Fault early warning method and system for wind disturbance resistant multi-rotor unmanned aerial vehicle

Abstract

The application discloses a fault early warning method and system for a wind interference resistant multi-rotor unmanned aerial vehicle. The method comprises the following steps: synchronously collecting multiple data through a main and auxiliary processing link; generating a future disturbance torque sequence and a first risk assessment value oriented to wind interference priority; generating a flight attitude deviation sequence and a second risk assessment value oriented to attitude priority; obtaining an early warning level and a corresponding composite feedforward control instruction; monitoring the attitude stability of the unmanned aerial vehicle; when the main processing link fails, generating an emergency control instruction based on the assessment result of the auxiliary processing link; if the attitude angular velocity does not decrease to a safety threshold after differential compensation, upgrading the early warning level and switching the control link. The application realizes situational adaptive parallel early warning control through synchronous collection of the main and auxiliary links, task context label driven asymmetric fusion and redundancy switching, thereby solving the problems of single early warning upgrade path and lack of situational adaptive parallel decision in the prior art.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) technology, and in particular to a fault early warning method and system for wind-resistant multi-rotor UAVs. Background Technology

[0002] With the rapid development of drone technology, multi-rotor drones have been widely used in agricultural and forestry patrols, aerial surveying, and logistics transportation due to their high maneuverability and simple structure. However, in actual operation, wind disturbances (such as sudden airflow and turbulence) are one of the important external factors affecting the flight stability of multi-rotor drones. In existing technologies, traditional fault warning systems usually rely on single-path data processing or only focus on a single risk dimension, such as only performing attitude monitoring or only assessing wind disturbance intensity, making it difficult to take into account the synergistic effects of multiple risk factors. Furthermore, when the main processing link fails, existing solutions often cannot achieve rapid and reliable generation of emergency control commands, resulting in blind spots in drone flight safety.

[0003] Therefore, in summary, existing technologies suffer from problems such as a lack of collaborative integration in multi-path risk assessment, poor emergency response capabilities after link failures, and insufficient early warning and control closed loops. Summary of the Invention

[0004] This application provides a fault early warning method and system for wind-resistant multi-rotor UAVs, which solves the problems of single early warning upgrade path and lack of situation-adaptive parallel decision-making in the prior art. It realizes situation-adaptive parallel early warning control that dynamically integrates wind disturbance and attitude risk based on task context labels and supports redundancy switching between primary and secondary links.

[0005] This application provides a fault warning method and system for wind-resistant multi-rotor UAVs, including the following steps: synchronously collecting multiple data through a main and secondary processing link, parsing the navigation commands output by the UAV in real time, and generating task context labels;

[0006] Based on the task context label, the main processing link processes the multi-source sensor data through the first prediction model to generate a future disturbance torque sequence and a first risk assessment value oriented towards wind disturbance priority.

[0007] Based on the task context label, the secondary processing link processes the multi-source sensor data through the second prediction model to generate a flight attitude deviation sequence and a second risk assessment value oriented towards attitude priority.

[0008] The first risk assessment value and the second risk assessment value are asymmetrically weighted and fused using the task context label to obtain the globally optimal early warning level and the corresponding composite feedforward control command.

[0009] The rotor motor is driven to perform differential speed compensation according to the warning level, while the attitude stability of the UAV is monitored.

[0010] When the main processing link fails, emergency control commands are generated independently from the evaluation results of the secondary processing link;

[0011] If the attitude angular velocity does not drop to the safe threshold after differential compensation, the warning level will be upgraded and the control link will be switched according to the preset rules.

[0012] Furthermore, the steps of parsing the navigation commands output by the drone in real time and generating task context labels include:

[0013] The system continuously receives and caches navigation command data streams sent by the UAV flight control unit, the navigation command data streams including command timestamps, command codes, and associated parameters;

[0014] The navigation command data stream is parsed line by line to extract the drone's flight mode commands;

[0015] The kinematic target values ​​are extracted from the associated parameters, including the target's three-dimensional spatial coordinates, the target's horizontal and vertical velocities, and the target's attitude angles;

[0016] The extracted flight mode commands are matched with a preset set of task classification rules;

[0017] After matching is completed, a structured data entity is obtained, and the structured data entity is defined as a task context label.

[0018] Furthermore, the step of processing the multi-source sensor data through the first prediction model to generate a future disturbance torque sequence and a first risk assessment value based on wind disturbance priority includes:

[0019] The task context label is received from the data bus through the main processing link, and the task category name defined in the task context label is called.

[0020] Based on the task category name, load a first prediction model parameter set corresponding to the task category name from the model library;

[0021] Multi-source physical data is collected synchronously through the main processing link, and the multi-source physical data is timestamped to obtain a unified multi-dimensional input time series.

[0022] The multidimensional input time series is fed into the first prediction model with loaded parameters;

[0023] After the first prediction model is calculated, the output is a sequence of predicted disturbance torque values ​​that will affect the roll axis, pitch axis and yaw axis of the UAV at multiple future time steps.

[0024] The main processing link analyzes the output disturbance moment sequence to obtain the maximum magnitude and the maximum rate of change of time of all moment vectors in the disturbance moment sequence, and obtains the first risk assessment value.

[0025] Furthermore, the step of processing the multi-source sensor data through a second prediction model to generate a flight attitude deviation sequence oriented towards attitude priority and a second risk assessment value includes:

[0026] The secondary processing link obtains the task context label from the data bus and configures the running parameters of the second prediction model based on the task category information in the task context label.

[0027] The secondary processing link synchronously collects attitude angle and angular velocity data from the UAV's secondary inertial measurement unit to obtain the optimal UAV three-axis attitude information at the current moment in real time.

[0028] The secondary processing link obtains the desired target attitude command from the UAV flight control unit;

[0029] The optimal three-axis attitude information of the UAV acquired in real time is used to perform vector operations with the desired target attitude command to obtain the real-time three-dimensional attitude deviation vector.

[0030] The three-dimensional attitude deviation vector is continuously input into the second prediction model, and the flight attitude deviation evolution sequence within the future time window is output.

[0031] By using the secondary processing link based on the flight attitude deviation evolution sequence, the predicted attitude deviation peak value and convergence speed are extracted to obtain a second risk assessment value characterizing the attitude instability risk.

[0032] Furthermore, the step of asymmetrically weighted fusion of the first risk assessment value and the second risk assessment value using the task context label includes:

[0033] Receive the first risk assessment value generated by the main processing link, the second risk assessment value generated by the secondary processing link, and the current task context label;

[0034] Based on the task category name contained in the task context label, a pair of weight coefficient values ​​are retrieved from the pre-set weight strategy database;

[0035] Among them, the first weighting coefficient corresponds to the consideration of wind disturbance priority, and the second weighting coefficient corresponds to the consideration of attitude priority.

[0036] Perform a weighted calculation by multiplying the first risk assessment value by the first weighting coefficient and the second risk assessment value by the second weighting coefficient to obtain a preliminary fusion value.

[0037] Obtain the correlation coefficient between the first risk assessment value sequence and the second risk assessment value sequence within the past sliding time window, and generate a global risk score.

[0038] Furthermore, the steps to obtain the globally optimal early warning level and the corresponding composite feedforward control command include:

[0039] Based on the generated global risk score, it is compared with the preset multi-level early warning threshold table;

[0040] The optimal global warning level that should be triggered at the moment is determined by judging the specific numerical range in which the global risk score falls.

[0041] While determining the globally optimal early warning level, composite feedforward control commands are generated in parallel.

[0042] The generation process of the composite feedforward control command includes two calculation branches: the first branch retrieves the future disturbance torque sequence generated and cached by the main processing link, performs inverse calculation through the rotational inertia matrix of the UAV, and obtains the feedforward torque command components.

[0043] The second branch retrieves the flight attitude deviation sequence generated by the sub-processing link and inputs it into the preset proportional, integral, and derivative controllers to obtain feedback control command components.

[0044] The feedforward torque command component and the feedback control command component are vector-synthesized to form the final output composite feedforward control command.

[0045] Furthermore, the steps for driving the rotor motor to perform differential compensation based on the warning level include:

[0046] Based on the generated composite feedforward control command and the determined early warning level, the composite feedforward control command is input into the preset control allocation algorithm;

[0047] The control allocation algorithm decomposes the composite feedforward control command into speed adjustment amounts applied to each independent rotor motor.

[0048] The base speed of each motor is added to the corresponding speed adjustment amount by the control allocation algorithm to generate the final speed target command;

[0049] The target rotation speed command is sent to the electronic speed controllers of each rotor of the UAV, and the electronic speed controllers drive the motors to execute the final target rotation speed command.

[0050] Furthermore, when the main processing link fails, the step of independently generating emergency control instructions from the evaluation results of the secondary processing link includes:

[0051] An independent monitoring program is run inside the processor of the secondary processing link to periodically detect the heartbeat communication signal between the secondary processing link and the primary processing link;

[0052] The monitoring program maintains a timer, which is reset whenever a heartbeat signal is received.

[0053] If the timer count is not less than the preset communication timeout threshold without being reset, the monitoring program will determine that the main processing link has failed to communicate.

[0054] If the main processing link is determined to have a communication failure, the secondary processing link will trigger the control takeover logic and switch to emergency mode.

[0055] In an emergency, the secondary processing link calls the generated second risk assessment value and flight attitude deviation sequence;

[0056] Emergency control commands are generated by inputting the flight attitude deviation sequence into a preset emergency controller;

[0057] The emergency controller uses a time-varying anti-saturation integrator.

[0058] ;

[0059] In the formula, This is an emergency control command vector. For the secondary link proportional gain, For the secondary link integral gain, This is the attitude deviation vector. For integral anti-saturation factor, Let L2 be the norm of the attitude deviation vector;

[0060] Integral anti-saturation factor for:

[0061] ;

[0062] In the formula, This represents the minimum value of the integral anti-saturation factor. This represents the maximum value of the integral anti-saturation factor. The attenuation slope coefficient, The current system time. When the primary link fails This represents the remaining battery power percentage.

[0063] Furthermore, if the attitude angular velocity does not drop to the safe threshold after differential compensation, the steps to upgrade the warning level and switch the control link according to preset rules include:

[0064] After the rotor motor performs differential compensation, the real-time collected triaxial angular velocity values ​​are compared with the preset safety threshold.

[0065] If the angular velocity value of any axis is continuously not less than the preset safety threshold within a specified time window, it is determined that the current control compensation measures have failed to achieve the expected effect, and the preset upgrade and switching rules are activated.

[0066] This application provides a fault warning system for wind-resistant multi-rotor UAVs, which implements a fault warning method for wind-resistant multi-rotor UAVs, including: a command parsing module, a first risk assessment value generation module, a second risk assessment value generation module, a warning level acquisition module, a differential speed compensation module, a control command generation module, and a link switching module.

[0067] The instruction parsing module is used to synchronously collect multiple data through the main and secondary processing links, parse the navigation instructions output by the UAV in real time, and generate task context labels.

[0068] The first risk assessment value generation module is used to process the multi-source sensor data based on the task context label through the main processing link and through the first prediction model to generate a future disturbance torque sequence and a first risk assessment value oriented towards wind disturbance priority.

[0069] The second risk assessment value is generated by processing the multi-source sensor data through the second prediction model based on the task context label via the sub-processing link to generate a flight attitude deviation sequence oriented towards attitude priority and a second risk assessment value.

[0070] The warning level acquisition module is used to perform asymmetric weighted fusion of the first risk assessment value and the second risk assessment value through the task context label to obtain the globally optimal warning level and the corresponding composite feedforward control command.

[0071] The differential speed compensation module is used to drive the rotor motor to perform differential speed compensation according to the warning level, while monitoring the attitude stability of the UAV.

[0072] The control command generation module is used to independently generate emergency control commands based on the evaluation results of the secondary processing link when the main processing link fails.

[0073] The link switching module is used to upgrade the warning level and switch the control link according to preset rules if the attitude angular velocity does not drop to the safety threshold after differential compensation.

[0074] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0075] 1. By synchronously collecting multi-source sensor and navigation command data through the main and secondary processing links and parsing and generating task context labels, efficient real-time identification of flight status is achieved. This enables rapid perception and response to the UAV flight environment in complex wind disturbance environments, solving the problems of lagging flight status monitoring and low identification accuracy in existing technologies.

[0076] 2. By extracting key risk features from the wind disturbance risk assessment value generated by the main road and the attitude deviation assessment value generated by the secondary road, and then weighting and fusing them according to the task context label, the risk considerations of wind disturbance priority and attitude priority are dynamically balanced, thereby obtaining the globally optimal early warning level and composite feedforward control command. This achieves the effect of maintaining accurate early warning decision-making and avoiding misjudgment or omission caused by a single factor even when multiple risk factors coexist.

[0077] 3. By driving the rotor motor to perform differential speed compensation based on the early warning level, and monitoring the change in attitude angular velocity in real time after compensation, and adjusting the link switching and compensation strategy in combination with preset upgrade and switching rules, the UAV can maintain its attitude angular velocity within the safe threshold at each stage, thereby ensuring flight safety and high system reliability. Attached Figure Description

[0078] Figure 1 A flowchart illustrating a fault early warning method for a wind-resistant multi-rotor UAV provided in an embodiment of this application;

[0079] Figure 2 This is a schematic diagram of the structure of a fault warning system for a wind-resistant multi-rotor UAV provided in an embodiment of this application. Detailed Implementation

[0080] This application provides a fault early warning method and system for wind-resistant multi-rotor UAVs, which solves the problems of single early warning upgrade path and lack of situation-adaptive parallel decision-making in the prior art. By synchronously collecting multiple data through the main and secondary links, parsing navigation instructions to generate task context labels, and performing asymmetric weighted fusion and emergency switching of redundant links based on the task context to predict wind disturbance and attitude risks respectively, situation-adaptive parallel early warning and continuous wind-resistant control are realized.

[0081] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.

[0082] like Figure 1The diagram shown is a flowchart of a fault warning method for a wind-resistant multi-rotor UAV provided in an embodiment of this application. The method is applied to a fault warning system for a wind-resistant multi-rotor UAV and includes the following steps: synchronously collecting multiple data through a main and secondary processing link, the multiple data including millimeter-wave radar wind field data, nine-axis IMU attitude data and barometer altitude information, and parsing the navigation commands output by the UAV in real time to generate task context labels.

[0083] Acquire multi-source sensor data synchronously collected by the main and secondary processing links, and determine the current mission context label based on the real-time flight data of the UAV;

[0084] Based on the task context label, the main processing link processes the multi-source sensor data through the first prediction model to generate a future disturbance torque sequence and a first risk assessment value oriented towards wind disturbance priority.

[0085] Based on the task context label, the secondary processing link processes the multi-source sensor data through the second prediction model to generate a flight attitude deviation sequence and a second risk assessment value oriented towards attitude priority.

[0086] The first risk assessment value and the second risk assessment value are asymmetrically weighted and fused using the task context label to obtain the globally optimal early warning level and the corresponding composite feedforward control command.

[0087] The rotor motor is driven to perform differential speed compensation according to the warning level, while the attitude stability of the UAV is monitored.

[0088] When the main processing link fails, an emergency control command is independently generated from the evaluation result of the secondary processing link, driving the control mechanism to execute the composite feedforward control command or the emergency control command to suppress wind disturbance.

[0089] If the attitude angular velocity does not drop to the safe threshold after differential compensation, the warning level will be upgraded and the control link will be switched according to the preset rules.

[0090] Furthermore, the steps of parsing the navigation commands output by the drone in real time and generating task context labels include:

[0091] The system continuously receives and caches navigation command data streams sent by the UAV flight control unit, the navigation command data streams including command timestamps, command codes, and associated parameters;

[0092] The navigation command data stream is parsed line by line to extract the UAV's flight mode commands, including hovering commands, waypoint cruise commands, automatic take-off commands, or automatic landing commands.

[0093] The kinematic target values ​​are extracted from the associated parameters, including the target's three-dimensional spatial coordinates, the target's horizontal and vertical velocities, and the target's attitude angles;

[0094] The extracted flight mode commands are matched with a preset set of task classification rules;

[0095] The task classification rule set defines the mapping relationship between different flight mode commands and abstract task categories.

[0096] After matching is completed, a structured data entity is obtained, and the structured data entity is defined as a task context label;

[0097] The label contains two fields: the first field stores the name of the successfully matched abstract task category, and the second field stores the performance constraint parameters related to the task category extracted from the navigation instructions, such as the position keeping accuracy range or the maximum allowable tracking error.

[0098] The generated task context label is encapsulated by the system and distributed to the main processing link and the secondary processing link via the internal data bus for subsequent steps to call.

[0099] Furthermore, the step of processing the multi-source sensor data through the first prediction model to generate a future disturbance torque sequence and a first risk assessment value based on wind disturbance priority includes:

[0100] The task context label is received from the data bus through the main processing link, and the task category name defined in the task context label is called.

[0101] Based on the task category name, load a first prediction model parameter set corresponding to the task category name from the model library;

[0102] Multi-source physical data is collected synchronously through the main processing link. The multi-source physical data includes wind speed and direction data from millimeter-wave radar, body attitude and angular velocity data from the main inertial measurement unit, and altitude data from the barometer. The multi-source physical data is then timestamped to obtain a unified multi-dimensional input time series.

[0103] The multidimensional input time series is fed into the first prediction model with loaded parameters; the model extracts time-series features from the input series through a series of recurrent neural network layers to capture the dynamic relationship between wind field changes and UAV response;

[0104] The first prediction model calculates the disturbance moment using the turbulence-body coupling equation:

[0105] ;

[0106] In the formula, To predict the disturbance moment vector (N·m). It is a time-varying turbulence intensity factor, and , The wind speed gradient is expressed in seconds (s⁻¹). This is the IMU angular velocity vector (rad / s). This represents the radar wind speed vector (m / s). The sensitivity coefficient is obtained from the task context label. The height attenuation constant (default 0.02m⁻¹) The height drift is expressed in meters. The base torque (taken as the historical average, N·m);

[0107] After the first prediction model is calculated, the output is a sequence of predicted disturbance torque values ​​that will affect the roll axis, pitch axis and yaw axis of the UAV at multiple future time steps.

[0108] The main processing link analyzes the output disturbance moment sequence to obtain the maximum magnitude and the maximum rate of change of time of all moment vectors in the disturbance moment sequence. It then compares these two calculation results with the wind disturbance threshold called by the task context label to generate a quantitative first risk assessment value.

[0109] Furthermore, the step of processing the multi-source sensor data through a second prediction model to generate a flight attitude deviation sequence oriented towards attitude priority and a second risk assessment value includes:

[0110] The secondary processing link obtains the task context label from the data bus and configures the running parameters of the second prediction model based on the task category information in the task context label.

[0111] The secondary processing link synchronously collects attitude angle and angular velocity data from the UAV's secondary inertial measurement unit, and combines it with the altitude information of the global navigation satellite system to obtain the optimal three-axis attitude information of the UAV at the current moment in real time through a state estimation algorithm.

[0112] The secondary processing link obtains the desired target attitude command from the UAV flight control unit;

[0113] The optimal three-axis attitude information of the UAV acquired in real time is used to perform vector operations with the desired target attitude command to obtain the real-time three-dimensional attitude deviation vector.

[0114] The three-dimensional attitude deviation vector is continuously input into the second prediction model. The model processes the input historical attitude deviation vector sequence through its internal filtering and trend analysis modules and outputs the flight attitude deviation evolution sequence within the future time window.

[0115] The second prediction model uses the attitude manifold projection operator:

[0116] ;

[0117] In the formula, For attitude instability, This is the attitude deviation vector (rad). It is the unit vector in the direction of gravity. This is the angular velocity deviation vector (rad / s). This is the cumulative gain coefficient (default 0.05s⁻¹).

[0118] Based on the flight attitude deviation evolution sequence, the predicted attitude deviation peak and convergence speed are extracted through the secondary processing link. These extracted indicators are then quantitatively compared with the attitude stability threshold defined in the mission context label to obtain a second risk assessment value that characterizes the risk of attitude instability.

[0119] Furthermore, the step of asymmetrically weighted fusion of the first risk assessment value and the second risk assessment value using the task context label includes:

[0120] Receive the first risk assessment value generated by the main processing link, the second risk assessment value generated by the secondary processing link, and the current task context label;

[0121] Based on the task category name contained in the task context label, a pair of weight coefficient values ​​are retrieved from the pre-set weight strategy database;

[0122] Among them, the first weighting coefficient corresponds to the consideration of wind disturbance priority, and the second weighting coefficient corresponds to the consideration of attitude priority.

[0123] Perform a weighted calculation by multiplying the first risk assessment value by the first weight coefficient, multiplying the second risk assessment value by the second weight coefficient, and adding the two products to obtain a preliminary fusion value; the weight coefficients here are asymmetrically set, and their sum is not limited to one.

[0124] Obtain the correlation coefficient between the first risk assessment value sequence and the second risk assessment value sequence within the past sliding time window;

[0125] The correlation coefficient is multiplied by a moderating factor also determined by the task context label, and then algebraically summed with the aforementioned preliminary fusion value to correct the assessment results when two risks occur concurrently or influence each other, ultimately generating a single, quantitative global risk score.

[0126] Furthermore, the steps to obtain the globally optimal early warning level and the corresponding composite feedforward control command include:

[0127] Based on the generated global risk score, it is compared with the preset multi-level early warning threshold table;

[0128] The table defines several non-overlapping numerical ranges, and each range corresponds to a unique warning level;

[0129] The optimal global warning level that should be triggered at the moment is determined by judging the specific numerical range in which the global risk score falls.

[0130] While determining the globally optimal early warning level, composite feedforward control commands are generated in parallel.

[0131] The generation process of the composite feedforward control command includes two calculation branches: the first branch retrieves the future disturbance torque sequence generated and cached by the main processing link, performs inverse operation through the rotational inertia matrix of the UAV, and obtains the feedforward torque command component used to directly offset the predicted wind disturbance.

[0132] The second branch retrieves the flight attitude deviation sequence generated by the sub-processing link, inputs it into the preset proportional, integral, and derivative controller, and obtains the feedback control command component used to correct the current attitude error.

[0133] The feedforward torque command component and the feedback control command component are vector-synthesized to form the final output composite feedforward control command.

[0134] Furthermore, the steps for driving the rotor motor to perform differential compensation based on the warning level include:

[0135] Based on the generated composite feedforward control command and the determined early warning level, the composite feedforward control command is input into the preset control allocation algorithm;

[0136] Based on the pre-stored rotor physical layout and motor dynamics model of the UAV, the algorithm decomposes the composite feedforward control command into speed adjustment amounts applied to each independent rotor motor through the control allocation algorithm; these adjustment amounts define the speed differences that need to be generated between the motors, i.e., the differential compensation values.

[0137] The base speed of each motor is added to the corresponding speed adjustment amount by the control allocation algorithm to generate the final speed target command;

[0138] The target rotation speed command is sent to the electronic speed controllers of each rotor of the UAV, and the electronic speed controllers drive the motors to execute the final target rotation speed command.

[0139] After the command is issued, the system immediately starts the attitude stability monitoring process. By continuously collecting three-axis angular velocity data from the primary and secondary inertial measurement units, the system obtains the actual attitude change rate of the UAV after performing differential compensation actions in real time, and uses this data for subsequent decision-making.

[0140] Furthermore, when the main processing link fails, the step of independently generating emergency control instructions from the evaluation results of the secondary processing link includes:

[0141] An independent monitoring program is run inside the processor of the secondary processing link to periodically detect the heartbeat communication signal between the secondary processing link and the primary processing link;

[0142] The monitoring program maintains a timer, which is reset whenever a heartbeat signal is received.

[0143] If the timer count is not less than the preset communication timeout threshold without being reset, the monitoring program will determine that the main processing link has failed to communicate.

[0144] If the main processing link is determined to have a communication failure, the secondary processing link will trigger the control takeover logic and switch to emergency mode.

[0145] The control takeover logic first blocks all instructions from the arbitration fusion module, and then switches its own operating mode to emergency mode;

[0146] In an emergency, the secondary processing link will no longer perform any fusion calculations, but will instead call upon the generated second risk assessment value and flight attitude deviation sequence;

[0147] By inputting the flight attitude deviation sequence into a preset emergency controller, an emergency control command aimed at maintaining attitude stability is generated. This command is then directly sent to the flight dynamics control module for execution.

[0148] The emergency controller uses a time-varying anti-saturation integrator.

[0149] ;

[0150] In the formula, This is an emergency control command vector. The gain of the secondary link is fixed at 0.8 (dimensionless). The integral gain of the secondary link is fixed at 0.3 (dimensionless). The attitude deviation vector ( (The variable is the integral variable, representing a historical moment). For integral anti-saturation factor, Let L2 be the norm of the attitude deviation vector;

[0151] Integral anti-saturation factor for:

[0152] ;

[0153] In the formula, The minimum value of the integral anti-saturation factor is fixed at 0.1 (dimensionless). The maximum value of the integral anti-saturation factor is fixed at 1.0 (dimensionless). The attenuation slope coefficient is (s⁻¹). The current system time (s) The time (s) when the primary link fails. This represents the remaining battery power percentage.

[0154] Furthermore, if the attitude angular velocity does not drop to the safe threshold after differential compensation, the steps to upgrade the warning level and switch the control link according to preset rules include:

[0155] After the drive rotor motor performs differential compensation, the attitude stability monitoring program will compare the real-time collected triaxial angular velocity values ​​with the preset safety threshold.

[0156] If the angular velocity value of any axis is continuously not less than the preset safety threshold within the specified time window, it is determined that the current control compensation measures have failed to achieve the expected effect, and the preset upgrade and switching rules are activated.

[0157] The rule first instructs the early warning system module to forcibly upgrade the current global optimal early warning level to the next higher level, and then broadcasts this new level information to the ground station or other monitoring terminals;

[0158] The rule command control flow switching module performs an operation that interrupts the data path of the composite feedforward control command currently output by the arbitration fusion module, and establishes a new data path that connects the output of the sub-processing link directly to the input of the flight power control module.

[0159] This switching operation ensures that all subsequent control commands are generated independently and uniquely by the secondary processing link, thereby forcing the flight control strategy to switch to attitude priority mode.

[0160] like Figure 2 The diagram shown is a structural schematic of a fault warning system for a wind-resistant multi-rotor UAV provided in an embodiment of this application. The fault warning system for a wind-resistant multi-rotor UAV provided in an embodiment of this application includes: a command parsing module, a first risk assessment value generation module, a second risk assessment value generation module, a warning level acquisition module, a differential speed compensation module, a control command generation module, and a link switching module.

[0161] The instruction parsing module is used to synchronously collect multiple data through the main and secondary processing links, parse the navigation instructions output by the UAV in real time, and generate task context labels.

[0162] The first risk assessment value generation module is used to process the multi-source sensor data based on the task context label through the main processing link and through the first prediction model to generate a future disturbance torque sequence and a first risk assessment value oriented towards wind disturbance priority.

[0163] The second risk assessment value is generated by processing the multi-source sensor data through the second prediction model based on the task context label via the sub-processing link to generate a flight attitude deviation sequence oriented towards attitude priority and a second risk assessment value.

[0164] The warning level acquisition module is used to perform asymmetric weighted fusion of the first risk assessment value and the second risk assessment value through the task context label to obtain the globally optimal warning level and the corresponding composite feedforward control command.

[0165] The differential speed compensation module is used to drive the rotor motor to perform differential speed compensation according to the warning level, while monitoring the attitude stability of the UAV.

[0166] The control command generation module is used to independently generate emergency control commands based on the evaluation results of the secondary processing link when the main processing link fails.

[0167] The link switching module is used to upgrade the warning level and switch the control link according to preset rules if the attitude angular velocity does not drop to the safety threshold after differential compensation.

[0168] Through the above description of the embodiments, those skilled in the art will understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the specific device can be divided into different functional modules to complete all or part of the functions described above.

[0169] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0170] In the embodiments covered by this application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the division of modules is merely a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connections shown or discussed may be indirect coupling or communication connections through some interfaces, devices, or modules, or they may be electrical, mechanical, or other forms of connection.

[0171] The modules described as separate components may or may not be physically separate. Similarly, the components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of the embodiments of this application, depending on actual needs.

[0172] Furthermore, the functional modules in the various embodiments of this application can be implemented either in hardware or as software functional modules. If these functional modules are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a computer program product, which includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the flow or function according to the embodiments of this application is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer program product is stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0173] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

Claims

1. A fault early warning method for wind-resistant multi-rotor unmanned aerial vehicles, characterized in that, Includes the following steps: Multiple data sources are collected synchronously through the main and secondary processing links, and navigation commands output by the UAV in real time are parsed to generate mission context labels; The steps for parsing navigation commands output in real time by the drone and generating task context labels include: The system continuously receives and caches navigation command data streams sent by the UAV flight control unit, the navigation command data streams including command timestamps, command codes, and associated parameters; The navigation command data stream is parsed line by line to extract the drone's flight mode commands; The kinematic target values ​​are extracted from the associated parameters, including the target's three-dimensional spatial coordinates, the target's horizontal and vertical velocities, and the target's attitude angles; The extracted flight mode commands are matched with a preset set of task classification rules; After matching is completed, a structured data entity is obtained, and the structured data entity is defined as a task context label; Based on the task context label, the main processing link processes multi-source sensor data through the first prediction model to generate a future disturbance torque sequence and a first risk assessment value oriented towards wind disturbance priority. Based on the task context label, the secondary processing link processes multi-source sensor data through a second prediction model to generate a flight attitude deviation sequence and a second risk assessment value oriented towards attitude priority. The first risk assessment value and the second risk assessment value are asymmetrically weighted and fused using the task context label to obtain the globally optimal early warning level and the corresponding composite feedforward control command. The rotor motor is driven to perform differential speed compensation according to the warning level, while the attitude stability of the UAV is monitored. When the main processing link fails, emergency control commands are generated independently from the evaluation results of the secondary processing link; If the attitude angular velocity does not drop to the safe threshold after differential compensation, the warning level will be upgraded and the control link will be switched according to the preset rules.

2. The fault early warning method for wind-resistant multi-rotor UAVs as described in claim 1, characterized in that, The steps for processing multi-source sensor data using a first prediction model to generate a future disturbance moment sequence and a first risk assessment value based on wind disturbance priority include: The task context label is received from the data bus through the main processing link, and the task category name defined in the task context label is called. Based on the task category name, load a first prediction model parameter set corresponding to the task category name from the model library; Multi-source physical data is collected synchronously through the main processing link, and the multi-source physical data is timestamped to obtain a unified multi-dimensional input time series. The multidimensional input time series is fed into the first prediction model with loaded parameters; After the first prediction model is calculated, the output is a sequence of predicted disturbance torque values ​​that will affect the roll axis, pitch axis and yaw axis of the UAV at multiple future time steps. The main processing link analyzes the output disturbance moment sequence to obtain the maximum magnitude and the maximum rate of change of time of all moment vectors in the disturbance moment sequence, and obtains the first risk assessment value.

3. The fault early warning method for wind-resistant multi-rotor UAVs as described in claim 1, characterized in that, The steps for processing multi-source sensor data using a second prediction model to generate attitude-priority-oriented flight attitude deviation sequences and second risk assessment values ​​include: The secondary processing link obtains the task context label from the data bus and configures the running parameters of the second prediction model based on the task category information in the task context label. The secondary processing link synchronously collects attitude angle and angular velocity data from the UAV's secondary inertial measurement unit to obtain the optimal UAV three-axis attitude information at the current moment in real time. The secondary processing link obtains the desired target attitude command from the UAV flight control unit; The optimal three-axis attitude information of the UAV acquired in real time is used to perform vector operations with the desired target attitude command to obtain the real-time three-dimensional attitude deviation vector. The three-dimensional attitude deviation vector is continuously input into the second prediction model, and the flight attitude deviation evolution sequence within the future time window is output. By using the secondary processing link based on the flight attitude deviation evolution sequence, the predicted attitude deviation peak value and convergence speed are extracted to obtain a second risk assessment value characterizing the attitude instability risk.

4. The fault early warning method for wind-resistant multi-rotor UAVs as described in claim 1, characterized in that, The step of performing asymmetric weighted fusion of the first risk assessment value and the second risk assessment value using the task context label includes: Receive the first risk assessment value generated by the main processing link, the second risk assessment value generated by the secondary processing link, and the current task context label; Based on the task category name contained in the task context label, a pair of weight coefficient values ​​are retrieved from the pre-set weight strategy database; Among them, the first weighting coefficient corresponds to the consideration of wind disturbance priority, and the second weighting coefficient corresponds to the consideration of attitude priority. Perform a weighted calculation by multiplying the first risk assessment value by the first weighting coefficient and the second risk assessment value by the second weighting coefficient to obtain a preliminary fusion value. Obtain the correlation coefficient between the first risk assessment value sequence and the second risk assessment value sequence within the past sliding time window, and generate a global risk score.

5. The fault early warning method for wind-resistant multi-rotor UAVs as described in claim 1, characterized in that, The steps to obtain the globally optimal early warning level and the corresponding composite feedforward control command include: Based on the generated global risk score, it is compared with the preset multi-level early warning threshold table; The optimal global warning level that should be triggered at the moment is determined by judging the specific numerical range in which the global risk score falls. While determining the globally optimal early warning level, composite feedforward control commands are generated in parallel. The generation process of the composite feedforward control command includes two calculation branches: the first branch retrieves the future disturbance torque sequence generated and cached by the main processing link, performs inverse calculation through the rotational inertia matrix of the UAV, and obtains the feedforward torque command components. The second branch retrieves the flight attitude deviation sequence generated by the sub-processing link and inputs it into the preset proportional, integral, and derivative controllers to obtain feedback control command components. The feedforward torque command component and the feedback control command component are vector-synthesized to form the final output composite feedforward control command.

6. The fault early warning method for wind-disruption-resistant multi-rotor UAVs as described in claim 1, characterized in that, The steps for driving the rotor motor to perform differential compensation based on the warning level include: Based on the generated composite feedforward control command and the determined early warning level, the composite feedforward control command is input into the preset control allocation algorithm; The control allocation algorithm decomposes the composite feedforward control command into speed adjustment amounts applied to each independent rotor motor. The base speed of each motor is added to the corresponding speed adjustment amount by the control allocation algorithm to generate the final speed target command; The target rotation speed command is sent to the electronic speed controllers of each rotor of the UAV, and the electronic speed controllers drive the motors to execute the final target rotation speed command.

7. The fault early warning method for wind-resistant multi-rotor UAVs as described in claim 1, characterized in that, When the primary processing link fails, the step of independently generating emergency control instructions based on the evaluation results of the secondary processing link includes: An independent monitoring program is run inside the processor of the secondary processing link to periodically detect the heartbeat communication signal between the secondary processing link and the primary processing link; The monitoring program maintains a timer, which is reset whenever a heartbeat signal is received. If the timer count is not less than the preset communication timeout threshold without being reset, the monitoring program will determine that the main processing link has failed to communicate. If the main processing link is determined to have a communication failure, the secondary processing link will trigger the control takeover logic and switch to emergency mode. In an emergency, the secondary processing link calls the generated second risk assessment value and flight attitude deviation sequence; Emergency control commands are generated by inputting the flight attitude deviation sequence into a preset emergency controller; The emergency controller uses a time-varying anti-saturation integrator. ； In the formula, For emergency control command vectors, For the secondary link proportional gain, For the secondary link integral gain, This is the attitude deviation vector. For integral anti-saturation factor, Let L2 be the norm of the attitude deviation vector; Integral anti-saturation factor for: ； In the formula, This represents the minimum value of the integral anti-saturation factor. This represents the maximum value of the integral anti-saturation factor. The attenuation slope coefficient, The current system time. When the primary link fails This represents the remaining battery power percentage.

8. The fault early warning method for wind-disruption-resistant multi-rotor UAVs as described in claim 1, characterized in that, If the attitude angular velocity does not drop to the safe threshold after differential compensation, the steps to upgrade the warning level and switch the control link according to preset rules include: After the rotor motor performs differential compensation, the real-time collected triaxial angular velocity values ​​are compared with the preset safety threshold. If the angular velocity value of any axis is continuously not less than the preset safety threshold within a specified time window, it is determined that the current control compensation measures have failed to achieve the expected effect, and the preset upgrade and switching rules are activated.

9. A fault warning system for wind-resistant multi-rotor unmanned aerial vehicles (UAVs), used to implement the fault warning method for wind-resistant multi-rotor UAVs as described in any one of claims 1-8, characterized in that, include: The module includes an instruction parsing module, a first risk assessment value generation module, a second risk assessment value generation module, a warning level acquisition module, a differential speed compensation module, a control instruction generation module, and a link switching module. The instruction parsing module is used to synchronously collect multiple data through the main and secondary processing links, parse the navigation instructions output by the UAV in real time, and generate task context labels. The first risk assessment value generation module is used to process the multi-source sensor data based on the task context label through the main processing link and through the first prediction model to generate a future disturbance torque sequence and a first risk assessment value oriented towards wind disturbance priority. The second risk assessment value is generated by processing the multi-source sensor data through the second prediction model based on the task context label via the sub-processing link to generate a flight attitude deviation sequence oriented towards attitude priority and a second risk assessment value. The warning level acquisition module is used to perform asymmetric weighted fusion of the first risk assessment value and the second risk assessment value through the task context label to obtain the globally optimal warning level and the corresponding composite feedforward control command. The differential speed compensation module is used to drive the rotor motor to perform differential speed compensation according to the warning level, while monitoring the attitude stability of the UAV. The control command generation module is used to independently generate emergency control commands based on the evaluation results of the secondary processing link when the main processing link fails. The link switching module is used to upgrade the warning level and switch the control link according to preset rules if the attitude angular velocity does not drop to the safety threshold after differential compensation.

Images