An unmanned aerial vehicle emergency event deduction and response method based on digital twinning

Abstract

This invention discloses a method for emergency event simulation and response for unmanned aerial vehicles (UAVs) based on digital twins. The method includes constructing a digital twin, acquiring flight status data, and determining whether an emergency event is triggered based on emergency triggering rules. When an emergency event is triggered, the current flight status data is used as initial conditions. The digital twin is used to simulate and simulate multiple pre-selected candidate emergency strategies, obtaining various simulation results. The optimal emergency strategy is selected based on a comprehensive scoring function, and the optimal emergency strategy is converted into corresponding flight control commands and issued for execution. During execution, the actual flight status data is compared in real time with the expected data under the optimal emergency strategy to continuously correct the flight control commands. This invention overcomes the limitations of traditional single emergency response strategies by achieving multi-strategy simulation and optimal decision-making through digital twins. Combined with closed-loop control, online correction, and self-learning calibration mechanisms, it significantly improves the safety and reliability of UAV emergency response.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation technology, and specifically relates to a method for simulation and response of unmanned aerial vehicle (UAV) emergency events based on digital twins. Background Technology

[0002] With the rapid development of drone technology, drones have been widely used in surveying and inspection, emergency rescue, logistics and transportation, agricultural plant protection, and urban security. During missions, drones typically rely on autonomous navigation and flight control systems for real-time control. However, due to complex and variable environments, limited energy resources, and unstable communication, various emergencies can still easily occur during operation. For example, when a drone's battery runs low after prolonged flight, or when it experiences attitude deviation, flight path deviation, or signal loss under conditions of strong winds or magnetic interference, failure to promptly identify and take effective countermeasures can often lead to mission failure, equipment damage, or even safety accidents.

[0003] Existing emergency response methods for drones mainly rely on predefined rules or simple threshold triggers. While these methods can ensure drone safety to a certain extent, they have obvious limitations: First, the response strategy is singular and cannot comprehensively determine the optimal solution based on mission status, environmental constraints, and remaining energy. Second, the threshold settings are fixed and cannot adapt to dynamic changes in different mission scenarios. Third, the response decision-making process lacks model support for the actual flight status, making it impossible to predict the execution consequences of different solutions in advance, which can easily lead to misjudgment or overreaction.

[0004] To enhance the intelligence level of drones, academia and industry have begun exploring the application of digital twin technology in the drone field in recent years. However, existing research on drone digital twins mainly focuses on modeling and visualization, such as displaying 3D models of drones, simulating flight trajectories, or verifying control algorithms, primarily serving the design verification and offline analysis stages.

[0005] In emergency management scenarios, leveraging digital twins to achieve real-time simulation and intelligent response for unmanned aerial vehicles (UAVs) remains a significant challenge. On one hand, the suddenness and time-sensitive nature of emergency events demand that systems complete state perception, scenario modeling, strategy simulation, and decision output within a very short timeframe. On the other hand, emergency response processes are influenced by multiple constraints, including energy, flight path, environment, and communication, making traditional static simulations unable to reflect the dynamic changes during actual flight. Current technologies lack a multi-scenario real-time simulation mechanism that can integrate the UAV's physical state, energy model, mission objectives, and environmental factors. Furthermore, most current digital twin systems are one-way mappings, i.e., data synchronization from the physical entity to the virtual model, lacking the closed-loop control capability to drive the physical aircraft's actions from the simulation results. Moreover, existing UAV systems generally lack self-learning capabilities. Data, strategies, and results from each emergency response process are not effectively accumulated and reused, preventing the system from optimizing strategies based on historical experience. Summary of the Invention

[0006] To address the problems existing in the background technology, this invention provides a method for simulated and responded to unmanned aerial vehicle (UAV) emergency events based on digital twins. This method solves the technical problems of existing technologies, such as reliance on predefined rules and static thresholds, single response strategies lacking forward-looking simulation capabilities, inability to comprehensively evaluate the optimal strategy based on real-time status and dynamic environmental changes, lack of closed-loop control capabilities to drive entity actions from simulation results, and lack of a self-learning mechanism based on historical experience.

[0007] The technical solution adopted in this invention is:

[0008] I. A method for simulated and responded to unmanned aerial vehicle (UAV) emergencies based on digital twins:

[0009] S1. Construct a digital twin of the UAV, wherein the digital twin includes at least a three-dimensional structural model, a dynamic model, an energy consumption model, and an environmental scene model.

[0010] S2. Acquire the drone's flight status data in real time, and determine whether to trigger an emergency event based on the flight status data and preset emergency triggering rules.

[0011] When an emergency event is triggered, the current flight status data of the UAV and the information of the environmental scene model are used as initial conditions. The digital twin is used to perform parallel simulation and deduction of multiple pre-generated candidate emergency strategies to obtain the deduction results of each candidate emergency strategy.

[0012] S3. Based on the preset comprehensive scoring function, evaluate the simulation results of all candidate emergency strategies and select the optimal emergency strategy from multiple candidate emergency strategies.

[0013] S4. The optimal emergency strategy is converted into corresponding flight control commands and sent to the UAV for execution. During the execution process, the actual flight status data of the UAV is compared with the expected data under the optimal emergency strategy in real time, so as to continuously correct the flight control commands.

[0014] Step S1 specifically involves:

[0015] S11. Construct a three-dimensional structural model that includes the UAV's geometric and inertial parameters.

[0016] S12. Construct a dynamic model of the UAV.

[0017] S13. Construct an energy consumption model that predicts future changes in electricity consumption at a given time step.

[0018] S14. Construct a communication link model that includes UAV signal strength attenuation and packet loss rate.

[0019] S15. Construct a dynamic environment scene model that includes a terrain model, a dynamic wind field model, and safety constraints.

[0020] S16. The three-dimensional structural model, dynamic model, energy consumption model, communication link model and dynamic environment scene model are encapsulated to form a digital twin, so as to output the predicted UAV state sequence in response to the input of control command.

[0021] Step S2 specifically involves:

[0022] S21. Real-time acquisition of flight status data of the UAV, including position, speed, attitude angle, remaining battery power and communication status data.

[0023] S22. Calculate the drone's attitude yaw angle, lateral offset distance, packet loss rate, and communication interruption duration based on the real-time collected flight status data and the current mission track point index. Then, determine whether the drone has triggered an emergency event based on the attitude yaw angle, lateral offset distance, packet loss rate, communication interruption duration, and remaining battery power.

[0024] S23. When the drone triggers an emergency event, according to the type of emergency event, the corresponding set of candidate emergency strategies is obtained from the preset strategy library, and the set of candidate emergency strategies is parameterized into a set of control command sequences.

[0025] S24. Using the UAV's flight status data and the information of the current dynamic environment scenario model as initial conditions, the digital twin is invoked according to the control command sequence to perform time-series simulation and deduction of each candidate emergency strategy in the acquired candidate emergency strategy set under a preset simulation duration. During the simulation and deduction process, the position, attitude angle and remaining power of the UAV under the current candidate emergency strategy are calculated at each time step, and the feasibility of the path at each time step is checked in real time, thereby obtaining the deduction results of the UAV including trajectory sequence, attitude change and power change.

[0026] The specific steps for determining whether a drone has triggered an emergency event are as follows:

[0027] When the remaining power is below the minimum return power threshold or the power decline rate exceeds the maximum decline rate threshold within several consecutive sampling periods, it is determined to be a power shortage or high-risk power consumption emergency event.

[0028] When the attitude yaw angle exceeds the yaw angle threshold or the lateral deviation distance of the track exceeds the lateral deviation distance threshold, it is determined to be an attitude yaw or track deviation emergency event.

[0029] When the packet loss rate exceeds the maximum allowable packet loss rate or the communication interruption duration reaches the maximum communication interruption duration, it is determined to be a communication anomaly emergency event.

[0030] Specifically, S3 is:

[0031] S31. Based on the simulation results of all candidate emergency strategies, select candidate emergency strategies that do not fall below the minimum safe flight altitude, do not cross no-fly zones, and whose attitude angles do not exceed the safe angle range, and summarize them into a set of feasible candidate emergency strategies.

[0032] S32. Construct a comprehensive scoring function for candidate emergency strategies based on the safety risk level, remaining power, and execution duration of the candidate emergency strategies, thereby calculating the comprehensive score of each candidate emergency strategy in the set of feasible candidate emergency strategies.

[0033] S33. Based on the comprehensive score, sort the candidate emergency strategies in the set of all feasible candidate emergency strategies, and select the candidate emergency strategy with the highest comprehensive score as the optimal emergency strategy.

[0034] The comprehensive score of each candidate emergency strategy in step S32 is obtained by processing it according to the following formula:

[0035]

[0036]

[0037] Where j is the index; This represents the j-th candidate emergency strategy in the set of feasible candidate emergency strategies; Indicate candidate emergency strategies Overall score; , and All represent the overall score weighting coefficients; Indicate candidate emergency strategies Corresponding security risk level; Indicate candidate emergency strategies The remaining battery power at the end of the simulation; Indicates the rated capacity of the drone's battery; Indicate candidate emergency strategies The execution time consumed by the simulation; , and All represent safety risk weighting coefficients; Indicate candidate emergency strategies Flight time of the drone in high-wind areas during the execution; This indicates the minimum battery level threshold for the drone to return to base. Indicate candidate emergency strategies The average packet loss rate during the simulation process; Indicates the maximum allowable packet loss rate; Indicate candidate emergency strategies The cumulative communication interruption duration during the simulation process; Indicates the maximum allowed communication interruption duration; and These represent the weights for packet loss rate and communication interruption duration, respectively.

[0038] Step S4 is as follows:

[0039] S41. Generate flight control commands that can be directly parsed and executed by the UAV's flight control system based on the control command sequence corresponding to the optimal emergency strategy.

[0040] S42. Send the flight control command of the optimal emergency strategy to the UAV through the communication link, instructing the UAV to execute the corresponding optimal emergency strategy.

[0041] S43. During execution, continuously collect actual flight status data of the UAV, compare the actual flight status data with the expected data of the optimal emergency strategy in the simulation in real time, and when the deviation between the actual flight status data and the expected data exceeds the preset deviation threshold, trigger the strategy correction process, generate the corrected flight control command and issue it.

[0042] The trigger strategy correction process is as follows: based on the current deviation magnitude and environmental changes, the optimal emergency strategy currently being executed is locally fine-tuned; or, using the current UAV flight status data and dynamic environment scenario model information as new initial conditions, a time-series simulation is performed on the candidate emergency strategies that were not selected as optimal and whose scores are in the top k, and a new optimal emergency strategy is selected as an alternative.

[0043] The specific steps for model calibration are as follows:

[0044] By comparing the actual flight status data collected throughout the execution process with the expected data from the corresponding simulation, error analysis is performed on the parameters of at least one key model in the digital twin.

[0045] When the error obtained from the analysis exceeds the preset calibration threshold, the parameters of the key model are updated according to the preset rules or optimization algorithm to calibrate the digital twin and improve the accuracy of subsequent inferences.

[0046] The key models are the dynamics model, energy consumption model, or environmental scenario model in the digital twin.

[0047] II. A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0048] The beneficial effects of this invention are:

[0049] 1. This invention utilizes a digital twin to perform parallel simulations of multiple candidate emergency strategies under UAV emergency events, and selects the optimal emergency strategy based on a comprehensive scoring function. It can comprehensively consider factors such as flight status, remaining battery power, environmental constraints, and communication status, overcoming the problems of single emergency response strategies and lack of forward-looking simulation and comprehensive evaluation capabilities in existing technologies, thereby improving the accuracy and safety of UAV emergency decision-making.

[0050] 2. This invention transforms the optimal emergency strategy into flight control commands for execution, and combines real-time deviation correction during execution and model calibration after execution to construct a closed-loop response mechanism from state perception, strategy deduction, decision output to execution feedback. This mechanism can improve the robustness, reliability, and subsequent deduction accuracy of UAV emergency response. Attached Figure Description

[0051] Figure 1 This is a flowchart of the method of the present invention.

[0052] Figure 2 This is a diagram showing the internal model composition of the digital twin of the present invention.

[0053] Figure 3This is a trajectory diagram of several feasible candidate emergency strategies in Example 1.

[0054] Figure 4 The graph shows the remaining power changes of several feasible candidate emergency strategies in Example 1. Detailed Implementation

[0055] The present invention will now be described in more detail with reference to the accompanying drawings and embodiments. However, the present invention is not limited thereto. For those skilled in the art, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention. Contents not described in detail in this specification are prior art known to those skilled in the art.

[0056] like Figure 1 As shown, the drone emergency event simulation and response method of this embodiment includes the following steps:

[0057] S1, such as Figure 2 As shown, a digital twin of a physical drone is constructed. The digital twin includes at least a three-dimensional structural model, a dynamic model, an energy consumption model, and an environmental scene model.

[0058] S11. Construct a three-dimensional structural model that includes the UAV's geometric and inertial parameters.

[0059] Specifically: A three-dimensional structural model containing the UAV's geometric and inertial parameters is constructed using computer-aided design tools in a preset reference coordinate system. In this embodiment, the reference coordinate system is the navigation coordinate system n. The geometric parameters include the UAV's fuselage, propellers, and sensors, etc.; the inertial parameters include the UAV's total mass and inertial tensor; the three-dimensional structural model determines the UAV's physical dimensions, geometric center, mass distribution, and aerodynamic projection area in complex environments (such as wind fields).

[0060] S12. Construct a dynamic model of the UAV.

[0061] The dynamic model is set up according to the following formula:

[0062]

[0063] in, This represents the derivative of the drone's position with respect to time, i.e., the rate of change of position. Represents the velocity vector of the UAV; This represents the derivative of the attitude angle with respect to time. This represents the angular velocity vector of the UAV in system b. Indicates the attitude angle of the drone; Indicates the attitude angle The determined attitude transformation matrix is ​​used to map angular velocity to attitude angle change rate; This represents the acceleration vector of the UAV in the navigation coordinate system n, that is, the derivative of the velocity vector with respect to time; The mass of the drone is represented by g; g represents the acceleration due to gravity. This represents the velocity vector of the UAV in the navigation coordinate system; The drag coefficient represents the coefficient of friction that is linearly related to speed. This represents the nonlinear drag coefficient related to the square of the velocity; Represents the velocity vector The modulus length; Indicates the attitude angle The rotation matrix determined from the slave coordinate system b to the navigation coordinate system n; This represents the total thrust generated by the drone's propellers. This represents the derivative of angular velocity with respect to time. The matrix representing the rotational inertia of the UAV; This represents the inverse of the rotational inertia matrix of the UAV. This represents the control torque generated by the propeller; It represents the aerodynamic torque generated by aerodynamic action.

[0064] S13. Construct an energy consumption model that predicts future changes in electricity consumption at a given time step.

[0065] The energy consumption model is set according to the following formula:

[0066]

[0067]

[0068]

[0069]

[0070]

[0071]

[0072] in, For the current time step The current state of battery charge (remaining percentage of charge). This is the simulation time step; For the next time step The current state of battery charge; This refers to the battery's rated capacity. For the current time step To the next time step Average total power within; This refers to the battery terminal voltage. For the current time step The remaining energy below; For the next time step The remaining energy below; For the current time step To the next time step Energy consumed internally; For the current time step The instantaneous total power; The power consumption of the onboard electronic equipment of the drone; For indexing; The number of rotors (motors / propellers) included in the drone; For the first Power consumption of each rotor; For the first The thrust generated by each rotor; For the first The incoming flow velocity at the location of each rotor; For the first The rotational speed of each rotor; air density; For rotor propulsion efficiency; The rotor area; This refers to the aerodynamic drag coefficient related to the rotor. Where is the rotor radius.

[0073] S14. Construct a communication link model that includes UAV signal strength attenuation and packet loss rate.

[0074] The communication link model simulates signal strength attenuation. and packet loss rate With the location of the drone change.

[0075] S15. Construct a dynamic environment scene model that includes a terrain model, a dynamic wind field model, and safety constraints.

[0076] The dynamic environment scene model is as follows:

[0077] A terrain model is constructed using a digital elevation model (DEM) to accurately simulate terrain undulations.

[0078] Using a three-dimensional wind field model A dynamic wind field model is constructed that takes into account both the variation of wind speed with height (e.g., logarithmic law) and time variation (e.g., gust model), where Represents three-dimensional coordinates. Indicates time.

[0079] Safety constraints include maximum permissible wind speed, minimum safe flight altitude, attitude angle constraints, and no-fly zone constraints.

[0080] The maximum permissible wind speed sets the upper limit of the allowable wind speed within the dynamic environment scenario model. Minimum safe flight altitude. :constraint Where z represents the drone's flight altitude, This indicates the minimum safe flight altitude, used to ensure sufficient safe distance between the drone and the ground or obstacles. Indicates the drone's horizontal position The corresponding terrain altitude value is obtained from the terrain model. Attitude angle constraints include the maximum safe angles for the UAV's pitch angle θ, roll angle φ, and heading angle ψ. The no-fly zone (NFZ) constraint is defined as a set of spatial polygons. Constraining the position of drones .

[0081] S16. The three-dimensional structural model, dynamic model, energy consumption model, communication link model and dynamic environment scene model are encapsulated to form a digital twin, so as to output the predicted UAV state sequence in response to the input of control command.

[0082] S2. Acquire the drone's flight status data in real time, and determine whether to trigger an emergency event based on the flight status data and preset emergency triggering rules.

[0083] When an emergency event is triggered, the current flight status data of the UAV and the information of the environmental scene model are used as initial conditions. The digital twin is used to perform parallel simulation and deduction of multiple pre-generated candidate emergency strategies to obtain the deduction results of each candidate emergency strategy.

[0084] S21. Periodically collect flight status data of the UAV, including position, speed, attitude angles (including pitch angle θ, roll angle φ and heading angle ψ), remaining battery power and communication status data, through the flight control system and airborne sensors on the UAV.

[0085] In practice, the flight status data of the UAV is periodically collected through telemetry data from the flight control system and airborne sensors, and set according to the following formula:

[0086]

[0087] in, This refers to the flight state data at the k-th time step; These are the three-dimensional position coordinates of the UAV; These are the x, y, and z coordinates of the UAV in the navigation coordinate system n, respectively. These are the roll angle, pitch angle, and yaw angle of the drone, respectively. This is the current remaining battery level.

[0088] S22. Calculate the drone's attitude yaw angle, lateral offset distance, packet loss rate, and communication interruption duration based on the real-time collected flight status data and the current mission track point index. Then, determine whether the drone has triggered an emergency event based on the attitude yaw angle, lateral offset distance, packet loss rate, communication interruption duration, and remaining battery power.

[0089] In practice, the UAV's planned flight path is obtained based on the current mission waypoint index. This includes both already flown destinations and target destinations to be reached. Based on the planned route. Based on the drone's current position, calculate the lateral offset distance of its flight path (i.e., the shortest distance from the current position to the current flight segment). Simultaneously, based on the predetermined flight path... The desired heading is determined, and the attitude yaw angle is calculated based on the UAV's current attitude angle. UAV communication status data is obtained from the communication link model, from which packet loss rate and communication interruption duration can be calculated.

[0090] The specific steps for determining whether a drone has triggered an emergency event (based on preset emergency triggering rules) are as follows:

[0091] When the remaining power is below the minimum return power threshold or the power decline rate exceeds the maximum decline rate threshold within several consecutive sampling periods, it is determined to be a power shortage or high-risk power consumption emergency event.

[0092] When the attitude yaw angle exceeds the yaw angle threshold or the lateral deviation distance of the track exceeds the lateral deviation distance threshold, it is determined to be an attitude yaw or track deviation emergency event.

[0093] When the packet loss rate exceeds the maximum allowable packet loss rate or the communication interruption duration reaches the maximum communication interruption duration, it is determined to be a communication anomaly emergency event.

[0094] In practice, multiple sets of threshold conditions will be set according to the mission type and flight environment, including minimum return battery power threshold, maximum descent rate threshold, yaw angle threshold, lateral offset distance threshold, maximum allowable packet loss rate, and maximum communication interruption duration.

[0095] In practice, a composite threshold triggering mechanism is adopted to improve the sensitivity to potential risks.

[0096] The following situations trigger low battery or high-risk power consumption emergency events: when the drone's real-time remaining battery power... Minimum return battery threshold (e.g., 25%); or when the rate of battery depletion... Exceeding the maximum descent rate threshold for several consecutive sampling periods (e.g., 0.5% / s).

[0097] The following situations trigger an attitude yaw or track deviation emergency: when the drone's position deviates from the planned flight path. Lateral offset distance of the flight path Horizontal offset distance threshold When; or when the drone's attitude yaw angle exceeds the yaw angle threshold. (e.g., 35 degrees Celsius).

[0098] The following situations trigger communication anomaly emergency events: when the packet loss rate... Maximum allowable packet loss rate (e.g., 50%); or when the communication interruption duration reaches the maximum communication interruption duration. hour.

[0099] S23. When the drone triggers an emergency event, the corresponding set of candidate emergency strategies is retrieved from the preset strategy library according to the type of emergency event. ,in A set of candidate emergency response strategies, A set of candidate emergency strategies The first strategy in the process, A set of candidate emergency strategies The second strategy in the process, A set of candidate emergency strategies The first in (One strategy), and parameterize the set of candidate emergency strategies into a set of control command sequences.

[0100] The default strategy library is as follows:

[0101] For power shortage or high-risk power consumption emergencies, the candidate emergency strategy set includes turning back along the original route, returning directly along the shortest path, choosing the nearest safe alternate landing point, and descending to the preset flight altitude and returning directly.

[0102] For emergency events involving deviations from the course or significant track deviations, the candidate emergency strategies include a single course correction to return to the original route, multiple segmented course corrections to return to the original route, and adjusting altitude to avoid areas of strong winds.

[0103] For communication anomaly emergencies, the set of candidate emergency strategies includes returning along the pre-set lost contact route, hovering in place and waiting for the link to be restored, and landing at the nearest safe alternate landing point.

[0104] S24. Using the UAV's flight status data and the information of the current dynamic environment scenario model as initial conditions, the digital twin is invoked according to the control command sequence to perform time-series simulation on each candidate emergency strategy in the acquired candidate emergency strategy set for a preset simulation duration. During the simulation, the position, attitude angle and remaining battery power of the UAV under the current candidate emergency strategy are calculated at each time step, and the feasibility of the path at each time step is checked in real time, thereby obtaining the simulation results of the UAV, including the trajectory sequence, attitude change and battery power change on the time axis.

[0105] In practice: Each candidate emergency strategy Initiate an independent simulation process. For each simulation process, use the constructed digital twin to iteratively calculate the position, attitude angle, and remaining battery power at each time step, and check the feasibility of the path at each time step. Preset the simulation duration. After completion, the drone's trajectory sequence, attitude changes, and battery level changes were obtained.

[0106] The iterative process is as follows:

[0107] F1. Use the drone's flight status data and the information from the current dynamic environment scene model as the initial conditions for the digital twin.

[0108] F2. Input the control command at the current time step k in the control command sequence into the digital twin.

[0109] F3. Based on the dynamic model and communication link model of the digital twin, calculate the flight status data for the next time step k+1, including position, velocity, attitude angle, remaining battery power, and communication status data.

[0110] F4. Calculate the remaining energy at the next time step k+1 based on the energy consumption model. .

[0111] F5. Based on the dynamic environment scenario model with safety constraints, check whether the path of the drone at the current time step k is feasible.

[0112] F6. Repeat steps F2-F5 to achieve continuous iterative calculation.

[0113] S3. Based on the preset comprehensive scoring function, evaluate the simulation results of all candidate emergency strategies and select the optimal emergency strategy from multiple candidate emergency strategies.

[0114] S31. Based on the simulation results of all candidate emergency strategies, including trajectory sequence, attitude change and power change, select candidate emergency strategies that do not fall below the minimum safe flight altitude, do not cross no-fly zones and whose attitude angles do not exceed the safe angle range, and summarize them into a set of feasible candidate emergency strategies.

[0115] S32. Construct a comprehensive scoring function for candidate emergency strategies based on the safety risk level, remaining power, and execution duration of the candidate emergency strategies, thereby calculating the comprehensive score of each candidate emergency strategy in the set of feasible candidate emergency strategies.

[0116] The overall score for each candidate emergency strategy is obtained using the following formula:

[0117]

[0118]

[0119] Where j is the index; This represents the j-th candidate emergency strategy in the set of feasible candidate emergency strategies; Indicate candidate emergency strategies Overall score; , and All represent the overall score weighting coefficients; Indicate candidate emergency strategies Corresponding security risk level; Indicate candidate emergency strategies The remaining battery power at the end of the simulation; Indicates the rated capacity of the drone's battery; Indicate candidate emergency strategies The execution time consumed by the simulation; , and All represent safety risk weighting coefficients; Indicate candidate emergency strategies The flight time of the drone in high-wind areas during the execution (areas with wind speeds of 8 m / s or higher are pre-defined as high-wind areas); This indicates the minimum battery level threshold for the drone to return to base. Indicate candidate emergency strategies The average packet loss rate during the simulation process; Indicates the maximum allowable packet loss rate; Indicate candidate emergency strategies The cumulative communication interruption duration during the simulation process; Indicates the maximum allowed communication interruption duration; and These represent the weights for packet loss rate and communication interruption duration, respectively.

[0120] S33. Based on the comprehensive score, sort the candidate emergency strategies in the set of all feasible candidate emergency strategies, and select the candidate emergency strategy with the highest comprehensive score as the optimal emergency strategy.

[0121] S4. The optimal emergency strategy is converted into corresponding flight control commands and sent to the UAV for execution. During the execution process, the actual flight status data of the UAV is compared with the expected data under the optimal emergency strategy in real time, so as to continuously correct the flight control commands.

[0122] S41. Generate flight control commands that can be directly parsed and executed by the UAV's flight control system based on the control command sequence corresponding to the optimal emergency strategy.

[0123] Specifically, flight control commands include parameters such as target waypoint coordinates, target altitude, target heading, and desired speed.

[0124] S42. Send the flight control command of the optimal emergency strategy to the UAV through the communication link, instructing the UAV to execute the corresponding optimal emergency strategy.

[0125] S43. During execution, continuously collect actual flight status data of the UAV, compare the actual flight status data with the expected data of the optimal emergency strategy in the simulation in real time, and when the deviation between the actual flight status data and the expected data exceeds the preset deviation threshold, trigger the strategy correction process, generate the corrected flight control command and issue it.

[0126] The trigger strategy correction process is as follows: based on the current deviation magnitude and environmental changes, the optimal emergency strategy currently being executed is locally fine-tuned; or, using the current UAV flight status data and dynamic environment scenario model information as new initial conditions, a time-series simulation is performed on the candidate emergency strategies that were not selected as optimal and whose scores are in the top k, and a new optimal emergency strategy is selected as an alternative.

[0127] Specifically, step S43 is as follows:

[0128] Real-time calculation of actual flight status data Compared with expected data deviation .

[0129] Local correction trigger: when position deviation Or attitude angle deviation Exceeding the preset threshold When this happens, the strategy correction process is triggered.

[0130] Local fine-tuning: If the deviation is within the preset range, adjust the original control command sequence. Perform smooth fine-tuning.

[0131] Re-analysis: If the deviation exceeds the preset range, or environmental conditions (such as wind speed) If a significant change occurs, the current real-time position, speed, attitude angle, remaining battery power, and current environmental scene information of the UAV are immediately used as new initial conditions. The time-series simulation is performed on the candidate emergency strategies that were not selected as the optimal ones and whose scores are in the top k. A new optimal emergency strategy is selected as an alternative, and then the corrected flight control commands are generated and issued to ensure the robustness and safety of the emergency response process.

[0132] Furthermore, after or during the execution of the optimal emergency strategy, a model calibration step is also included.

[0133] By comparing the actual flight status data collected throughout the execution process with the expected data from the corresponding simulation, error analysis is performed on at least one key model parameter in the digital twin.

[0134] When the error obtained from the analysis exceeds the preset calibration threshold, the key model parameters are updated according to the preset rules or optimization algorithm to calibrate the digital twin and improve the accuracy of subsequent inferences.

[0135] Key models include dynamic models, energy consumption models, and environmental scenario models in digital twins.

[0136] In practice, the preset rules or optimization algorithms can be one or more combinations of Recursive Least Squares (RLS), Extended Kalman Filter (EKF), Gradient Descent, and Particle Swarm Optimization (PSO), and the selection can be made according to the parameter type, observability, and calibration timeliness requirements.

[0137] For linearization parameters in the dynamic model (such as the drag coefficient which is linearly related to velocity) and the nonlinear drag coefficient related to the square of the velocity The drag coefficient can be recursively identified using the recursive least squares method for real-time online updates, and the speed and acceleration data collected during flight can be used for recursive identification.

[0138] For key parameters in the energy consumption model, different calibration methods can be selected according to the parameter type: rotor-related aerodynamic drag coefficient and rotor propulsion efficiency Quasi-online estimation can be performed using extended Kalman filtering; this constitutes the battery terminal voltage. The parameters (such as polynomial fitting coefficients or lookup table values) can be optimized offline using the gradient descent method, and the voltage curve can be fitted and calibrated using complete discharge data collected from multiple flight missions.

[0139] For large-scale parameter identification related to environmental scene models, such as 3D wind field models The parameters in the model and the parameters in the terrain model can be globally searched using the particle swarm optimization algorithm to identify these spatially distributed parameters by minimizing the model prediction error over a long period of time.

[0140] The calibrated model parameters are stored in a database to improve the accuracy of subsequent emergency simulations.

[0141] The model calibration step is to enable the updated digital twin model to more accurately predict UAV trajectories and energy changes in subsequent emergency event simulations, thereby achieving the adaptive and self-learning capabilities of the UAV emergency response system.

[0142] Example 2:

[0143] S1. Construct a digital twin of the physical UAV using the same method as in Example 1. The digital twin includes a three-dimensional structural model, a dynamic model, an energy consumption model, and an environmental scene model.

[0144] S21. Real-time acquisition of flight status data of the UAV, including position, speed, attitude angle, remaining battery power and communication status data.

[0145] S22. Calculate the drone's attitude yaw angle, lateral offset distance, packet loss rate, and communication interruption duration based on the real-time collected flight status data and the current mission track point index. Then, determine whether the drone has triggered an emergency event based on the attitude yaw angle, lateral offset distance, packet loss rate, communication interruption duration, and remaining battery power.

[0146] The following conditions were detected in this embodiment:

[0147] Remaining battery power Minimum return battery threshold ;

[0148] Rate of decrease in battery power over the past 5 consecutive sampling periods Maximum descent rate threshold ;

[0149] No attitude yaw or communication anomalies.

[0150] Based on this, the system determines that a "low battery emergency event" has been triggered and marks the event and the current state as the initial conditions for the simulation.

[0151] S23. When the drone triggers an emergency event, the corresponding set of candidate emergency strategies is retrieved from the preset strategy library according to the type of emergency event. The candidate emergency strategy set is parameterized into a sequence of control commands. In this embodiment, the current position of the UAV is far from the return point. The distance is 10km from the nearest alternate landing point. For a distance of 4km, the generated candidate emergency strategies include:

[0152] Shortest path return (target) Maintain current altitude Original speed ).

[0153] Economic slow return (target) Target height Economic speed ).

[0154] Alternate landing (target) Minimum safe height Safe speed ).

[0155] : Detour around the wind (planned route avoids predicted strong wind areas, then performs economical low-speed return).

[0156] Each strategy Corresponding to a sequence of control commands .

[0157] S24. Using the current UAV flight status data and the information from the current dynamic environment scenario model as initial conditions, initiate an independent simulation process for each candidate emergency strategy, within a preset simulation duration. Timing simulations are performed under s. Each simulation process uses a constructed digital twin for iterative computation:

[0158] H1. Use the current flight status data and environmental scene information as the initial conditions for the digital twin.

[0159] H2. Input the control command at the current time step k in the control command sequence into the digital twin.

[0160] H3. Based on the dynamics model and the communication link model, calculate the flight status data (position, velocity, attitude angle, and communication status) for the next time step k+1.

[0161] H4. Calculate the remaining energy at the next time step k+1 based on the energy consumption model. .

[0162] H5. Based on the dynamic environment scenario model with safety constraints, check whether the path at the current time step k is feasible (whether it crosses a no-fly zone, whether it is below the safe altitude, and whether the attitude angle exceeds the limit).

[0163] H6. Repeat steps H2-H5 until the simulation ends.

[0164] S31. Based on the simulation results of all candidate emergency strategies, candidate emergency strategies that do not fall below the minimum safe flight altitude, do not cross no-fly zones, and whose attitude angles do not exceed the safe angle range are selected and summarized into a set of feasible candidate emergency strategies. The selection results of this embodiment are as follows, resulting in the trajectories of 3 feasible candidate emergency strategies:

[0165] Not feasible (encountering headwinds, predicted power depletion).

[0166] Feasible (sufficient power supply, acceptable wind field impact).

[0167] : Feasible (the nearest alternate landing point can be reached).

[0168] Feasible (avoiding strong winds, but the path is longer and time is increased).

[0169] like Figure 3 To obtain representative candidate emergency response strategies , and The trajectory. For example... Figure 4 To obtain feasible candidate emergency response strategies , and The remaining power consumption curve.

[0170] S32. Construct a comprehensive scoring function for candidate emergency strategies based on safety risk level, remaining power, and execution duration of candidate emergency strategies, thereby calculating the comprehensive score of each candidate emergency strategy in the set of feasible candidate emergency strategies.

[0171] In the embodiment, the comprehensive scoring weighting coefficient , and The coefficients were set to 0.6, 0.3, and 0.1 respectively, and the security risk weight coefficients were set to 0.4, 0.4, and 0.2 respectively. The comprehensive scores for each strategy were then calculated. , , The results show The highest overall score was due to its low risk, sufficient remaining battery power, and moderate execution time.

[0172] S33. Based on the comprehensive score, rank all feasible candidate emergency strategies and select the candidate strategy with the highest comprehensive score. As the optimal emergency response strategy.

[0173] S41. Based on the optimal emergency response strategy The control command sequence generates flight control commands that can be directly parsed and executed by the UAV flight control system, including target waypoint coordinates (return point). ), target height Economic speed Parameters such as these.

[0174] S42. Send the flight control command of the optimal emergency strategy to the UAV through the communication link, instructing the UAV to execute the economical low-speed return strategy according to the predetermined path.

[0175] S43. Continuously collect actual flight status data of the UAV during execution. The expected data of the optimal emergency response strategy in the simulation. Perform real-time comparison and calculate deviation. .

[0176] When position deviation Or attitude angle deviation Exceeding the preset threshold When this occurs, the policy correction process is triggered:

[0177] If the deviation is within the preset range, the original control command sequence will be smoothly fine-tuned (e.g., the heading will be slightly adjusted to return to the expected trajectory).

[0178] If the deviation exceeds the preset range, or if environmental conditions (such as wind speed) are detected... If a significant change occurs, the current real-time position, speed, attitude angle, remaining battery power, and current environmental scene information of the drone will be used as new initial conditions. The candidate emergency strategies that were not selected as optimal but ranked in the top two (e.g., ...) will be evaluated. and The timing simulation is performed, and a new optimal emergency strategy is selected as an alternative. Then, the modified flight control commands are generated and issued.

[0179] S44. After the optimal emergency strategy is executed (or after sufficient data is accumulated during execution), compare the actual flight status data collected throughout the execution process with the expected data from the corresponding simulation, and perform error analysis on the key model parameters in the digital twin. When the error obtained from the analysis exceeds the preset calibration threshold, update the key model parameters according to preset rules or optimization algorithms.

[0180] Comparative Example 1:

[0181] Drones use a traditional fixed rule: when the remaining battery power... Minimum return battery threshold At that time, execute " The method follows the "shortest path return" rule; it does not consider wind field distribution; it does not assess future power consumption; it does not check no-fly zones; it lacks multi-strategy simulation and optimization decision-making, etc. It employs traditional methods for execution. The shortest path return route revealed that the return route passes through a strong headwind area, significantly increasing the actual power demand; the energy model predicts that the power will drop to 0 before reaching the return point; the mission will inevitably fail, and there is even a risk of forced landing or crash.

[0182] 1. This invention utilizes a digital twin to perform parallel simulations of multiple candidate emergency strategies under UAV emergency events, and selects the optimal emergency strategy based on a comprehensive scoring function. It can comprehensively consider factors such as flight status, remaining battery power, environmental constraints, and communication status, overcoming the problems of single emergency response strategies and lack of forward-looking simulation and comprehensive evaluation capabilities in existing technologies, thereby improving the accuracy and safety of UAV emergency decision-making.

[0183] 2. This invention transforms the optimal emergency strategy into flight control commands for execution, and combines real-time deviation correction during execution and model calibration after execution to construct a closed-loop response mechanism from state perception, strategy deduction, decision output to execution feedback. This mechanism can improve the robustness, reliability, and subsequent deduction accuracy of UAV emergency response.

[0184] This invention utilizes a digital twin to simulate multiple candidate emergency strategies in parallel. By integrating factors such as flight status, remaining battery power, environmental constraints, and communication status, it selects the optimal solution using a comprehensive scoring function. This overcomes the shortcomings of traditional methods, such as limited strategy simplification and lack of forward-looking evaluation, thus improving the accuracy and safety of decision-making. Simultaneously, the optimal strategy is translated into flight control command execution, and combined with real-time deviation correction and model calibration, a closed-loop response mechanism is constructed from perception, simulation, decision-making to feedback, enhancing the robustness, reliability, and simulation accuracy of emergency response.

[0185] The above-described embodiments are merely preferred embodiments provided to fully illustrate the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are all within the scope of protection of the present invention. The scope of protection of the present invention is defined by the claims.

Claims

1. A method for unmanned aerial vehicle emergency event deduction and response based on digital twinning, characterized in that, Includes the following steps: S1. Construct a digital twin of the UAV, wherein the digital twin includes at least a three-dimensional structural model, a dynamic model, an energy consumption model, and an environmental scene model; S2. Acquire the drone's flight status data in real time, and determine whether to trigger an emergency event based on the flight status data and preset emergency triggering rules; When an emergency event is triggered, the current flight status data of the UAV and the information of the environmental scene model are used as initial conditions. The digital twin is used to perform parallel simulation and deduction of multiple pre-generated candidate emergency strategies to obtain the deduction results of each candidate emergency strategy. S3. Based on the preset comprehensive scoring function, evaluate the simulation results of all candidate emergency strategies and select the optimal emergency strategy from multiple candidate emergency strategies. S4. The optimal emergency strategy is converted into corresponding flight control commands and sent to the UAV for execution. During the execution process, the actual flight status data of the UAV is compared with the expected data under the optimal emergency strategy in real time, so as to continuously correct the flight control commands.

2. A method for unmanned aerial vehicle emergency event deduction and response based on digital twinning, characterized in that, Step S1 specifically involves: S11. Construct a three-dimensional structural model that includes the UAV's geometric and inertial parameters; S12. Construct a dynamic model of the UAV; S13. Construct an energy consumption model that predicts future changes in electricity consumption at a given time step; S14. Construct a communication link model that includes UAV signal strength attenuation and packet loss rate; S15. Construct a dynamic environmental scene model that includes a terrain model, a dynamic wind field model, and safety constraints; S16. The three-dimensional structural model, dynamic model, energy consumption model, communication link model and dynamic environment scene model are encapsulated to form a digital twin, so as to output the predicted UAV state sequence in response to the input of control command.

3. The digital-twin-based UAV emergency event deduction and response method according to claim 1, wherein, Step S2 specifically involves: S21. Real-time acquisition of flight status data of the UAV, including position, speed, attitude angle, remaining battery power and communication status data; S22. Calculate the drone's attitude yaw angle, lateral offset distance, packet loss rate, and communication interruption duration based on the real-time collected flight status data and the current mission track point index. Then, determine whether the drone has triggered an emergency event based on the attitude yaw angle, lateral offset distance, packet loss rate, communication interruption duration, and remaining battery power. S23. When the UAV triggers an emergency event, according to the type of emergency event, the corresponding set of candidate emergency strategies is obtained from the preset strategy library, and the set of candidate emergency strategies is parameterized into a set of control command sequences. S24. Using the UAV's flight status data and the information of the current dynamic environment scenario model as initial conditions, the digital twin is invoked according to the control command sequence to perform time-series simulation and deduction of each candidate emergency strategy in the acquired candidate emergency strategy set under a preset simulation duration. During the simulation and deduction process, the position, attitude angle and remaining power of the UAV under the current candidate emergency strategy are calculated at each time step, and the feasibility of the path at each time step is checked in real time, thereby obtaining the deduction results of the UAV including trajectory sequence, attitude change and power change.

4. The digital-twin-based UAV emergency event deduction and response method according to claim 3, characterized in that, The specific steps for determining whether a drone has triggered an emergency event are as follows: When the remaining power is lower than the minimum return power threshold or the power decrease rate exceeds the maximum decrease rate threshold within several consecutive sampling periods, it is determined to be a power shortage or high-risk power consumption emergency event. When the attitude yaw angle exceeds the yaw angle threshold or the lateral deviation distance of the track exceeds the lateral deviation distance threshold, it is determined as an attitude yaw or track deviation emergency event. When the packet loss rate exceeds the maximum allowable packet loss rate or the communication interruption duration reaches the maximum communication interruption duration, it is determined to be a communication anomaly emergency event.

5. The digital-twin-based UAV emergency event deduction and response method according to claim 1, wherein, Specifically, S3 is: S31. Based on the simulation results of all candidate emergency strategies, select candidate emergency strategies that do not fall below the minimum safe flight altitude, do not cross no-fly zones, and whose attitude angles do not exceed the safe angle range, and summarize them into a set of feasible candidate emergency strategies. S32. Construct a comprehensive scoring function for candidate emergency strategies based on safety risk level, remaining power, and execution duration of candidate emergency strategies, thereby calculating the comprehensive score of each candidate emergency strategy in the set of feasible candidate emergency strategies. S33. Based on the comprehensive score, sort the candidate emergency strategies in the set of all feasible candidate emergency strategies, and select the candidate emergency strategy with the highest comprehensive score as the optimal emergency strategy.

6. The method for simulated and responded to unmanned aerial vehicle (UAV) emergency events based on digital twins according to claim 5, characterized in that: The comprehensive score of each candidate emergency strategy in step S32 is obtained by processing it according to the following formula: Where j is the index; This represents the j-th candidate emergency strategy in the set of feasible candidate emergency strategies; Indicate candidate emergency strategies Overall score; , and All represent the overall score weighting coefficients; Indicate candidate emergency strategies Corresponding security risk level; Indicate candidate emergency strategies The remaining battery power at the end of the simulation; Indicates the rated capacity of the drone's battery; Indicate candidate emergency strategies The execution time consumed by the simulation; , and All represent safety risk weighting coefficients; Indicate candidate emergency strategies Flight time of the drone in high-wind areas during the execution; This indicates the minimum battery level threshold for the drone to return to base. Indicate candidate emergency strategies The average packet loss rate during the simulation process; Indicates the maximum allowable packet loss rate; Indicate candidate emergency strategies The cumulative communication interruption duration during the simulation process; Indicates the maximum allowed communication interruption duration; and These represent the weights for packet loss rate and communication interruption duration, respectively.

7. The method for simulated and responded to unmanned aerial vehicle (UAV) emergency events based on digital twins according to claim 1, characterized in that, Step S4 is as follows: S41. Generate flight control commands that can be directly parsed and executed by the UAV's flight control system based on the control command sequence corresponding to the optimal emergency strategy; S42. Send the flight control command of the optimal emergency strategy to the UAV through the communication link, instructing the UAV to execute the corresponding optimal emergency strategy; S43. During execution, continuously collect actual flight status data of the UAV, compare the actual flight status data with the expected data of the optimal emergency strategy in the simulation in real time, and when the deviation between the actual flight status data and the expected data exceeds the preset deviation threshold, trigger the strategy correction process, generate the corrected flight control command and issue it.

8. The method for simulated and responded to unmanned aerial vehicle (UAV) emergency events based on digital twins according to claim 7, characterized in that, The specific process for triggering policy correction is as follows: Based on the current deviation magnitude and environmental changes, the optimal emergency strategy currently being implemented can be fine-tuned locally; or, using the current UAV flight status data and dynamic environmental scenario model information as new initial conditions, time-series simulations can be performed on the top k candidate emergency strategies that were not selected as optimal, and a new optimal emergency strategy can be selected as an alternative.

9. The method for simulated and responded to unmanned aerial vehicle (UAV) emergency events based on digital twins according to claim 1, characterized in that, It also includes a model calibration step, specifically: By comparing the actual flight status data collected throughout the execution process with the expected data from the corresponding simulation, error analysis is performed on the parameters of at least one key model in the digital twin. When the error obtained from the analysis exceeds the preset calibration threshold, the parameters of the key model are updated according to the preset rules or optimization algorithm to calibrate the digital twin and improve the accuracy of subsequent inferences. The key models are the dynamics model, energy consumption model, or environmental scenario model in the digital twin.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 9.

Images