Airborne task planning and management system for small aircraft

Abstract

The invention relates to an aircraft autonomous control technology, and aims to realize mode conversion from passive risk response to active predictive risk avoidance. The core system comprises a state fusion module which collects and fuses multi-source information in real time and generates a standardized instantaneous task state; and the forward deduction module is used for performing Monte Carlo type deduction by using an airborne digital twin model to generate various possible state trajectories in the future. And the robustness evaluation module is used for quantitatively evaluating the future task risk based on the deduction result and predicting the comprehensive robustness. And the preventive intervention decision module is used for actively outputting a fine tuning instruction to the flight control system when the risk is predicted. The system can recognize potential crisis accumulated by multiple uncertainties in advance, grasp an optimal intervention window, change passive response into active management, and significantly improve the survivability of the unmanned aerial vehicle in a complex environment.

Description

Technical Field

[0001] This invention relates to the field of autonomous control technology for aircraft, specifically to an airborne mission planning and management system for small aircraft. Background Technology

[0002] When unmanned aerial vehicles (UAVs) perform missions, especially in highly dynamic and uncertain environments, their success and safety depend heavily on the continuous and effective management of multiple dimensions such as flight path, fuel, and external threats.

[0003] To address unforeseen circumstances during flight, existing technologies generally employ an event-triggered passive replanning mode. This mode monitors the aircraft's status in real time, and only initiates a mission replanning procedure to address the actual risk event when a critical parameter, such as fuel reserves or distance to a threat, exceeds a preset safety threshold.

[0004] However, this approach has significant drawbacks. First, its response is inherently delayed, only able to react passively after the risk has materialized, often missing the optimal intervention window. Second, this model lacks the ability to predict future risks and cannot effectively identify and mitigate potential crises resulting from the accumulation of multiple uncertainties. Finally, once replanning is triggered, it typically requires substantial global path adjustments, which not only incurs high computational costs but may also lead to reduced task efficiency and unnecessary fuel consumption; in complex environments, it may even fail to find a safe solution.

[0005] Therefore, how to shift from a passive and delayed event response model to a proactive management model that can predict and continuously maintain mission robustness in order to improve the mission success rate and survivability of aircraft in complex and uncertain environments has become an urgent technical problem to be solved. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides a forward-looking mission robustness maintenance system based on airborne digital twins. Specifically, the technical solution of this invention includes: The state fusion module is used to collect and fuse information from multiple heterogeneous sources in real time to generate a standardized multidimensional instantaneous task state vector. The forward inference module is used to perform multiple Monte Carlo forward inferences based on the instantaneous task state vector generated by the state fusion module and the preset airborne digital twin model, so as to generate a set of future state trajectories. The robustness assessment module is used to quantitatively assess task risks based on the set of future state trajectories generated by the forward inference module and a preset robustness boundary margin model, so as to generate comprehensive robustness prediction data. The preventive intervention decision module is used to determine and output preventive fine-tuning instructions to the flight control system when mission risks are predicted, based on the comprehensive robustness prediction data generated by the robustness assessment module.

[0007] Preferably, the comprehensive robustness prediction data includes: the robustness lower bound prediction curve and the state envelope prediction horizon.

[0008] Preferably, the robustness assessment module generates the comprehensive robustness prediction data, including: For each trajectory in the set of future state trajectories, calculate its robustness boundary margin curve over future time series; Based on all robustness boundary margin curves, statistical quantiles are calculated to determine the robustness lower bound prediction curve; The state envelope prediction horizon is determined based on the robustness lower bound prediction curve and the preset minimum robustness threshold. Preferably, the calculation of the robustness boundary margin curve includes: Define the fuel margin function and the threat exposure margin function as sub-margin functions; The robustness boundary margin is calculated by using a multiplicative model based on the weighted geometric average concept, combining each of the sub-margin functions with preset weight coefficients. Preferably, the preventive intervention decision module is specifically used for: Continuously monitor the state envelope prediction horizon and the robustness lower bound prediction curve; When the state envelope prediction horizon is less than the preset minimum safe horizon threshold, or when it is predicted that the robustness lower bound prediction curve will be lower than the preset minimum robustness threshold in the future, preventive intervention is triggered. Preferably, after the preventive intervention decision module triggers a preventive intervention, it is further used to: Based on the principle of minimizing disturbances, candidate preventive fine-tuning instructions are searched on the current route. The preventive fine-tuning instruction is determined from the candidate preventive fine-tuning instructions that enables the new prediction robustness margin after applying the instruction to be maintained above the minimum robustness threshold throughout the entire prediction horizon. Preferably, the forward inference module is specifically used for: Instantiate the airborne digital twin model containing elements of aircraft dynamics, aerodynamics, and environmental dynamics; Random perturbations are introduced into the environmental uncertainty parameters, and large-scale parallel Monte Carlo forward extrapolation is performed to generate the set of future state trajectories. Preferably, the fuel margin function is used to measure whether, under a certain state, the current fuel is sufficient to complete the remaining tasks and maintain a minimum safety reserve; the threat exposure margin function is used to measure the minimum safe distance between the aircraft and all known threats. Preferably, the weighting coefficient is used as an index to adjust the sensitivity of the robustness boundary margin to changes in each of the sub-margins.

[0009] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention achieves a shift from passive response to proactive, predictive risk avoidance. Addressing the shortcomings of existing technologies that only respond after a risk has occurred, this invention utilizes airborne digital twins for large-scale forward simulations far exceeding real-time capabilities, generating multi-dimensional predictions for the future. This allows the system to identify potential crises arising from multiple uncertainties before new risks materialize, thus seizing the optimal intervention window, transforming passive response into proactive management, and significantly enhancing the survivability of unmanned aerial vehicles in highly dynamic environments. 2. Improved sensitivity and reliability of risk assessment: To overcome the shortcomings of traditional weighted summation models that may mask fatal risks in a single dimension, this invention innovatively adopts a multiplicative model based on the weighted geometric average concept to calculate comprehensive robustness. This method follows the "weakest link" principle, ensuring that any sharp deterioration in any critical sub-margin (such as fuel or threat exposure) will be significantly amplified and reflected in the final comprehensive risk index, thus making risk assessment more sensitive and reliable. 3. It achieves efficient and low-cost continuous task fine-tuning; unlike traditional solutions that involve computationally expensive and potentially inefficient global path replanning after triggering, this invention, when a risk is predicted, prioritizes minimizing disturbance and searches for and executes preventative fine-tuning instructions only based on the current route. By rapidly and proactively verifying candidate instructions, the effectiveness and safety of intervention actions are ensured, maintaining task robustness at minimal cost and avoiding unnecessary fuel consumption and time waste. 4. Enhanced robustness of decision-making in environments with high uncertainty: Faced with strong environmental uncertainty, this invention systematically explores the space of future possibilities by introducing random perturbations and conducting Monte Carlo parallel extrapolation. The decision-making basis is not a single optimal prediction, but rather a pessimistic prediction (robust lower bound) with high confidence derived from statistical analysis of massive extrapolation results. This decision-making approach, based on statistically high probability safety, ensures that the system can make more robust and reliable judgments in complex and ever-changing real-world environments. Attached Figure Description

[0010] The present invention will be further explained below with reference to the accompanying drawings and embodiments: Figure 1 This is a structural diagram of the system of the present invention. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0012] Example 1: Please see Figure 1 A forward-looking mission robustness maintenance system based on airborne digital twins includes: a state fusion module for real-time acquisition and fusion of information from multiple heterogeneous sources to generate a standardized multidimensional instantaneous mission state vector; a forward extrapolation module for performing multiple Monte Carlo forward extrapolations based on the instantaneous mission state vector generated by the state fusion module and a preset airborne digital twin model to generate a set of future state trajectories; a robustness assessment module for quantitatively assessing mission risks based on the set of future state trajectories generated by the forward extrapolation module and a preset robustness boundary margin model to generate comprehensive robustness prediction data; and a preventive intervention decision module for determining and outputting preventive fine-tuning commands to the flight control system when mission risks are predicted, based on the comprehensive robustness prediction data generated by the robustness assessment module.

[0013] In a preferred embodiment of the invention, the system is deployed in the onboard edge computing unit of an unmanned aerial vehicle (UAV) to address highly dynamic and uncertain mission environments. The system's four core modules work collaboratively to form a complete closed loop of perception-prediction-evaluation-decision-control. The state fusion module aims to provide unified, real-time, and accurate data input for the entire system. This module acquires heterogeneous information in real time from GPS, IMU, fuel sensors, radar, weather sensors, and the mission planning database via the onboard data bus. This information includes the aircraft's three-dimensional position, velocity, attitude, and remaining fuel. As well as known enemy radar locations and dynamic characteristics, wind field data, preset waypoints, mission time windows, and minimum reserve fuel. The module incorporates a Kalman filter algorithm to clean and smooth the collected data, align the timestamps, and finally fuse them into a standardized multidimensional instantaneous task state vector. This vector serves as the sole initial condition for all subsequent analyses, ensuring that system decisions are based on an accurate snapshot of the current physical world. The forward inference module aims to transform task management from a passive, reactive approach to a proactive, pre-emptive one. This module receives real-time updates from the state fusion module. Immediately afterwards, this is used as the initial condition to instantiate a high-fidelity airborne digital twin model. Subsequently, the module performs... Sub-massive parallel Monte Carlo forward inference generates a set of future state trajectories. This collection constitutes a systematic exploration of the space of future possibilities; The robustness assessment module aims to transform the raw, high-dimensional set of future state trajectories generated by the forward extrapolation module into an intuitive and decision-oriented risk metric. After receiving the trajectory set, this module uses a pre-defined robustness boundary margin model to assess the risk of each trajectory at each future time point. Calculations were performed to obtain A separate robust evolution curve. Subsequently, the module... Statistical analysis was performed on the curves to generate comprehensive robust prediction data; The preventative intervention decision-making module serves as the decision-making and execution unit of this invention, making decisions and taking action based on risk prediction. This module continuously monitors the comprehensive robustness prediction data generated by the robustness assessment module; when this data indicates an unacceptable risk in the future, the module does not wait for the disruptive event to actually occur but immediately initiates the intervention procedure. Based on the principle of minimizing disturbances, it searches for an optimal preventative fine-tuning instruction based on the current flight path. For example, the heading is deflected 2 degrees to the left, and the climb is 50 meters. Once this instruction is determined, it is immediately sent to the flight control system via the flight control interface for execution, thereby proactively avoiding risks before they materialize. This embodiment constructs a complete, forward-looking mission robustness maintenance system through the close collaboration of the four modules described above. It transforms the traditional event-triggered, passive replanning model into a continuous state envelope-maintaining mission fine-tuning model. This system proactively predicts and mitigates potential risks, rather than passively waiting for disruptive events to occur. This achieves a shift from discrete, delayed responses to continuous, forward-looking maintenance, significantly improving the mission success rate and survivability of aircraft in complex and uncertain environments.

[0014] Example 2: The comprehensive robustness prediction data includes: the robustness lower bound prediction curve and the state envelope prediction horizon.

[0015] In this embodiment, the technical solution of Embodiment 1 is further specified. The comprehensive robustness prediction data output by the robustness assessment module is not a single risk value, but is composed of two interrelated core indicators to provide richer and more valuable risk information for decision-making. Robustness lower bound prediction curve The curve indicates each future time point. For all Robustness boundary margin of the extrapolated trajectory After statistically sorting the values, a specific percentile, such as the 5th percentile, is taken to form a curve. Its purpose is to provide a robust worst-case prediction with high confidence, effectively addressing the uncertainties in Monte Carlo simulations. Its source is the curve calculated by the robustness assessment module. strip The curve was obtained through statistical processing. By focusing on the 5th percentile rather than the mean, the system's decision-making will be based on a 95% probability that a worse outcome will not occur, thus exhibiting a more conservative and safer behavior. State envelope predicts horizon This indicator refers to the maximum future time for which the predicted robustness lower bound can be maintained above the acceptable minimum robustness threshold. It is derived from the robustness lower bound prediction curve. With a preset minimum robustness threshold This minimum robustness threshold is calculated. Refers to a pre-defined, system-acceptable A minimum value, such as 0.3, is set below which the task is considered to have entered an unacceptable risk state. Its function is to serve as a baseline for judging whether the task's robustness meets the standard. A specific calibration method includes the following steps: First, run thousands of Monte Carlo simulations containing both successful and failed task scenarios, recording all task data; second, filter out all failed task cases and extract the RBM value sequence for each failed case within a fixed time window (e.g., 30 seconds) before the disruptive event occurs; finally, construct a statistical sample set from all extracted RBM values, calculate a specific low percentile (e.g., 5th or 10th percentile) of this sample set, and use this statistical value as the minimum robustness threshold. ; The computational logic is to find future times. The maximum value of , such that all from the current point in time to that point in time. All greater than ; By concretizing comprehensive robustness prediction data into robustness lower bound prediction curves and state envelope prediction horizons, this invention achieves a deep characterization of future risks. The former provides a pessimistic estimate of the degree of risk, while the latter transforms risk into an intuitive safety time window. This combination makes the decision-making basis of the preventive intervention module more comprehensive and robust, knowing not only how dangerous it is, but also when the danger will arrive, thus enabling more precise and timely intervention and resulting in a significant improvement in decision accuracy.

[0016] Example 3: The robustness assessment module generates comprehensive robustness prediction data, including: calculating the robustness boundary margin curve of each trajectory in the future state trajectory set over future time series; calculating statistical quantiles based on all robustness boundary margin curves to determine the robustness lower bound prediction curve; and determining the state envelope prediction horizon based on the robustness lower bound prediction curve and a preset minimum robustness threshold.

[0017] This embodiment further illustrates the process of generating comprehensive robustness prediction data in Embodiment 2, which is executed sequentially within the robustness assessment module; In one processing flow, the robustness assessment module receives a set of future state trajectories generated by the forward inference module. The module will iterate through every trajectory in the set. ...each state trajectory represents a future possibility. Each of these is mapped to a corresponding robustness boundary margin curve. It should be noted that the predictive horizon here... It refers to the total future duration simulated in a single inference process, set by the forward inference module. It is a fixed input parameter, distinct from the state envelope prediction horizon, which is the output of the system calculation. After completing this step, the system will receive... strip curve; In obtaining After establishing independent robustness boundary margin curves, the module at each discrete future time point... Up, will this indivual value The data is collected to form a sample set. The module sorts this sample set and calculates its statistical quantiles; in this embodiment, the 5th percentile is used. All time points are then considered. Connecting the calculated 5th percentiles forms the final robustness lower bound prediction curve. ; The module predicts based on the robustness lower bound curve generated in the previous step. and the preset minimum robustness threshold To determine the state envelope and predict the horizon. The specific algorithm is as follows: starting from the current moment... Begin, check The value of. If Then continue checking the next time point; if If the check stops, the previous time point will be used as the reference. The value of . Right now Continue to maintain The maximum duration above; This embodiment, by clearly defining the above three-step calculation process, ensures the feasibility and logical rigor of the comprehensive robustness prediction data generation process. It effectively converges the uncertainty generated by Monte Carlo simulations into a high-confidence pessimistic prediction using statistical quantiles, and further transforms this state prediction into an intuitive time-based early warning metric. This approach enables the system to make decisions based on statistically high probabilities of safety, greatly enhancing the robustness and reliability of the decisions. Example 4: The calculation of the robustness boundary margin curve includes: defining the fuel margin function and the threat exposure margin function as sub-margin functions; and using a multiplication model based on the weighted geometric average idea, combining each sub-margin function with a preset weight coefficient to calculate the robustness boundary margin.

[0018] The fuel margin function measures whether, under a given condition, the current fuel is sufficient to complete the remaining tasks and maintain a minimum safety reserve; the threat exposure margin function measures the minimum safe distance between the aircraft and all known threats. Weighting coefficients, used as an index, are sensitivity coefficients of the robustness boundary margin to changes in each sub-margin. This embodiment addresses the robustness boundary margin in Embodiment 3. The core calculation method is described in detail. To transform the abstract concept of task safety into a computable quantifiable metric, this invention constructs... Mathematical model; To implement this model, sub-margin functions are defined for several key dimensions that play a decisive role in the success or failure of the task. In this embodiment, at least two core sub-margin functions are included: Fuel margin function The purpose of this function is to measure the state given. Given the following, how much fuel does the aircraft currently have relative to the minimum requirements for completing the mission and ensuring a safe return? The calculation logic is as follows: .in, Current state The amount of fuel is provided by an onboard fuel sensor. From the current position The minimum fuel required for return is calculated in real time by the aircraft dynamics model; It is the minimum reserve of fuel that must be maintained, a constant value stipulated by law or mission requirements; This is the planned additional reserve fuel, used as a normalization baseline; in this embodiment, It is set to a positive positive number to ensure the validity of the calculation; the clip function is a clipping function, which restricts the calculation result to a specified closed interval, i.e. Indicates if The result is ;like The result is Otherwise, the result is The clip function restricts the calculation result to... The range; the function outputs a value of 1 to indicate that there is plenty of fuel, and a value of 0 to indicate that there is not enough fuel to safely return. Threat Exposure Margin Function The purpose of this function is to measure the degree of spatial isolation between an aircraft and all known threats, i.e., the safety level of distance threats; its specific normalization calculation logic can be expressed by the following formula: .in, For aircraft and the first The distance to a threat can be measured in real time using airborne sensing systems such as radar. For the first The inherent danger radius of a threat is derived from the threat database input before the mission; This is a predetermined safety buffer distance, the value of which is determined to ensure that the aircraft has sufficient maneuvering space for avoidance. In this embodiment, It is also set to a positive positive number to avoid calculation errors. The function outputs a value of 1 to indicate that the area is far from all threats, and a value of 0 to indicate that the safe buffer distance has been breached. Based on the aforementioned sub-margin function, this embodiment abandons the traditional weighted summation model because it suffers from risk masking. To reflect the principle of the weakest link, this embodiment innovatively employs a multiplicative model based on the weighted geometric average concept to calculate the overall robustness boundary margin. The purpose of this model is to ensure that a sharp deterioration in any sub-proportion will lead to a decline in the overall... The value drops sharply. The calculation formula is as follows: ; in, The value of the i-th sub-margin function, for example or It is a dimensionless floating-point number with a range of 1 / 2π. Its source is obtained by normalizing specific physical quantities within a safe operating range; The range of values ​​is And satisfy Its source is the preset value injected during the mission planning phase, or it can be dynamically adjusted according to the mission phase. For example, in a typical reconnaissance mission, the weighting coefficient can be dynamically adjusted according to the following rules: During the "entry" phase towards the target area, fuel consumption is the main issue, and at this time, it can be set... , During the "reconnaissance operations over the target area" phase, avoiding potential threats is the primary task, and the weighting can be adjusted accordingly. , During the "return" phase, ensuring sufficient fuel for a safe return becomes crucial again, and its weight can be restored to [previous value]. , Other weights are adjusted accordingly to ensure the sum is 1; in this model, As an index, its physical meaning is defined as For the i-th sub-margin Sensitivity coefficient to change. The larger the value, the greater the impact of a small change in that sub-margin on the overall [macro-margin]. The value has a more dramatic impact, indicating that this dimension is more critical at the current stage of the task; : The final calculated robustness boundary margin. Since all All in The interval, and the result is also in An interval is a dimensionless comprehensive robustness indicator; By defining specific sub-margin functions, this invention provides a clear physical basis for robustness assessment. More importantly, one of the core innovations of this invention is the adoption of a multiplicative model based on weighted geometric averages and the introduction of weighting coefficients as sensitivity coefficients. This model addresses the risk masking defect of traditional weighted summation models, realizing the assessment principle of the weakest link. That is, any fatal risk in any single dimension can be significantly amplified in the final comprehensive index, thereby greatly improving the sensitivity and reliability of risk assessment and generating key technical gains.

[0019] Example 5: The preventive intervention decision module is specifically used to: continuously monitor the state envelope prediction horizon and the robustness lower bound prediction curve; when the state envelope prediction horizon is less than the preset minimum safe horizon threshold, or when it is predicted that the robustness lower bound prediction curve will be lower than the preset minimum robustness threshold in the future, preventive intervention is triggered.

[0020] This embodiment describes in detail the specific conditions under which the preventive intervention decision-making module in Embodiment 2 triggers intervention, demonstrating the rigor of its decision-making logic; The core task of this module is to continuously monitor two key metrics output by the robustness assessment module: the state envelope prediction horizon. and robustness lower bound prediction curve The module internally sets two corresponding trigger thresholds: Minimum safe view threshold This threshold refers to the minimum level that the system must maintain to ensure sufficient reaction time. Length. Its function is to define a final decision point in time, ensuring the system does not fall into danger due to delayed response. It is derived from the aircraft's maneuverability and the system's preventative intervention delay. A comprehensive determination is needed. For example, if the system takes 30 seconds from decision-making to completing the avoidance action, then... It may be set to 60 seconds to allow for ample margin; Minimum robustness threshold The definition and source of this threshold have been detailed in the implementation of Example 2; The triggering logic for preventative intervention consists of two parallel conditions; either condition must be met to trigger the intervention: Time-based alerts: When the calculated state envelope predicts the horizon. Shorten to less than the preset minimum safety field threshold. The system immediately triggers intervention. This indicates that while the current state is acceptable, a situation is predicted to worsen within a short period of time. Within a certain timeframe, the task's robustness is about to fall below an acceptable level. This is a trigger mechanism based on time urgency. Based on state deterioration: predicted At some point in the future Will pass under Even the current Still greater than However, if the preventive intervention decision module analyzes the shape of the robustness lower bound prediction curve—for example, if its downward slope is too large—it predicts that the curve will soon intersect the minimum robustness threshold at a certain point within the prediction horizon. Intersections can also trigger intervention earlier. This is a more proactive triggering mechanism based on the trend of state deterioration; This embodiment greatly improves the intelligence and timeliness of intervention decisions by setting dual, complementary triggering conditions. Based on The trigger ensures the system has a minimum reaction time, serving as a bottom-line defense; while based on Predictive triggering of deteriorating trends enables earlier identification and response to risks, representing proactive defense. This combination allows the system to intervene at the optimal time, avoiding route disruptions and fuel waste caused by premature intervention, and preventing missed avoidance windows due to late intervention, thus achieving a better balance between maintaining mission efficiency and ensuring safety.

[0021] Example 6: After the preventive intervention decision module triggers preventive intervention, it is also used to: search for candidate preventive fine-tuning instructions based on the current route with the principle of minimizing disturbance; and determine the preventive fine-tuning instructions that can keep the new forecast robustness margin above the minimum robustness threshold throughout the entire forecast horizon.

[0022] This embodiment further describes how the preventive intervention decision module calculates the specific optimal intervention action after the conditions in Embodiment 5 are triggered; Once preventative intervention is triggered, the module immediately enters the optimization calculation phase. Its core objective is not to perform disruptive global replanning, but to find a solution that minimizes the disturbance to the current task. In an optimized computational flow, the module, based on the current flight path plan and adhering to the principle of minimizing disturbances, generates a finite set of candidate preventative fine-tuning instructions. These instructions are typically small, discrete adjustments to control variables, such as: {heading +2°, heading -2°, speed +5m / s, speed -5m / s, altitude +50m, altitude -50m}. This approach avoids complex continuous spatial optimization and is highly suitable for resource-constrained airborne platforms. The module virtually executes and evaluates each candidate preventative fine-tuning instruction. Specifically, for each candidate instruction, the module uses it as a new input and returns it to the forward inference module to quickly perform a new round of forward inference and robustness evaluation, thereby obtaining the new prediction robustness lower bound prediction curve after applying the instruction. ; The module filters the new prediction curves generated by all candidate instructions. The filtering criterion is: the instruction corresponds to... Is it within the entire prediction horizon, i.e., for all They can all be stably maintained at the minimum robustness threshold. Above, the module will select the candidate instruction with the smallest perturbation or the smallest comprehensive cost function as the final preventative fine-tuning instruction. and output it to the flight control system; The intervention decision-making mechanism described in this embodiment has significant benefit effects. The principle of minimizing perturbations ensures the continuity and efficiency of the task, avoiding fuel waste and extended task time caused by frequent and large-scale replanning. By performing rapid forward-looking verification of candidate instructions, it ensures that the final output fine-tuning instructions are indeed effective and safe, avoiding blind decision-making. This generation-verification closed-loop optimization process enables the system to address potential risks at minimal cost while ensuring safety, achieving both accuracy and economy in decision-making.

[0023] Example 7: The forward inference module is specifically used to: instantiate an airborne digital twin model containing elements of aircraft dynamics, aerodynamics, and environmental dynamics; introduce random perturbations into environmental uncertainty parameters and perform large-scale parallel Monte Carlo forward inference to generate a set of future state trajectories.

[0024] This embodiment provides a detailed explanation of the internal working mechanism of the forward inference module in Embodiment 1; For simulation purposes, the instantiated airborne digital twin model in this module is a high-fidelity set of mathematical and physical models designed to simulate the future behavior of the aircraft in a computational environment as accurately as possible. This model comprises at least three core components: Aircraft dynamics model: describes the six-degree-of-freedom motion response of an aircraft under various control inputs; Aerodynamic model: Calculates the lift and drag forces acting on an aircraft based on its speed, altitude, attitude, etc. Environmental dynamics model: Models the dynamic factors in the task environment. For example, for known mobile threats, the model will include their motion model; for meteorology, the model will include a stochastic model of wind field changes over time and space. To simulate real-world uncertainties, the module introduces random perturbations into environmental uncertainty parameters during simulations. Environmental uncertainty parameters refer to variables in the model that cannot be precisely known or are inherently random, such as the future location of enemy radar, fluctuations in wind intensity and direction, etc. Their purpose is to ensure that the simulation results cover a broader range of possibilities, avoiding planning vulnerabilities caused by relying on a single deterministic prediction. These parameters are derived from probability distributions, such as Gaussian distributions, established based on historical data, sensor measurement errors, or expert experience. Before each simulation begins, the system randomly samples from these probability distributions and assigns a specific value to the uncertainty parameter. Based on this model, the module utilizes the multi-core processing power of the onboard edge computing unit to perform large-scale parallel Monte Carlo forward inference. This means The independent extrapolation processes are initiated and computed simultaneously, with each extrapolation using a different set of random perturbation parameters. This parallel processing significantly reduces the time required to generate the set of future state trajectories, ensuring that the entire system meets real-time requirements. This embodiment delivers significant technical gains by specifying the implementation method of the forward inference module. A high-fidelity airborne digital twin model ensures the accuracy of the inference results. Introducing random perturbations into environmental uncertainty parameters and employing a massively parallel Monte Carlo method are key to achieving both foresight and robustness in this invention. This allows the system to move beyond relying on a single, fragile optimal prediction, enabling it to understand and assess multiple future possibilities, especially low-probability but high-risk events. This provides a solid data foundation for subsequent robustness assessments and decision-making, significantly enhancing the system's ability to cope with the complexity and uncertainty of the real world.

[0025] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A forward-looking mission robustness maintenance system based on airborne digital twins, characterized in that, include: The state fusion module is used to collect and fuse information from multiple heterogeneous sources in real time to generate a standardized multidimensional instantaneous task state vector. The forward inference module is used to perform multiple Monte Carlo forward inferences based on the instantaneous task state vector generated by the state fusion module and the preset airborne digital twin model, so as to generate a set of future state trajectories. The robustness assessment module is used to quantitatively assess task risks based on the set of future state trajectories generated by the forward inference module and a preset robustness boundary margin model, so as to generate comprehensive robustness prediction data. The preventive intervention decision module is used to determine and output preventive fine-tuning instructions to the flight control system when mission risks are predicted, based on the comprehensive robustness prediction data generated by the robustness assessment module.

2. The forward-looking mission robustness maintenance system based on airborne digital twin as described in claim 1, characterized in that, The comprehensive robustness prediction data includes: the robustness lower bound prediction curve and the state envelope prediction horizon.

3. The forward-looking mission robustness maintenance system based on airborne digital twin as described in claim 2, characterized in that, The robustness assessment module generates the comprehensive robustness prediction data, including: For each trajectory in the set of future state trajectories, calculate its robustness boundary margin curve over future time series; Based on all robustness boundary margin curves, statistical quantiles are calculated to determine the robustness lower bound prediction curve; The state envelope prediction horizon is determined based on the robustness lower bound prediction curve and the preset minimum robustness threshold.

4. A forward-looking mission robustness maintenance system based on airborne digital twins according to claim 3, characterized in that, The calculation of the robustness boundary margin curve includes: Define the fuel margin function and the threat exposure margin function as sub-margin functions; The robustness boundary margin is calculated by using a multiplication model based on the weighted geometric average concept, combining each sub-margin function with a preset weight coefficient.

5. A forward-looking mission robustness maintenance system based on airborne digital twins according to claim 2, characterized in that, The preventive intervention decision-making module is specifically used for: Continuously monitor the state envelope prediction horizon and the robustness lower bound prediction curve; Preventive intervention is triggered when the predicted horizon of the state envelope is less than the preset minimum safe horizon threshold, or when it is predicted that the predicted robustness lower bound curve will fall below the preset minimum robustness threshold in the future.

6. A forward-looking mission robustness maintenance system based on airborne digital twins according to claim 5, characterized in that, After the preventive intervention decision module triggers a preventive intervention, it is also used for: Based on the principle of minimizing disturbances, candidate preventive fine-tuning instructions are searched on the current route. The preventive fine-tuning instruction is identified as the instruction that enables the new prediction robustness margin after the instruction is applied to remain above the minimum robustness threshold throughout the entire prediction horizon.

7. A forward-looking mission robustness maintenance system based on airborne digital twins according to claim 1, characterized in that, The forward inference module is specifically used for: Instantiate the airborne digital twin model containing elements of aircraft dynamics, aerodynamics, and environmental dynamics; Random perturbations are introduced into the environmental uncertainty parameters, and large-scale parallel Monte Carlo forward extrapolation is performed to generate the set of future state trajectories.

8. A forward-looking mission robustness maintenance system based on airborne digital twins according to claim 4, characterized in that, The fuel margin function measures whether, under a given condition, the current fuel is sufficient to complete the remaining tasks and maintain a minimum safety reserve; the threat exposure margin function measures the minimum safe distance between the aircraft and all known threats.

9. A forward-looking mission robustness maintenance system based on airborne digital twins according to claim 4, characterized in that, The weighting coefficients serve as an index to adjust the sensitivity of the robustness boundary margin to changes in each of the sub-margins.

Images