversion, and a unique identifier is assigned to it, such as giving each character an identity card, which facilitates accurate identification and management in complex processing flow. These identifiers can be numbers, special symbols or codes, etc. in the form of establishing a one-to-one correspondence between characters and identifiers. The characters are arranged in the initial character sequence according to the identifier order, so that the characters have a clear arrangement order, providing an ordered basis for subsequent allocation, sorting and other operations. This ordered processing avoids the chaos of characters in the operation process, ensuring that each character can participate in the information processing flow under certain rules.

[0151] After completing the curve area division and character ordered arrangement, the characters in the initial character sequence are allocated to each sector sub-area. The allocation process can be based on various strategies, such as equal allocation, so that each sub-area obtains approximately the same number of characters; or differential allocation according to factors such as the importance and area size of the sub-area. Through this allocation method, the characters are linked to different areas of the UAV flight path, making the character set in each sub-area become an information carrier related to the flight characteristics of the area, creating conditions for subsequent operations on characters based on area characteristics, and realizing the preliminary fusion of UAV path information and character information.

[0152] The geometric center of the two-dimensional closed curve is taken as the polar point to establish a polar coordinate system, which determines the precise position measurement method of the characters in the curve space for each sector sub-area. In the polar coordinate system, the angle coordinate of the character is obtained by calculating the included angle between the character and the polar axis. This angle reflects the relative position relationship of the character in the space surrounded by the curve, providing a quantitative basis for subsequent sorting of characters based on spatial position. Through the polar coordinate system, the position information of the characters is closely combined with the spatial distribution of the UAV flight path, so that the processing of the characters can be mapped to the actual flight space, enhancing the relevance of information processing and UAV flight state.

[0153] After obtaining the angle of each character in the sector sub-area, the characters in the area are sorted in ascending order according to the angle size. This sorting method arranges the characters in each sub-area in the order of their position in the curve space, forming a character sequence with spatial logic. From the perspective of spatial distribution, characters with smaller angles are located near the polar axis, and characters with larger angles are far from the polar axis. Ascending sorting ensures that the characters are arranged in accordance with the spatial position rule. Through this angle-based sorting, the order of characters in each sub-area matches the spatial characteristics of the UAV flight path in that area, providing an ordered local basis for the final integration of characters from each sub-area to form a rearranged character table.

[0154] Finally, according to the clockwise arrangement order of the fan-shaped sub-regions on the two-dimensional closed curve, the characters in each sub-region are sequentially connected after the angle sorting. The above connection mode follows the spatial trend of the UAV flight path, so that the character order in the rearranged character table not only reflects the regional division characteristics of the UAV flight path, but also embodies the ordered arrangement of the characters in each region based on the spatial position. The formation of the rearranged character table realizes the deep fusion of the UAV path information and the character information, and constructs a unique character mapping rule. The rule can be used for subsequent conversion of the task information data stream, and by associating the character order with the flight path characteristics, the confidentiality and uniqueness of information transmission are increased, which provides a new technical means for data security transmission and management.

[0155] With reference to Figure 2 In another embodiment of the present application, a UAV flight control device based on multi-sensor data fusion is also provided, comprising:

[0156] A fusion module is configured to denoise the UAV flight data collected by the multi-sensor to obtain pure data, and perform clustering and weighted fusion processing on the pure data to obtain fusion data.

[0157] An estimation module is configured to perform state estimation processing on the fusion data based on an extended state observer to obtain real-time flight state parameters of the UAV.

[0158] A compensation module is configured to compensate the real-time flight state parameters of the UAV based on a preset target flight state parameter to obtain a compensated deviation value.

[0159] An optimization module is configured to perform global optimization on the parameters of the fuzzy adaptive PID control algorithm by combining the fuzzy adaptive PID control algorithm with a genetic algorithm optimization parameter, to obtain an optimal UAV flight control amount.

[0160] A control module is configured to control the UAV flight based on the optimal UAV flight control amount.

[0161] In the present embodiment, the specific implementation of each module in the above device embodiment is described above in the method embodiment, and will not be repeated here.

[0162] With reference to Figure 3 In the present embodiment, the internal structure of the UAV can be as follows Figure 3As shown in the figure. The UAV includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the processor of the computer design is used to provide computing and control ability. The memory of the UAV includes non-volatile storage medium, internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the UAV is used to store the corresponding data in the embodiment. The network interface of the UAV is used for communication connection with the external terminal through the network. The computer program is executed by the processor to realize the above method.

[0163] Those skilled in the art can understand that, Figure 3 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the application, and does not constitute a limitation on the UAV to which the scheme of the application is applied.

[0164] The embodiment of the application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by the processor to realize the above method. It can be understood that the computer readable storage medium in the embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0165] To sum up, the UAV flight control method based on multi-sensor data fusion provided in the embodiment of the application comprises the following steps: denoising the UAV flight data collected by the multi-sensor to obtain pure data; performing clustering and weighted fusion processing on the pure data to obtain fusion data; performing state estimation processing on the fusion data based on an extended state observer to obtain real-time UAV flight state parameters; performing deviation compensation on the real-time UAV flight state parameters based on preset target flight state parameters to obtain a compensated deviation value; according to the compensated deviation value, performing global optimization on the parameters of fuzzy adaptive PID by combining a fuzzy adaptive PID control algorithm with a genetic algorithm optimization parameter to obtain optimal UAV flight control quantity; and controlling the UAV flight based on the optimal UAV flight control quantity. In the application, the real-time UAV flight state parameters are compensated, and the global optimization is performed on the parameters of fuzzy adaptive PID by combining the fuzzy adaptive PID control algorithm with the genetic algorithm optimization parameter to obtain the optimal UAV flight control quantity, which overcomes the defect that the current UAV cannot realize global optimization of control parameters, resulting in insufficient flight control precision.

[0166] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium provided by the present application and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0167] It should be noted that in this document, the terms "comprising", "including", or any other variant thereof are intended to cover a non-exclusive inclusion, such that a process, device, article or method that comprises a list of elements does not only include those elements, but can also include other elements not expressly listed or inherent to such process, device, article or method. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, device, article or method that includes the element.

[0168] The above description is only the preferred embodiment of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation, or direct or indirect application in other related technical fields, based on the content of the present application specification and drawings, are also included in the patent protection scope of the present application.

Claims

1. A UAV flight control method based on multi-sensor data fusion, characterized in that: The following steps are involved: De-noise the UAV flight data collected by multiple sensors to obtain pure data; Performing clustering and weighted fusion processing on the clean data to obtain fused data; Performing state estimation processing on the fused data based on an extended state observer to obtain real-time flight state parameters of the UAV; Performing deviation compensation on the real-time flight state parameters of the UAV based on preset target flight state parameters to obtain a compensated deviation value; According to the compensated deviation value, the parameters of the fuzzy adaptive PID control algorithm are globally optimized by combining the genetic algorithm to optimize the parameters, thereby obtaining the optimal UAV flight control quantity; Based on the optimal UAV flight control quantity, the UAV flight is controlled.

2. The UAV flight control method based on multi-sensor data fusion according to claim 1 is characterized in that: De-noise the drone flight data collected by multiple sensors to obtain clean data, including: The UAV flight data collected by multiple sensors is subjected to correlation analysis based on mutual information entropy to obtain the data correlation matrix; According to the data correlation matrix, the flight data is subjected to dimensionality reduction and denoising by a sparse representation reconstruction algorithm to obtain the pure data.

3. The UAV flight control method based on multi-sensor data fusion according to claim 2 is characterized in that: The clean data is subjected to clustering and weighted fusion processing to obtain fused data, including: The clean data is input into a long short-term memory network model including an attention mechanism; the attention mechanism adaptively assigns weights according to the data association matrix, performs multi-scale feature extraction and weighted processing, and obtains a weighted feature vector; Performing cluster analysis on the weighted feature vector to obtain multiple data feature clusters; For each data feature cluster, the adaptive particle swarm optimization algorithm is used to adjust the inertia weight and learning factor according to the dynamic changes of the feature cluster, and the weight parameter optimization process is performed to obtain the optimal fusion weight of each cluster; According to the optimal fusion weight of each cluster, the clean data is subjected to clustering weighted fusion processing to obtain fused data.

4. The UAV flight control method based on multi-sensor data fusion according to claim 1, characterized in that: The fusion data is subjected to state estimation processing based on an extended state observer to obtain real-time flight state parameters of the UAV, including: Performing state variable division on the fused data to obtain a state variable group; According to the state variable group, a state space equation including the dynamic characteristics of the UAV is constructed to obtain a UAV state model; Introducing extended state variables into the UAV state model, treating unmodeled dynamics and external disturbances as new state variables, and obtaining an extended state space equation; According to the extended state space equation, a preset feedback gain matrix is ​​used to perform real-time feedback calculation on the fusion data to obtain an observation error; Based on the observation error, updating the parameters of the feedback gain matrix to obtain an optimized feedback gain matrix; The optimized feedback gain matrix is ​​used to iteratively calculate the extended state space equation to obtain the real-time flight state parameters of the UAV.

5. The UAV flight control method based on multi-sensor data fusion according to claim 4 is characterized in that: The state variable group includes position, velocity, attitude angle and angular velocity.

6. The UAV flight control method based on multi-sensor data fusion according to claim 1, characterized in that: According to the compensated deviation value, the fuzzy adaptive PID control algorithm is combined with the genetic algorithm to optimize the parameters, and the parameters of the fuzzy adaptive PID are globally optimized to obtain the optimal UAV flight control quantity, including: Encode the proportional coefficient, integral coefficient and differential coefficient of the fuzzy adaptive PID control algorithm; The deviation change is calculated based on the compensated deviation values ​​at adjacent time points, and the deviation change speed is obtained by combining the time interval; Constructing a fitness evaluation function based on the compensated deviation value and the deviation change rate to generate an initial population; Based on the fuzzy adaptive PID control algorithm, selection, crossover and mutation operations are performed on the initial population to screen out the coding sequence with the highest evaluation value; Decoding the coding sequence to obtain optimized PID parameters; According to the optimized PID parameters and the compensated deviation values, a global optimization calculation of the PID parameters is performed to obtain the optimal UAV flight control quantity.

7. A UAV flight control device based on multi-sensor data fusion, characterized in that: include: The fusion module is used to denoise the UAV flight data collected by multiple sensors to obtain pure data; Performing clustering and weighted fusion processing on the clean data to obtain fused data; An estimation module is used to perform state estimation processing on the fusion data based on an extended state observer to obtain real-time flight state parameters of the UAV; A compensation module is used to perform deviation compensation on the real-time flight state parameters of the UAV based on preset target flight state parameters to obtain a compensated deviation value; An optimization module is used to perform global optimization on the parameters of the fuzzy adaptive PID control algorithm based on the compensated deviation value, thereby obtaining the optimal UAV flight control quantity; The control module is used to control the flight of the drone based on the optimal drone flight control quantity.

Images