Data-driven method and system for detecting abnormality in unmanned aerial vehicle attitude control

Abstract

The application discloses a kind of based on data-driven unmanned vehicle attitude control anomaly detection method and system, belong to unmanned vehicle detection technical field, this method is first based on the sensor data and flight attitude data of quadrotor unmanned vehicle of historical time period, respectively train complementary filter model and LSTM model, the first attitude data and second attitude data are outputted in trained complementary filter model and LSTM model;Then based on the attitude data error of last time attitude data and corresponding time unmanned vehicle actual attitude data determined by LSTM model output, according to attitude data error, the first attitude data and second attitude data of current time are weighted fusion, obtain unmanned vehicle attitude prediction data;Finally, according to the residual error of unmanned vehicle attitude prediction data and true unmanned vehicle attitude data in historical data determines confidence interval, according to confidence interval, the attitude of unmanned vehicle is detected;The method effectively improves the accuracy and reliability of unmanned vehicle attitude control anomaly detection, and has important practical significance for improving the safety and reliability of unmanned vehicle flight.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of unmanned aerial vehicle detection, in particular to an unmanned aerial vehicle attitude control anomaly detection method and system based on data driving. BACKGROUND

[0002] In recent years, the rapid development of unmanned aerial vehicle technology has enabled its wide application in military and civilian fields and played an increasingly important strategic value. However, with the increase in the complexity of the unmanned aerial vehicle system structure and the increase in the difficulty of task execution, unmanned aerial vehicle failures occur frequently, thereby causing serious safety accidents and huge economic losses. For unmanned aerial vehicles, flight attitude control is a core technology for ensuring the stable and accurate control of unmanned aerial vehicles. Flight data, as an indirect reflection of the running state of unmanned aerial vehicles, is an important basis for evaluating flight performance and monitoring the health condition of the system. Therefore, developing an unmanned aerial vehicle attitude control anomaly detection method based on data driving can provide accurate attitude in real time, enable a quadcopter unmanned aerial vehicle to maintain a stable flight attitude, and avoid unnecessary casualties and losses, which has important practical significance for improving the safety and reliability of unmanned aerial vehicle flight. The common anomaly detection methods include anomaly detection methods based on prior knowledge, models and data driving.

[0003] A kind of unmanned aerial vehicle flight data anomaly detection method based on LSTM with publication number CN108960303A discusses the establishment of deep learning model based on data driving for unmanned aerial vehicle attitude control. The method reconstructs unmanned aerial vehicle telemetry data phase space data and establishes lstm model for prediction to check abnormal points to carry out unmanned aerial vehicle flight data anomaly detection;Publication number CN117851946A of the abnormal detection method and system and medium for multivariate flight time series data, input unmanned aerial vehicle historical time series data into rnn-cnn model for training to establish model, then input real-time flight data into prediction model for anomaly detection, solve the problem that it is difficult to find and locate anomalies in the current vast flight data, improve the ability of aircraft flight data anomaly detection.

[0004] However, in the above invention technical scheme, the current deep model training based on deep model in processing high-noise data in the real world, although the deep learning model has powerful feature extraction and pattern recognition ability, but may show that the model matching degree is insufficient, especially in the face of a large number of unpredictable noise or abnormal values in data. This matching degree deficiency is usually manifested as the generalization ability of the model decreases, the prediction result is unstable or the sensitivity to abnormal data increases. SUMMARY

[0005] In view of the influence of high noise on the test of flight data abnormal results of a deep learning model in the prior art and the problem of how to improve the cluster cooperation of a quadrotor unmanned aerial vehicle group, the application provides a data-driven unmanned aerial vehicle attitude control abnormality detection method and system, which deploys a quadrotor unmanned aerial vehicle flight task on an unmanned aerial vehicle cloud network end platform based on a docker container, and realizes remote state monitoring and task scheduling of the unmanned aerial vehicle.

[0006] The application is realized by the following technical solutions:

[0007] In a first aspect, the application provides a data-driven unmanned aerial vehicle attitude control abnormality detection method, which comprises the following steps:

[0008] Based on the sensor data and flight attitude data of the quadrotor unmanned aerial vehicle in a historical time period, a complementary filter model and an LSTM model are trained respectively, and the trained complementary filter model and LSTM model output first attitude data and second attitude data;

[0009] The attitude data error is determined based on the last moment attitude data output by the LSTM model and the actual attitude data of the unmanned aerial vehicle at the corresponding moment, and the first attitude data and the second attitude data at the current moment are weighted and fused according to the attitude data error, so as to obtain unmanned aerial vehicle attitude prediction data;

[0010] The confidence interval is determined according to the residual error between the unmanned aerial vehicle attitude prediction data and the real unmanned aerial vehicle attitude data in the historical data, and the attitude of the unmanned aerial vehicle is detected according to the confidence interval.

[0011] Preferably, the sensor data comprises three-axis accelerometer data, three-axis gyroscope data and three-axis geomagnetic sensor data.

[0012] The flight attitude data comprises a pitch angle, a roll angle and a yaw angle.

[0013] Preferably, the complementary filter model outputs the first attitude data, which comprises:

[0014] The sensor data in the navigation coordinate system is converted into sensor data in the carrier coordinate system, the accumulated error of the sensor data in the process of conversion between the navigation coordinate system and the carrier coordinate system is eliminated, the sensor data in the carrier coordinate system is converted into Euler angles, and the first attitude data is obtained.

[0015] Preferably, the sensor data conversion method in the carrier coordinate system is as follows:

[0016] The quaternion transformation matrix between the navigation coordinate system and the carrier coordinate system is determined, and the sensor data in the navigation coordinate system at each moment is converted into sensor data in the carrier coordinate system by using the quaternion transformation matrix.

[0017] Preferably, the determination method of the first attitude data comprises:

[0018] The vector product error generated in the sensor data conversion process is eliminated by using a PI controller, the compensation and correction of the sensor data are completed, and the obtained sensor data is converted into Euler angle form to obtain the first attitude data.

[0019] Preferably, the training method of the LSTM model comprises:

[0020] The sensor data and the corresponding attitude data of the unmanned aerial vehicle at the previous n time points are input into the LSTM model as training data, and the LSTM model is trained in sequence by combining the stochastic gradient descent method.

[0021] Preferably, the weighted fusion method of the first attitude data and the second attitude data is:

[0022]

[0023]

[0024]

[0025] wherein, is the error between the unmanned aerial vehicle attitude data calculated by the complementary filter prediction model at the previous time and the actual attitude data; is the error between the unmanned aerial vehicle attitude data output by the LSTM prediction model at the previous time and the actual attitude data; is the weight coefficient of the complementary filter model; is the weight coefficient of the LSTM model; is the unmanned aerial vehicle attitude prediction data; is the first attitude data; refers to the second attitude data.

[0026] Preferably, the detection of the attitude of the unmanned aerial vehicle according to the confidence interval comprises:

[0027] The sensor data of the unmanned aerial vehicle is input into the trained complementary filter model and the trained LSTM model respectively, the complementary filter model outputs the first attitude data, the LSTM model outputs the second attitude data, the first attitude data and the second attitude data are weighted and fused based on the error, the current time unmanned aerial vehicle attitude prediction data is obtained, the residual error between the unmanned aerial vehicle attitude prediction data and the actual attitude data is calculated, and the residual error is compared with the confidence interval. When the residual error exceeds the confidence interval, the attitude data at the current time point is abnormal.

[0028] In a second aspect, the application provides a data-driven unmanned aerial vehicle attitude control abnormality detection system, characterized in that it comprises:

[0029] The prediction module is configured to train a complementary filter model and an LSTM model based on sensor data and flight attitude data of the quadcopter in a historical time period, and the trained complementary filter model and the LSTM model output first attitude data and second attitude data;

[0030] The attitude fusion module is configured to determine attitude data error based on the last time attitude data output by the LSTM model and actual attitude data of the quadcopter at the corresponding time, and to perform weighted fusion on the first attitude data and the second attitude data at the current time according to the attitude data error to obtain quadcopter attitude prediction data.

[0031] The diagnosis module is configured to determine a confidence interval according to a residual error between the quadcopter attitude prediction data and real quadcopter attitude data in historical data, and to detect the attitude of the quadcopter according to the confidence interval.

[0032] In a third aspect, the present application provides a "cloud-network-terminal" quadcopter cluster control system, characterized in that it comprises a Kubernetes cluster and a ground station.

[0033] The Kubernetes cluster comprises a plurality of quadcopters, each serving as a worker node, and the ground station serves as a master node to control the worker nodes, and the above-mentioned data-driven quadcopter attitude control anomaly detection method is deployed on each worker node.

[0034] Compared with the prior art, the present application has the following beneficial technical effects:

[0035] The application provides a data-driven unmanned aerial vehicle attitude control anomaly detection method, which combines a complementary filtering model and an LSTM model, so that the advantages of the two models can be fully utilized. The complementary filtering model usually has good response and stability to real-time data, while the LSTM model is good at processing time series data and capturing long-term dependencies. This fusion helps to improve the matching degree and generalization ability of the model when facing high noise or abnormal value data, making the prediction result more stable and accurate. Secondly, the attitude data error is determined by the last time attitude data output by the LSTM model and the actual attitude data of the unmanned aerial vehicle at the corresponding time, and the first attitude data and the second attitude data at the current time are weighted and fused according to the attitude data error. This dynamic weight adjustment method can more accurately reflect the current real attitude of the unmanned aerial vehicle, thereby improving the accuracy of anomaly detection; in addition, the confidence interval is determined according to the residual error between the unmanned aerial vehicle attitude prediction data and the real unmanned aerial vehicle attitude data in the historical data, which provides a quantitative standard for the attitude detection of the unmanned aerial vehicle, making the anomaly judgment more objective and reliable. When the residual error between the actual attitude data and the prediction data exceeds the confidence interval, it can be determined as an anomaly, so that measures can be taken in time to avoid potential safety risks.

[0036] The application also provides a data-driven unmanned aerial vehicle attitude control anomaly detection system, an electronic device and a computer storage medium, which have all the advantages of the above-mentioned data-driven unmanned aerial vehicle attitude control anomaly detection method. BRIEF DESCRIPTION OF DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some embodiments of the application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0038] Figure 1 The control architecture of the "cloud-network-terminal" unmanned aerial vehicle cluster control system in the application;

[0039] Figure 2 The working node flowchart of the unmanned aerial vehicle configured as a K8s cluster in the application;

[0040] Figure 3 The first unmanned aerial vehicle attitude prediction flowchart of the complementary filtering model in the application;

[0041] Figure 4 The flowchart of the unmanned aerial vehicle attitude control anomaly detection method in the application. DETAILED DESCRIPTION

[0042] To make the purposes, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some but not all of the embodiments of the present application. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations.

[0043] Therefore, the detailed description of the embodiments of the present application provided below in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. All other embodiments obtained by a person of ordinary skill in the art without creative labor based on the embodiments in the present application are within the scope of protection of the present application.

[0044] It should be noted that similar reference numbers and letters represent similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings.

[0045] A data-driven unmanned aerial vehicle attitude control anomaly detection method, comprising the following steps:

[0046] Step 1, obtaining sensor data and flight attitude data of a quadrotor unmanned aerial vehicle in a historical time period.

[0047] The sensor data includes three-axis accelerometer data, three-axis gyroscope data, and three-axis geomagnetic sensor data.

[0048] The flight attitude data includes Pitch (pitch angle), Roll (roll angle), and Yaw (yaw angle)

[0049] Step 2, training a complementary filter model according to the sensor data and the flight attitude data, and the trained model predicting first attitude data of the quadrotor unmanned aerial vehicle.

[0050] Converting the sensor data in the navigation coordinate system obtained in step 1 into sensor data in the carrier coordinate system, eliminating the vector product error generated in the conversion process of the navigation coordinate system and the carrier coordinate system, and converting the sensor data in the carrier coordinate system into Euler angles to obtain the first attitude data.

[0051] The sensor data conversion method in the carrier coordinate system is as follows:

[0052] Determine the quaternion transformation matrix between the navigation coordinate system and the carrier coordinate system, and convert the sensor data in the navigation coordinate system at each time into sensor data in the carrier coordinate system using the quaternion transformation matrix.

[0053] The vector product error elimination method generated by the sensor data conversion is as follows:

[0054] As shown in Figure 3 The vector product error is composed of two parts. One part is the vector product error obtained by converting the gravitational acceleration from the navigation coordinate system to the carrier coordinate system, and taking the vector product of the gravitational acceleration in the carrier coordinate system with the gravitational acceleration in the carrier accelerometer. The other part is the vector product error obtained by converting the three-axis geomagnetic pole data in the carrier coordinate system at the current time point to the three-axis geomagnetic meter data in the navigation coordinate system, and then to the standard geomagnetic meter data, and then converting the standard three-axis geomagnetic meter data to the three-axis geomagnetic meter data in the carrier coordinate system, and then taking the vector product of the original three-axis geomagnetic meter data in the carrier coordinate system.

[0055] The PI controller is used to eliminate the vector product error generated in the sensor data conversion process, complete the compensation and correction of the sensor data, and then convert the gyroscope data into Euler angle form for output, to obtain the first attitude data of the quadrotor unmanned aerial vehicle at the current time node.

[0056] Step 3, training the LSTM model according to the sensor data and the flight attitude data, and the trained model predicts the second attitude data of the rotor unmanned aerial vehicle.

[0057] In the training process of the LSTM model, the sensor data and the corresponding attitude data of the quadrotor unmanned aerial vehicle at the previous n time points are input into the LSTM model as training data, and the random gradient descent method is used to optimize the sequence of the LSTM model. The trained LSTM model outputs the second attitude data at the current time.

[0058] The framework of the LSTM model includes an input layer, a hidden layer and an output layer. The input layer is divided into a training set and an experimental set. The hidden layer has 128 LSTM neural units, which are used to learn the relationship between the input features and the final model output and to perform state transmission. The output layer calculates the hidden state input by the hidden layer, the bias matrix and the bias term, and finally calculates the attitude angle at the current time through an activation function to output the final result.

[0059] Since the sensor data and the unmanned aerial vehicle attitude data are continuous time series data, the LSTM can capture the long-term dependence of the data through its internal memory unit, and is suitable for processing such continuous dynamic information. The model is trained using data and optimized using the random gradient descent method.

[0060] Step 4, determining the attitude data error based on the last time attitude data output by the LSTM model and the actual attitude data of the unmanned aerial vehicle at the corresponding time, and weighting and fusing the first attitude data and the second attitude data at the current time according to the attitude data error, to obtain the unmanned aerial vehicle attitude prediction data at the current time.

[0061] After the LSTM model outputs the predicted second attitude data of the current time and the complementary filter model outputs the predicted first attitude data of the current time, the first attitude data and the second attitude data are weighted and fused based on the error of the previous time, so as to reduce the influence of the noise in the unmanned aerial vehicle flight data on the unmanned aerial vehicle attitude data.

[0062] After the unmanned aerial vehicle attitude data output by the LSTM model and the unmanned aerial vehicle attitude data output by the complementary filter are dynamically weighted and fused based on the error, the unmanned aerial vehicle attitude prediction data of the current time is obtained. The calculation formula of the dynamic weighted fusion is:

[0063]

[0064]

[0065]

[0066] Among them, Error between the unmanned aerial vehicle attitude data calculated by the complementary filter prediction model at the previous time and the actual attitude data; Error between the unmanned aerial vehicle attitude data calculated by the LSTM prediction model at the previous time and the actual attitude data; Weight coefficient of the complementary filter prediction model in the weighted fusion formula at the current time; Weight coefficient of the complementary filter prediction model in the weighted fusion formula at the current time; Predicted unmanned aerial vehicle attitude data value at the current time obtained by weighted fusion; Predicted unmanned aerial vehicle attitude data value at the current time obtained by the complementary filter model; Predicted unmanned aerial vehicle attitude data value at the current time obtained by the LSTM model.

[0067] Since the LSTM model needs n time unmanned aerial vehicle flight data as input, the unmanned aerial vehicle flight data of the first n time cannot use the output result obtained by the LSTM model. Based on this, the unmanned aerial vehicle attitude data of the first n time adopts the unmanned aerial vehicle attitude data obtained by the complementary filter.

[0068] Step 5, determine the confidence interval according to the unmanned aerial vehicle attitude prediction data and the unmanned aerial vehicle attitude data in the historical data, and detect the attitude of the unmanned aerial vehicle according to the confidence interval.

[0069] The residual error of the predicted unmanned aerial vehicle attitude data and the real unmanned aerial vehicle attitude data is calculated, a confidence interval is determined according to the residual error, the residual error contains random noise, and the noise is normally distributed, and the confidence interval (μ-3σ, μ+3σ) is obtained through the residual error, μ is the average value of the residual data in the sample data, and σ is the standard deviation of the sample residual in the sample set.

[0070] The unmanned aerial vehicle attitude data anomaly detection algorithm fuses the complementary filter and the LSTM network, extracts and fuses the data features of the unmanned aerial vehicle sensors, devices and environment, and is used to support the anomaly detection of the unmanned aerial vehicle flight data. This comprehensive use of various data features can more comprehensively reflect the running state of the unmanned aerial vehicle and improve the comprehensiveness and accuracy of anomaly detection.

[0071] In summary, the technical scheme fuses the complementary filter model and the LSTM model, dynamically adjusts the weight, provides the confidence interval, and comprehensively utilizes various data features, thereby effectively improving the accuracy and reliability of the unmanned aerial vehicle attitude control anomaly detection, and having important practical significance for improving the safety and reliability of the unmanned aerial vehicle flight.

[0072] Based on the above data-driven unmanned aerial vehicle attitude control anomaly detection method, the application also provides a data-driven unmanned aerial vehicle attitude control anomaly detection system, characterized in that it comprises:

[0073] A prediction module is configured to train a complementary filter model and an LSTM model based on the sensor data and flight attitude data of a quadcopter unmanned aerial vehicle in a historical time period, and the trained complementary filter model and LSTM model output first attitude data and second attitude data.

[0074] An attitude fusion module is configured to determine attitude data error based on the last time attitude data output by the LSTM model and the actual attitude data of the unmanned aerial vehicle at the corresponding time, and to perform weighted fusion on the first attitude data and the second attitude data at the current time according to the attitude data error to obtain unmanned aerial vehicle attitude prediction data.

[0075] A diagnosis module is configured to determine a confidence interval according to the residual error of the unmanned aerial vehicle attitude prediction data and the real unmanned aerial vehicle attitude data in the historical data, and to detect the attitude of the unmanned aerial vehicle according to the confidence interval.

[0076] Based on the above data-driven unmanned aerial vehicle attitude control anomaly detection method, the application also provides a "cloud-network-terminal" unmanned aerial vehicle cluster control system, which comprises a Kubernetes cluster and a ground station.

[0077] The Kubernetes cluster includes multiple drones, each drone serving as a worker node, and a ground station serving as a master node to control the worker nodes, and the data-driven drone attitude control anomaly detection method is deployed on each worker node.

[0078] In the drone cluster control system, the ground station realizes centralized management and scheduling of the entire drone cluster through Kubernetes cluster technology. Each drone joins the Kubernetes cluster as an independent node in the cluster. The ground station as the master node is responsible for issuing instructions to each drone, monitoring its flight status and obtaining real-time flight data. The automatic deployment and expansion capabilities of Kubernetes enable the drone cluster to flexibly adjust to changing task requirements.

[0079] The construction method of the Kubernetes cluster is as follows:

[0080] The ip address of each drone in the drone cluster is used as a worker node, and the master node is the control end of the drone cluster, used to control the flight status of each drone.

[0081] Define a configuration file to set up the cloud network end drone master node, which runs on the ground station and is responsible for remotely controlling all drone devices under the cloud network end platform of the drone. Initialize the master node and start the corresponding processes such as etcd (a database responsible for storing the entire node state and cluster configuration), ApiServer (the only entry for resource operations, receives user input commands, provides authentication, authorization, API registration, other modules query or modify data through apiserver, only through apiserver can directly interact with etcd).

[0082] Set the ip address of the corresponding drone node, so that the ip address of the node in docker can be accessed through the API service interface of the master node, and ensure that there is network isolation in the docker container of different nodes to ensure resource isolation and fault isolation.

[0083] Use the kubeadm command to add the worker node to the k8s whole cluster (K8s cluster is a collection of a group of physical or virtual machines, which are organized into a single computing resource pool and run the Kubernetes platform on it. K8s cluster usually includes a master node and multiple worker nodes, the master node is usually responsible for the management and control of the cluster, while the worker node is responsible for running containerized applications. That is, the worker node is the corresponding drone device), and start the kubelet and kube-proxy services to realize request forwarding based on ip address and port to pod.

[0084] When the model is deployed on the drone node at the cloud-net-end, the actual attitude data in the flight state of the drone and the attitude data output by the model are subjected to residual error calculation. When the obtained residual error data is outside the confidence interval calculated by the sample data, it is determined that the attitude data of the drone at the current time point is abnormal. After the flight of the drone ends, the abnormal points and abnormal sequences in the flight state of the drone can be obtained, as follows:

[0085] During the flight, the drone inputs the sensor data of the drone into the trained complementary filter model and the trained LSTM model respectively, the complementary filter model outputs first attitude data, the LSTM model outputs second attitude data, and the first attitude data and the second attitude data are weighted and fused based on errors to obtain the drone attitude prediction data at the current time, and the residual error between the actual attitude data and the model prediction data, i.e., the difference between the actual data and the model prediction data, is calculated.

[0086] In order to effectively identify the abnormality, the calculation result of the residual error is compared with the confidence interval obtained by analyzing the sample data. When the residual error exceeds the preset confidence interval, the system immediately determines that the attitude data at the current time point is abnormal. Further, the system can not only mark the abnormality of a single time point, but also track and record the abnormal sequence, i.e., the abnormal data in continuous multiple time points. The detection of these abnormal points and abnormal sequences helps to discover problems in real time during the flight, such as sensor failure, abnormality of the flight control system or external interference. After the flight ends, all abnormal data points and abnormal sequences are extracted and analyzed to provide valuable information for subsequent flight data review and fault elimination, thereby improving the safety and reliability of the drone flight system.

[0087] It should be noted that in the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative, for example, the division of each module is only a logical function division, and actual implementation can have another division manner, for example, multiple modules can be combined or integrated into another device, or some features can be ignored or not executed. The modules described as separate components can be or can not be physically separated, and the components displayed as modules can be one physical unit or multiple physical units, i.e., can be located in one place or can be distributed to multiple different places. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0088] In addition, each module in various embodiments of the present application can be integrated in one processing unit, or each module can be physically present alone, or two or more modules can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0089] The electronic device provided in the embodiments of the present application includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the steps of the method for detecting abnormality of attitude control of a UAV based on data driving according to any one of the above embodiments are implemented.

[0090] The electronic device provided in another embodiment of the present application can further include: an input port connected to the processor, configured to transmit the multi-modal data collected by an external collection device to the processor; a display unit connected to the processor, configured to display the processing result of the processor to the outside world; and a communication module connected to the processor, configured to realize the communication between the electronic device and the outside world. The display unit can be a display panel, a laser scanning display, etc. The communication mode adopted by the communication module includes, but is not limited to, mobile high-definition link technology (HML), universal serial bus (USB), high-definition multimedia interface (HDMI), wireless connection (including wireless fidelity technology (WiFi), Bluetooth communication technology, low-power Bluetooth communication technology, and communication technology based on IEEE 802.11s).

[0091] The computer readable storage medium provided in the embodiments of the present application stores a computer program. When the computer program is executed by a processor, the steps of the method for detecting abnormality of attitude control of a UAV based on data driving according to any one of the above embodiments are implemented.

[0092] The related parts of the system for detecting abnormality of attitude control of a UAV based on data driving, the electronic device and the computer readable storage medium provided in the embodiments of the present application are described in detail in the corresponding part of the method for detecting abnormality of attitude control of a UAV based on data driving provided in the embodiments of the present application, which will not be described here. In addition, the parts of the above technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail, so as not to be too redundant.

[0093] The above content only illustrates the technical idea of the present application, and cannot limit the protection scope of the present application. Any modification made according to the technical idea of the present application on the basis of the technical solutions falls within the protection scope of the claims of the present application.

Claims

1. A data-driven based unmanned aerial vehicle (UAV) attitude control anomaly detection method, characterized in that, The application relates to a method for predicting the attitude of a quadrotor unmanned aerial vehicle (UAV) and a system thereof. The method comprises the following steps: The complementary filter model outputs the first attitude data, which comprises the following steps: The sensor data in the navigation coordinate system is converted into sensor data in the carrier coordinate system, the integral error of the sensor data generated in the conversion process of the navigation coordinate system and the carrier coordinate system is eliminated, the sensor data in the carrier coordinate system is converted into Euler angles, and the first attitude data is obtained. The training method of the LSTM model comprises the following steps: The sensor data and the corresponding attitude data of the UAV at the previous n time points are input into the LSTM model as training data, and the LSTM model is optimized in sequence by combining the stochastic gradient descent method; the second attitude data is output by the trained LSTM model. The attitude data error is determined based on the last time attitude data output by the LSTM model and the actual attitude data of the UAV at the corresponding time, and the first attitude data and the second attitude data at the current time are weighted and fused according to the attitude data error to obtain the UAV attitude prediction data. The weighted fusion method of the first attitude data and the second attitude data is as follows: wherein, is an error between the UAV attitude data calculated by the complementary filter prediction model at the previous time and the actual attitude data; is an error between the UAV attitude data output by the LSTM prediction model at the previous time and the actual attitude data; is a weight coefficient of the complementary filter model; is a weight coefficient of the LSTM model; is the UAV attitude prediction data; is the first attitude data; refers to the second attitude data; The confidence interval is determined according to the residual error between the UAV attitude prediction data and the real UAV attitude data in the historical data, and the attitude of the UAV is detected according to the confidence interval.

2. The data-driven based abnormal detection method for UAV attitude control according to claim 1, wherein, The sensor data comprises three-axis accelerometer data, three-axis gyroscope data and three-axis geomagnetic sensor data. 3.The data-driven based abnormal detection method for UAV attitude control according to claim 1, wherein, The sensor data conversion method in the carrier coordinate system is as follows: The quaternion transformation matrix between the navigation coordinate system and the carrier coordinate system is determined, and the sensor data in the navigation coordinate system at each time point is converted into sensor data in the carrier coordinate system by using the quaternion transformation matrix.

4. The data-driven based unmanned aerial vehicle attitude control anomaly detection method of claim 1, wherein, The determination method of the first attitude data comprises the following steps: The vector product error generated in the conversion process of the sensor data is eliminated by using a PI controller, the compensation and correction of the sensor data are completed, and the sensor data obtained is converted into Euler angles to obtain the first attitude data.

5. The data-driven based unmanned aerial vehicle attitude control anomaly detection method of claim 1, wherein, The detection of the attitude of the UAV according to the confidence interval comprises the following steps: The sensor data of the UAV is input into the trained complementary filter model and the trained LSTM model respectively, the first attitude data is output by the complementary filter model, the second attitude data is output by the LSTM model, the first attitude data and the second attitude data are weighted and fused based on the error, the UAV attitude prediction data at the current time is obtained, the residual error between the UAV attitude prediction data and the actual attitude data is calculated, the residual error is compared with the confidence interval, and the attitude data at the current time point is abnormal when the residual error exceeds the confidence interval.

6. A system for performing the data-driven based abnormality detection method for UAV attitude control according to any one of claims 1-5, characterized in that, The application relates to a method for predicting the attitude of a quadrotor unmanned aerial vehicle (UAV) and a system thereof. The prediction module is used for training a complementary filter model and an LSTM model based on the sensor data and the flight attitude data of the quadrotor UAV in a historical time period, and the trained complementary filter model and the trained LSTM model output first attitude data and second attitude data. The attitude fusion module is configured to determine attitude data error based on the last time attitude data output by the LSTM model and actual attitude data of the UAV at the corresponding time, and to perform weighted fusion on the first attitude data and the second attitude data at the current time according to the attitude data error, so as to obtain UAV attitude prediction data. The diagnosis module is configured to determine a confidence interval according to a residual error between the UAV attitude prediction data and real UAV attitude data in historical data, and to detect the attitude of the UAV according to the confidence interval.

7. A cloud-net-edge drone swarm control system, characterized in that, The system comprises a Kubernetes cluster and a ground station; The Kubernetes cluster comprises a plurality of UAVs, each UAV serving as a worker node, and the ground station serving as a master node to control the worker nodes, and each worker node is deployed with the data-driven UAV attitude control anomaly detection method according to any one of claims 1-5.

Images