Intelligent audit dynamic collaboration method and device based on risk grading and electronic equipment

Abstract

This application discloses a method, apparatus, and electronic device for intelligent collaborative auditing based on risk grading. This solution acquires flight plan data to be audited; performs multi-dimensional risk assessment on the flight plan data to determine a comprehensive risk index, which characterizes the overall risk level of the flight plan data across multiple dimensions; determines the corresponding human-machine collaboration depth level based on the comprehensive risk index; selects a matching target audit strategy from multiple preset human-machine collaboration audit strategies based on the human-machine collaboration depth level, where different human-machine collaboration audit strategies correspond to different levels of manual audit intensity; and executes the target audit strategy to complete the audit of the flight plan data. This achieves automated and intelligent operation of intelligent collaborative auditing based on risk grading, thereby improving the efficiency and effectiveness of flight plan auditing.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, and electronic device for intelligent collaborative auditing based on risk classification. Background Technology

[0002] With the rapid development of the low-altitude economy, low-altitude flight activities such as drones and general aviation are becoming increasingly frequent, leading to an explosive growth in the number of low-altitude flight plan approvals. This places higher demands on approval efficiency, review accuracy, and operational safety. To improve approval efficiency, existing low-altitude control systems are gradually introducing artificial intelligence technologies, such as intelligent review technology based on Large Language Models (LLM), which can assist in approval decisions.

[0003] The existing low-altitude flight plan review mainly includes three methods: traditional manual review, manual intervention in fixed stages, and fully automated approval. However, the existing methods generally suffer from problems such as rigid review mechanisms, unbalanced resource allocation, insufficient safety guarantees, and poor intelligent adaptability, which cannot meet the needs of large-scale and differentiated low-altitude flight approval management. Summary of the Invention

[0004] The purpose of this application is to propose a risk-based intelligent dynamic collaboration method for auditing, which aims to solve the problems of high dependence on manual intervention and low efficiency in existing risk-based intelligent dynamic collaboration methods.

[0005] This application provides a dynamic collaborative intelligent auditing method based on risk classification, the method including: Obtain flight plan data pending review; A multi-dimensional risk assessment is conducted on flight plan data to determine a comprehensive risk index. The comprehensive risk index is used to characterize the overall risk level of flight plan data across multiple dimensions. The corresponding level of human-machine collaboration depth is determined based on the comprehensive risk index. Based on the depth level of human-machine collaboration, a matching target review strategy is determined from multiple preset human-machine collaboration review strategies. Different human-machine collaboration review strategies correspond to different human review intensities. Implement the target review strategy to complete the review of flight plan data.

[0006] Accordingly, this application also provides a risk-based intelligent audit dynamic collaborative system, the system comprising: The data acquisition module is used to acquire flight plan data; The risk pre-assessment module is used to construct risk pre-assessment prompts and guide the risk assessment model based on the risk pre-assessment prompts to conduct multi-dimensional risk assessments of flight plan data and generate a comprehensive risk index. The risk grading module is used to determine the depth level of human-machine collaboration based on the comprehensive risk index. The collaboration scheme selection module is used to select the corresponding human-machine collaboration review strategy based on the depth level of human-machine collaboration. The collaboration scheme execution module is used to execute human-machine collaboration review strategies; The audit process recording module is used to record the audit process and results, forming an audit log.

[0007] Accordingly, this application also provides a risk-based intelligent audit dynamic collaborative device, including: The acquisition unit is used to acquire flight plan data to be reviewed; The assessment unit is used to conduct multi-dimensional risk assessments of flight plan data and determine the comprehensive risk index. The comprehensive risk index is used to characterize the overall risk level of flight plan data across multiple dimensions. The first determining unit is used to determine the corresponding human-machine collaboration depth level based on the comprehensive risk index. The second determining unit is used to determine the matching target review strategy from multiple preset human-machine collaboration review strategies based on the depth level of human-machine collaboration, wherein different human-machine collaboration review strategies correspond to different human review intensities. The execution unit is used to execute the target review strategy to complete the review of flight plan data.

[0008] Accordingly, this application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the program, implements the above-described intelligent audit dynamic collaborative method based on risk classification.

[0009] Accordingly, this application also provides a computer-readable storage medium storing a plurality of instructions adapted for loading by a processor to execute the above-described risk-based intelligent audit dynamic collaboration method.

[0010] This application acquires flight plan data to be reviewed; conducts multi-dimensional risk assessment of the flight plan data to determine a comprehensive risk index, which characterizes the overall risk level of the flight plan data across multiple dimensions; determines the corresponding human-machine collaboration depth level based on the comprehensive risk index; selects a matching target review strategy from multiple preset human-machine collaboration review strategies based on the human-machine collaboration depth level, wherein different human-machine collaboration review strategies correspond to different levels of manual review intensity; and executes the target review strategy to complete the review of the flight plan data. This achieves automated and intelligent operation of risk-level-based intelligent review dynamic collaboration, thereby improving the efficiency and effectiveness of flight plan review. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] in: Figure 1 This is a schematic diagram of a risk-based intelligent audit dynamic collaborative system provided in an embodiment of this application.

[0013] Figure 2 This is a flowchart illustrating a risk-based intelligent audit dynamic collaboration method provided in an embodiment of this application.

[0014] Figure 3 This is a flowchart illustrating another intelligent audit dynamic collaboration method based on risk classification provided in an embodiment of this application.

[0015] Figure 4 This is a flowchart illustrating another intelligent collaborative auditing method based on risk classification provided in an embodiment of this application.

[0016] Figure 5 This is a flowchart illustrating another intelligent audit dynamic collaboration method based on risk classification provided in an embodiment of this application.

[0017] Figure 6 This is a structural block diagram of a risk-based intelligent audit dynamic collaborative device provided in an embodiment of this application.

[0018] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] This application provides a risk-based intelligent audit dynamic collaboration method, apparatus, computer-readable storage medium, and electronic device. Specifically, the risk-based intelligent audit dynamic collaboration method of this application can be executed by an electronic device, which can be a terminal or a server. The server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, and big data and artificial intelligence platforms.

[0021] To address the aforementioned issues, this application provides a method, apparatus, and electronic device for intelligent review and dynamic collaboration based on risk grading. This enables automated and intelligent operation of intelligent review and dynamic collaboration based on risk grading, thereby improving the efficiency and effectiveness of flight plan review.

[0022] The following sections provide detailed descriptions of each example. It should be noted that the order in which the embodiments are described is not intended to limit the preferred order of the embodiments.

[0023] This application provides an intelligent audit dynamic collaborative system based on risk classification, the system comprising: The data acquisition module is used to acquire flight plan data; The risk pre-assessment module is used to construct risk pre-assessment prompts and guide the risk assessment model based on the risk pre-assessment prompts to conduct multi-dimensional risk assessments of flight plan data and generate a comprehensive risk index. The risk grading module is used to determine the depth level of human-machine collaboration based on the comprehensive risk index. The collaboration scheme selection module is used to select the corresponding human-machine collaboration review strategy based on the depth level of human-machine collaboration. The collaboration scheme execution module is used to execute human-machine collaboration review strategies; The audit process recording module is used to record the audit process and results, forming an audit log.

[0024] For example, please see Figure 1 , Figure 1 This is a schematic diagram of a risk-based intelligent audit dynamic collaborative system provided in an embodiment of this application.

[0025] This application provides a risk-based intelligent auditing dynamic collaboration system, including a data acquisition module, a risk pre-audit module, a risk classification module, a collaboration scheme selection module, a collaboration scheme execution module, and an auditing process recording module. Specifically, the functions of each module are as follows: The data acquisition module is configured to acquire multi-source data required for flight plan review, which may include flight plan data, airspace data, meteorological data, aircraft data, applicant data, etc. The data acquisition module can connect to external data sources (such as airspace management systems, meteorological service platforms, aircraft registration databases, etc.) via API (Application Programming Interface) to achieve automatic data acquisition and updating.

[0026] The risk pre-assessment module may include a prompt word construction submodule, a model invocation submodule, and a risk index extraction submodule; The prompt word construction submodule is used to construct risk pre-screening prompt words to guide the risk assessment model in conducting risk assessments. The model invocation submodule is used to input the risk pre-screening prompts and flight plan data (preferably in JSON format) obtained by the data acquisition module into the risk assessment model, invoke the risk assessment model to perform risk assessment, and generate a comprehensive risk index; The risk index extraction submodule is used to extract the comprehensive risk index from the output of the risk assessment model.

[0027] The risk grading module is configured to determine the corresponding human-machine collaboration depth level based on the comprehensive risk index output by the risk pre-assessment module and according to preset grading rules.

[0028] The collaboration scheme selection module is configured to select the corresponding human-machine collaboration review strategy based on the human-machine collaboration depth level determined by the risk classification module.

[0029] The collaboration scheme execution module is configured to execute the selected human-machine collaboration review strategy, specifically including Scheme 1 (rapid manual confirmation), Scheme 2 (manual review of key nodes) and Scheme 3 (manual review of the entire process).

[0030] The audit process recording module is configured to fully record the audit process and results during the execution of the human-machine collaborative audit strategy, forming an audit log.

[0031] For further details, please refer to Figure 2 , Figure 2 This is a flowchart illustrating a risk-based intelligent review dynamic collaboration method provided in an embodiment of this application. The specific process can be as follows: 101. Obtain flight plan data pending review.

[0032] In this embodiment, the flight plan and related data to be reviewed are first obtained from the low-altitude airspace control system. The low-altitude airspace control system refers to a comprehensive information platform responsible for the approval, monitoring, and management of low-altitude airspace flight plans.

[0033] In some embodiments, a data connection can be established with the low-altitude airspace control system through an API interface to achieve automatic acquisition of flight plan data, avoiding the inefficiency and error-prone problems caused by manual data entry.

[0034] In some embodiments, flight plan data may include multiple categories such as basic flight plan information, airspace data, meteorological data, aircraft data, and historical data.

[0035] The basic information of the flight plan may include the applicant ID, aircraft ID, takeoff point coordinates, landing point coordinates, route coordinate sequence, planned takeoff time, planned landing time, and flight purpose. The applicant ID and aircraft ID are used for subsequent queries of the applicant's historical records and aircraft qualifications, respectively; the takeoff point coordinates, landing point coordinates, and route coordinate sequence are used for airspace compliance checks and route analysis; the planned takeoff time and planned landing time are used for weather condition matching and spacetime conflict detection; and the flight purpose is used for mission complexity assessment and compliance judgment.

[0036] Airspace data can include the airspace type, airspace boundaries, and airspace usage status involved in the flight route. Airspace types are specifically divided into flyable airspace, controlled airspace, and sensitive airspace, used to determine whether a flight plan violates airspace management regulations; sensitive airspace requires special review. Airspace boundaries are used to determine whether a flight route enters restricted airspace. Airspace usage status reflects the current airspace availability, occupancy, or reservation status, used for spatiotemporal conflict detection.

[0037] Meteorological data can include wind speed, wind direction, visibility, weather phenomena, temperature, and weather forecasts. Wind speed is used to determine whether the aircraft's airworthiness wind speed limit is exceeded; wind direction, combined with the flight path, is used to analyze the impact of tailwinds or headwinds on flight; visibility is used to determine whether visual flight conditions are met; weather phenomena are used to identify whether adverse weather conditions such as rain, snow, fog, and thunderstorms affect flight safety; temperature is used to determine whether it affects the aircraft's battery performance or poses a risk of icing; and weather forecasts provide forward-looking weather prediction information for the planned flight period for risk assessment.

[0038] The aircraft data may include the aircraft model, maximum takeoff weight, airworthiness certificate number, and airworthiness certificate validity period. The aircraft model is used to match the aircraft's performance parameters; the maximum takeoff weight is used to determine whether it complies with airworthiness regulations; the airworthiness certificate number is used to verify the authenticity of the qualification; and the airworthiness certificate validity period is used to determine whether the airworthiness qualification is within its validity period.

[0039] The historical data includes the applicant's historical flight records and the aircraft's historical flight records. The applicant's historical flight records are used to assess the applicant's credit and flying experience; the aircraft's historical flight records are used to assess the aircraft's reliability and maintenance status.

[0040] 102. Conduct a multi-dimensional risk assessment of flight plan data to determine the comprehensive risk index.

[0041] In this embodiment, the comprehensive risk index is a quantitative indicator used to characterize the overall risk of flight plan data across multiple dimensions. This comprehensive risk index can be presented numerically; a higher value indicates a greater overall risk for the flight plan, and vice versa.

[0042] In some embodiments, the step "conducting a multi-dimensional risk assessment of flight plan data to determine a comprehensive risk index" may include the following operations: Construct risk pre-assessment prompts corresponding to multiple preset risk dimensions; Input the risk pre-screening prompts and flight plan data into the risk assessment model so that the risk assessment model can output the risk scores of the flight plan data under each risk dimension according to the guidance of the risk pre-screening prompts; The risk scores under each risk dimension are integrated to obtain a comprehensive risk index.

[0043] Among them, the multiple preset risk dimensions can include at least: airspace risk, weather risk, aircraft risk, spatiotemporal conflict risk, applicant risk, and mission complexity.

[0044] Airspace risk dimension: Used to assess the compliance risks of flight routes in terms of airspace use. Specifically, it analyzes whether the route coordinate sequence passes through sensitive airspace (such as military restricted areas or airspace above government offices), controlled airspace (such as airport clearance zones or air traffic control areas), or no-fly zones, and assesses the distance relationship between the route and the boundaries of these restricted airspaces; the closer the distance, the higher the risk. For example, a route directly crossing a no-fly zone is considered extremely high risk, while a route relatively close to the boundary of controlled airspace is considered medium risk.

[0045] Meteorological Risk Dimension: Used to assess whether the meteorological conditions for the planned flight period are suitable for airworthiness. It comprehensively analyzes factors such as wind speed, wind direction, visibility, weather phenomena, and temperature to determine whether the wind speed exceeds the aircraft's maximum wind resistance capability, whether the visibility meets the minimum requirements for visual flight, whether there are severe weather conditions such as rain, snow, fog, or thunderstorms, and whether the temperature is too low to cause icing or affect battery performance. It also considers weather forecasts to assess weather change trends. For example, if the wind speed is close to the limit and the forecast indicates that the wind speed may increase, the risk score will be increased accordingly.

[0046] Aircraft Risk Dimension: Used to assess the risks associated with an aircraft's qualifications and performance. This includes verifying the validity and authenticity of the airworthiness certificate, and determining whether the aircraft's performance meets mission requirements, including whether the maximum takeoff weight complies with regulations, whether the endurance covers the flight duration, whether the payload capacity meets mission requirements, and whether it possesses the capability for safe flight under planned weather conditions. For example, an airworthiness certificate with only 15 days remaining is considered critical, and the risk score will be increased accordingly.

[0047] Spatiotemporal conflict risk dimension: This dimension assesses whether this plan conflicts with other approved flight plans in terms of time and space. It analyzes the relationship between takeoff time, landing time, and flight path coordinate sequence and existing plans to determine if multiple aircraft simultaneously enter the same airspace, if there are intersections in the flight paths, and whether the spatial separation meets safety distance requirements. Factors such as speed and altitude differences are also considered. For example, if a plan passes through the same airspace node at the same time and the time interval is less than the safety threshold, the risk score will be increased accordingly.

[0048] Applicant Risk Dimension: This dimension assesses the creditworthiness and flight experience of the applicant or operating company. It analyzes historical flight records, violation records, flight frequency, and mission completion status to determine if there have been any violations such as unauthorized entry into controlled airspace, excessive altitude flight, or flight beyond permitted range, and whether the applicant's flight experience is sufficient to handle mission complexity. For example, if an applicant has one minor violation and less than one year of flight experience, their risk score will be increased accordingly.

[0049] Mission complexity dimension: This dimension assesses the technical and managerial complexity of a flight mission. It analyzes route length, flight duration, and whether special operations are involved. It determines whether the route exceeds the normal range, whether the flight duration is close to or exceeds the maximum endurance, whether the mission involves special operations such as beyond visual line of sight (BVR) flight, night flight, formation flight, or manned flight, and whether it is conducted in densely populated urban areas or over important facilities. For example, if BVR flight is involved and the flight duration exceeds 2 hours, the risk score will be increased accordingly.

[0050] In some embodiments, the risk assessment model may employ a Large Language Model (LLM), a natural language processing model trained using deep learning techniques and possessing a large number of parameters (typically in the billions to hundreds of billions). This type of model is capable of understanding and generating human language, exhibiting powerful language understanding, logical reasoning, text generation, and knowledge integration capabilities.

[0051] In this embodiment, the prompt word construction step is performed first. Specifically, a special risk pre-assessment prompt word is designed to guide the large language model to conduct a comprehensive risk assessment of the flight plan from multiple preset risk dimensions. This risk pre-assessment prompt word is presented in natural language form, simulating the thinking mode of a low-altitude flight plan review expert, enabling the large language model to output structured risk assessment results according to the preset dimensional framework.

[0052] The risk pre-assessment prompts employ a design approach combining role-playing and task instructions. Specifically, the prompts first assign a clear role to the large language model: a "low-altitude flight plan review expert," guiding the model to analyze the plan from a professional review perspective. Subsequently, the prompts explicitly require the large language model to evaluate the flight plan item by item from the aforementioned multiple preset risk dimensions, and specify the format of the output, including risk scores for each dimension, a comprehensive risk index, and a risk analysis explanation.

[0053] For example, a typical prompt word template is as follows: "You are a low-altitude flight plan review expert. Please conduct a risk assessment on the following flight plan, analyzing it from the following dimensions:" Airspace risks: Whether the flight path passes through sensitive airspace, controlled airspace, or no-fly zones; Meteorological risk: Are the weather conditions suitable for flight (wind speed, visibility, weather phenomena); Aircraft risks: Whether the aircraft has complete qualifications and whether its performance meets mission requirements; Spatiotemporal conflict risk: Whether there are spatiotemporal conflicts with other approved plans; Applicant risks: Applicant's past flight records and violation records; Mission complexity: The degree of complexity of the flight mission (route length, flight duration, special operations); Please output the following: Risk scores for each dimension (0-100 points, the higher the score, the greater the risk); Overall risk index (0-100 points); Risk analysis explanation; Flight plan data: {Flight plan JSON data}.

[0054] In this embodiment, after determining the risk pre-screening prompts and flight plan data, a large language model is invoked to perform a risk assessment. The large language model, acting as the risk assessment model, scores the flight plan data under each preset risk dimension, guided by the risk pre-screening prompts. The system then merges these risk scores from each dimension to generate a comprehensive risk index. This step realizes the technical process of converting multi-dimensional quantitative assessment into a single comprehensive indicator.

[0055] After receiving the constructed risk pre-screening prompts, the large language model analyzes and scores the flight plan data one by one according to the multiple risk dimensions specified in the prompts. The risk score for each dimension is represented by a numerical value from 0 to 100, with a higher score indicating a greater risk for that dimension and a lower score indicating a lower risk.

[0056] Specifically, the large language model processes flight plan data as follows: In the airspace risk dimension, the model analyzes whether the route coordinate sequence passes through sensitive airspace, controlled airspace, or no-fly zones, and assesses the distance relationship between the route and the boundaries of restricted airspace. If the route completely avoids all restricted airspaces and maintains a safe distance, a lower score is output, such as 15 points; if the route directly crosses a no-fly zone, a higher score is output, such as above 90 points.

[0057] In the meteorological risk dimension, the model analyzes meteorological factors such as wind speed, visibility, weather phenomena, and temperature to determine whether they exceed the aircraft's airworthiness conditions. If the meteorological conditions are good and all indicators are far better than the limits, a lower score is output, such as 10 points; if the meteorological conditions are at the edge, with wind speed close to the limit and forecasts indicating a possible deterioration, a moderate score is output, such as 25 points.

[0058] In the aircraft risk dimension, the model checks whether the validity period of the airworthiness certificate and the performance parameters of the aircraft meet the mission requirements. If the validity period of the airworthiness certificate is sufficient and the performance parameters fully meet the mission requirements, a lower score is output, such as 10 points; if the airworthiness certificate is about to expire or the performance is at a critical state, a medium score is output, such as 50 points.

[0059] In the dimension of spatiotemporal conflict risk, the model analyzes whether there is spatiotemporal overlap with the approved plan. If there is no conflict or sufficient interval, a lower score is output, such as 5 points; if there is potential conflict risk, the score is increased accordingly.

[0060] In terms of applicant risk, the model analyzes historical flight records and violation records. If the applicant has a good history and extensive flying experience, a lower score is output, such as 20 points; if there are multiple violation records, a higher score is output.

[0061] In terms of task complexity, the model analyzes the flight path length, flight duration, and whether special operations are involved. If the task is a regular short-haul flight, a lower score is output, such as 5 points; if the task involves beyond visual line of sight (BVR) flight and has a longer flight duration, a higher score is output, such as 30 points.

[0062] Through the above processing, the large language model outputs a dataset containing six dimensions of risk scores, with the following structure example: "risk_dimensions": { "airspace_risk": 15, "weather_risk": 25, "aircraft_risk": 10, "conflict_risk": 5, "applicant_risk": 20, "task_complexity": 30 }; Among them, risk_dimensions represents the risk dimension; "airspace_risk": 15, indicating an airspace risk score of 15; "airspace_risk": 15, indicating a weather risk score of 25; "aircraft_risk": 10, indicating an aircraft risk score of 10; "conflict_risk": 5, indicating a spacetime conflict risk score of 5; "applicant_risk": 20, indicating an applicant risk score of 20; "task_complexity": 30, indicating a task complexity score of 30.

[0063] After obtaining the risk scores for each dimension, a fusion processing step is performed to combine the scores from the six dimensions into a single comprehensive risk index. This index also uses a numerical range of 0 to 100 points to represent the overall risk level of the flight plan data.

[0064] In this embodiment, the fusion processing can employ a weighted average algorithm. Specifically, the system pre-configures corresponding weight coefficients for each risk dimension, with the sum of the weights for each dimension being 1. The formula for calculating the comprehensive risk index is as follows: Comprehensive Risk Index = Airspace Risk Score × Weight 1 + Meteorological Risk Score × Weight 2 + Aircraft Risk Score × Weight 3 + Spatiotemporal Conflict Risk Score × Weight 4 + Applicant Risk Score × Weight 5 + Mission Complexity Score × Weight 6.

[0065] The weights of each dimension can be adjusted according to the actual application scenario. For example, in areas with scarce airspace resources, the weight of the airspace risk dimension can be appropriately increased; in seasons with variable weather conditions, the weight of the weather risk dimension can be appropriately increased. System administrators can dynamically adjust the weights of each dimension through configuration files or the management interface to adapt to the risk assessment needs of different regions, seasons, and task types.

[0066] In a simplified implementation, the system can use the arithmetic mean method, where each dimension has equal weight, and the overall risk index is equal to the average of the six dimension scores.

[0067] In another implementation method, a weighted average method is used, and differentiated weight coefficients can be configured according to the actual needs of risk assessment. For example, if airspace safety and weather conditions have the most critical impact on flight safety, their weights can be set to higher values, such as 0.25 each; other dimensions can be set to lower values, such as 0.125 each.

[0068] After calculating the comprehensive risk index, the system encapsulates the risk scores for each dimension and the comprehensive risk index into structured output data. Simultaneously, the large language model generates a risk analysis description in natural language, clearly identifying the main risk points and corresponding solutions. A complete output example is shown below: { "risk_dimensions": { "airspace_risk": 15, "weather_risk": 25, "aircraft_risk": 10, "conflict_risk": 5, "applicant_risk": 20, "task_complexity": 30 }, "comprehensive_risk_index": 42, "risk_analysis": "The overall risk index for this flight plan is 42 points (medium risk). Key risk points: weather conditions are at a critical juncture, with wind speeds approaching limits; the mission is highly complex, with a long route involving multiple airspaces; the applicant has one minor violation in their historical record. Standard manual review is recommended." }

[0069] Here, "comprehensive_risk_index": 42 indicates a comprehensive risk index of 42. risk_analysis refers to risk analysis.

[0070] After calculating the weighted average, this average can be converted into a final comprehensive risk index of 0-100 points according to preset mapping rules. For example, a linear mapping relationship can be set to ensure that the average and the comprehensive risk index maintain a consistent proportion. Alternatively, a non-linear mapping curve can be set based on historical data statistics to make the distribution of the comprehensive risk index more consistent with the actual risk distribution pattern.

[0071] Through the above-mentioned technical solution of "multi-dimensional scoring + fusion processing", this embodiment has at least the following beneficial effects: comprehensive evaluation results: quantitative scoring is performed from multiple dimensions, which not only retains the fine-grained information of each dimension, but also generates a single comprehensive index through fusion processing, taking into account both the comprehensiveness of the evaluation and the simplicity of decision-making; configurable weights: it supports dynamic adjustment of the weights of each dimension, so that the evaluation model can adapt to the risk preferences and control requirements under different scenarios, and has good flexibility and scalability.

[0072] In some embodiments, the risk index can also be calculated based on predefined rules, as follows: Define the scoring rules for each risk dimension; calculate the scores for each dimension based on the flight plan data matching rules; and obtain the comprehensive risk index by weighted summation.

[0073] In some embodiments, the risk index can also be predicted using a machine learning model, as follows: The model is trained using historical flight plan data and risk labels; input flight plan features and output risk index; model options include: random forest, XGBoost, neural network, etc.

[0074] In some embodiments, the risk index can be calculated using LLM combined with predefined rules, as follows: LLM performs qualitative risk analysis; the rules engine performs quantitative risk calculation; and the results of both are combined to obtain the final risk index.

[0075] 103. Determine the corresponding depth level of human-machine collaboration based on the comprehensive risk index.

[0076] In this embodiment, the comprehensive risk index is further divided into three risk levels: 0-40 points are low risk, 41-70 points are medium risk, and 71-100 points are high risk.

[0077] In some embodiments, the step "determine the corresponding human-machine collaboration depth level based on the comprehensive risk index" may include the following operations: According to the preset risk index classification rules, the comprehensive risk index is mapped to the corresponding target risk level; Based on the target risk level, the depth level of human-machine collaboration is determined from the preset depth levels of human-machine collaboration. The preset depth levels of human-machine collaboration include at least: rapid manual confirmation level, key node manual review level, and full-process manual review level.

[0078] In this embodiment, the corresponding human-machine collaboration depth level is determined based on the risk level of the flight plan. The human-machine collaboration depth level is used to characterize the breadth and depth of human involvement in the review process. The appropriate review strategy is automatically selected based on the human-machine collaboration depth level, achieving a precise match between risk classification and review intensity.

[0079] In some embodiments, three collaboration depth levels are set, from low to high: Level 1 rapid manual confirmation level, Level 2 key node manual review level, and Level 3 full-process manual review level.

[0080] The preset risk index grading rules can be found in Table 1 below: Table 1 If the comprehensive risk index is within the range of 0-40, the corresponding target risk level is: low risk; the corresponding human-machine collaboration depth level is Level 1: rapid manual confirmation. If the comprehensive risk index is between 41 and 70, the corresponding target risk level is: medium risk; the corresponding human-machine collaboration depth level is Level 2: manual review of key nodes. If the comprehensive risk index is between 71 and 100, the corresponding target risk level is: high risk; the corresponding human-machine collaboration depth level is Level 3: full-process manual review.

[0081] Further details on the levels of human collaboration can be found in Table 2 below: Table 2 Level 1: Rapid Human Confirmation Level, corresponding to low-risk flight plans. At this level, human involvement is limited to "rapid confirmation after human review of AI audit results." This level is suitable for low-risk flight scenarios, such as routine aerial photography missions conducted within suitable airspace. These flight plans completely avoid sensitive and controlled airspace, have excellent weather conditions far exceeding restrictions, possess complete and valid aircraft qualifications, have a good applicant history, and involve relatively low mission complexity. In these scenarios, the credibility of AI audit results is high, and human verification is only required, significantly improving audit efficiency.

[0082] Level 2: Key Node Manual Review Level, corresponding to medium-risk flight plans. At this level, the degree of human involvement is "human review of AI-marked key risk nodes." This level is applicable to medium-risk flight scenarios, such as general commercial operations with weather conditions at the edge of their limits, or flight plans with routes approaching the boundaries of controlled airspace.

[0083] Level 3: Full-process manual review level, corresponding to high-risk flight plans. At this level, the degree of human involvement is "human review of every step of the AI ​​review process." This level is applicable to high-risk flight scenarios, such as routes involving sensitive airspace, severe weather conditions, missions involving beyond visual line of sight (BVR) or manned flights, or situations where the applicant has a large number of violations.

[0084] This embodiment achieves the following technical effects by setting three progressive levels of human-machine collaboration depth: First, risk adaptation. The appropriate level of human involvement is automatically matched according to the risk level of the flight plan. In low-risk situations, AI is the main force and human intervention is the auxiliary force, while in high-risk situations, human intervention is the main force and AI intervention is the auxiliary force. This fully leverages the high efficiency advantage of AI while preserving sufficient space for human judgment in high-risk scenarios.

[0085] In some embodiments, risk classification may also be carried out in at least one of the following ways: Dynamic thresholds are adjusted based on historical data: statistical analysis of the risk index distribution of historical flight plans; dynamic adjustment of tiered thresholds based on the distribution; ensuring a relatively balanced number of missions at each level. Configurable thresholds, manually configurable by administrators: Provides a configuration interface, allowing administrators to adjust thresholds for each level; adapts to management requirements in different regions and periods; adaptive thresholds: adjusts based on feedback from audit results; if AI suggestions for a certain level of task are frequently modified manually, the threshold will be automatically adjusted. This allows for adaptive optimization of the threshold.

[0086] In some embodiments, the depth level of human-computer collaboration can also be adjusted according to actual needs, such as 2-level division, 4-level division or more levels, dynamic level, etc.

[0087] For example, the two-level classification is simplified into two levels: low and high. Level 1: low risk, quick manual confirmation; Level 2: high risk, full-process manual review.

[0088] For example, a four-level or more classification system provides a finer-grained distinction between levels. This involves adding intermediate levels to the existing three levels, such as a "key node review" level. This is suitable for large systems with a wider variety of task types.

[0089] For example, the system can be dynamically adjusted based on system load: when the system load is high, the fast confirmation threshold is increased (more tasks move to Level 1); when the system load is low, the fast confirmation threshold is decreased (more tasks move to Level 2-3).

[0090] 104. Based on the depth level of human-machine collaboration, determine the matching target review strategy from multiple preset human-machine collaboration review strategies.

[0091] Different human-machine collaborative review strategies correspond to different levels of human review intensity.

[0092] In this embodiment, the intensity of manual review refers to the quantitative representation of the workload and depth of review by manual reviewers during the approval process of flight plans. The higher the intensity of manual review, the more time reviewers need to invest, the more review items they need to focus on, and the more in-depth their judgment and analysis; conversely, the lower the intensity of manual review, the lighter the workload of reviewers and the simpler the review process.

[0093] In some embodiments, the multiple preset human-machine collaborative review strategies include at least: a rapid human confirmation strategy, a key node human review strategy, and a full-process human review strategy.

[0094] Among them, the human-machine collaboration depth level corresponding to the rapid manual confirmation strategy is rapid manual confirmation; the human-machine collaboration depth level corresponding to the key node manual review strategy is key node manual review; and the human-machine collaboration depth level corresponding to the full-process manual review strategy is full-process manual review.

[0095] For example, if the depth level of human-machine collaboration is Level 1 with rapid human confirmation, then the rapid human confirmation strategy can be selected as the target review strategy; if the depth level of human-machine collaboration is Level 3 with key node human review, then the key node human review strategy can be selected as the target review strategy; if the depth level of human-machine collaboration is Level 3 with full-process human review, then the full-process human review strategy can be selected as the target review strategy.

[0096] 105. Implement the target review strategy to complete the review of flight plan data.

[0097] In some embodiments, if the target review strategy is a rapid manual confirmation strategy, the step "Execute the target review strategy" may include the following operations: Artificial intelligence is used to perform full-process automated review of flight plan data and generate review results. The full-process automated review includes at least one of the following: airspace compliance check, weather condition airworthiness check, aircraft qualification validity check, spatiotemporal conflict detection and applicant qualification verification. The review results include at least one of the following: artificial intelligence approval suggestions, risk index and risk level, results of each inspection item, and key information summary. The review results will be sent to a human reviewer for viewing and confirmation. In response to the reviewer's confirmation of the review results, the approval of the flight plan data is completed.

[0098] For example, please see Figure 3 , Figure 3This is a flowchart illustrating another intelligent audit dynamic collaboration method based on risk classification provided in an embodiment of this application. Figure 3 The execution flow corresponding to the rapid manual confirmation strategy is shown below: AI completes the entire automated review process: First, the artificial intelligence engine is invoked to perform a fully automated review of the flight plan data. This fully automated review includes at least the following five core checks: Airspace compliance check: Obtain the flight plan's route coordinate sequence and combine it with airspace boundary data obtained from low-altitude air traffic control to determine whether the route passes through sensitive airspace, controlled airspace, or no-fly zones, and calculate the shortest distance between the route and the boundaries of various restricted airspaces. If the route is entirely within the flyable airspace and maintains a safe distance from the controlled airspace boundary, the airspace compliance check result is "passed"; if the route enters controlled airspace or is too close to the boundary, the check result is "failed" or "requires attention".

[0099] Meteorological airworthiness check: Obtain meteorological data for the planned flight period, including wind speed, wind direction, visibility, weather phenomena, and temperature. Combine this data with the aircraft's performance parameters (such as maximum wind resistance and minimum visibility requirements) to determine whether the meteorological conditions meet airworthiness requirements. If all meteorological indicators are better than or equal to the airworthiness limits, the meteorological airworthiness check result is "passed"; if any indicator exceeds the limit or is at a boundary, the check result is "failed" or "requires attention".

[0100] Aircraft Qualification Validity Check: Obtain the aircraft's airworthiness certificate number and validity period, and verify them against the official registration database to determine whether the airworthiness certificate is genuine, valid, and within its validity period. Simultaneously, check whether the aircraft's performance parameters (such as maximum takeoff weight, range, and payload capacity) meet the technical requirements of this flight mission. If the airworthiness certificate is valid and the performance parameters meet the mission requirements, the aircraft qualification check result is "Pass"; if the airworthiness certificate has expired, is about to expire, or the performance parameters do not meet the mission requirements, the check result is "Fail" or "Requires Attention".

[0101] Spatiotemporal conflict detection: This involves acquiring the takeoff time, landing time, and route coordinate sequence of this flight plan, comparing and analyzing this data with corresponding data from other approved flight plans, determining whether multiple aircraft enter the same airspace within the same time period, whether there are intersections or junctions between routes, and whether the spatial spacing between routes meets safety distance requirements. If no spatiotemporal conflict exists, the detection result is "Pass"; if potential conflict risks exist, the result is "Fail" or "Requires Attention".

[0102] Applicant Qualification Verification: Obtain the applicant's historical flight records and violation records to determine whether the applicant possesses the qualifications and credibility to perform this flight mission. Specifically, this includes whether the applicant has engaged in any violations such as unauthorized entry into controlled airspace, flying beyond permitted altitudes, or flying outside designated areas, and whether the applicant's flight experience is sufficient to handle the complexity of this mission. If the applicant has a good historical record and possesses the corresponding qualifications, the applicant's qualification verification result is "Pass"; if there are violation records or insufficient qualifications, the verification result is "Fail" or "Requires Attention".

[0103] After the above five checks are completed, the results of each check are summarized to generate structured audit result data for use in subsequent steps.

[0104] Displaying a summary of the AI ​​review results: After the entire automated review process is completed, the generated review results will be pushed to the human review panel in a concise card format for display. The review results should include at least the following: AI-powered approval recommendations: Based on the results of the fully automated review process, the system makes a comprehensive judgment and provides a "reject" or "approval" recommendation. For example, if all inspection items are "passed," an "approval" recommendation is given; if any inspection item is "failed," a "rejection" recommendation is given or a note indicating that it requires special human attention is provided, based on preset rules.

[0105] Risk Index and Risk Level: Displays the comprehensive risk index (0-100 points) and its corresponding risk level. The comprehensive risk index quantifies the overall risk of the flight plan, while the risk level is categorized as low risk (0-40 points), medium risk (41-70 points), or high risk (71-100 points) based on the index range. In the rapid manual confirmation strategy, since it only applies to low-risk flight plans, the displayed risk index should be within the range of 0-40 points, and the risk level should be low.

[0106] Results of each inspection item: The summary results of the above five inspection items are displayed in a visual manner, usually using a concise expression such as "All Passed" or "Problems Exist". If all inspection items are "Passed", "All Passed" is displayed; if any inspection item is "Failed" or "Requires Attention", "Problems Exist" is displayed, and you can further expand to view the specific problem items.

[0107] Key Information Summary. Extract and summarize the core information of the flight plan, including at least the airspace type (e.g., suitable airspace, distance to controlled airspace boundaries), meteorological conditions (e.g., wind speed, visibility overview), aircraft type, and airworthiness status. The key information summary aims to help reviewers grasp the core elements of the flight plan in the shortest possible time, without needing to consult the complete raw data.

[0108] With the card-style summary display described above, reviewers can obtain the key information needed for review within seconds, greatly improving the efficiency of reviewing low-risk flight plans.

[0109] Reviewers quickly browse and confirm: After the review results are displayed, the above summary information is pushed to the human review platform, awaiting reviewer's viewing and confirmation. Specifically, the human review platform interface provides the following interactive options: Confirmation: If the reviewer agrees with the AI's approval suggestion after reviewing the summary information (e.g., the AI ​​suggests "Approve" and the reviewer deems it reasonable), they should click the "Confirm" button on the interface. In response to this confirmation, the approval process will proceed directly to completion, requiring no further action from the reviewer.

[0110] For more details: If reviewers have any questions about the summary information or wish to view more detailed review evidence, they can click the "Details" button. This will display the complete review report, including detailed data for each inspection item, the basis for the AI's analysis and judgment, and references to the original data. Reviewers can return to the summary interface to continue their decision-making after viewing the details.

[0111] Editing Procedure: If reviewers find the AI's approval suggestions unreasonable or discover AI misjudgments after reviewing the summary information or details, they can click the "Edit" button. This action will initiate a manual editing process, allowing reviewers to manually adjust the inspection results and provide a reason for the change. After the edits are complete, the manually edited results will be used for subsequent processing.

[0112] The above interactive design ensures that the rapid human confirmation strategy, while pursuing high efficiency, retains the necessary exception handling channels, enabling reviewers to intervene and correct in a timely manner when encountering situations where the AI ​​judgment is questionable.

[0113] When the reviewer clicks the "Confirm" button, the approval process is completed in response to this confirmation. This includes: Record approval results: Write data such as AI approval suggestions, reviewer confirmation actions, confirmation time, and reviewer identity information into the approval record.

[0114] Update flight plan status: Update the approval status of the flight plan to "approved" or "rejected" and feed back the approval result to the low-altitude airspace control.

[0115] Generate approval vouchers: Generate official approval vouchers or approval documents based on the approval results, which applicants can download or view.

[0116] Triggering subsequent processes: Based on the approval results, subsequent processes can be automatically triggered, such as including the approved flight plan in the spatiotemporal conflict detection database as a benchmark for subsequent flight plan conflict detection; or returning the rejected flight plan to the applicant with the reasons for rejection for them to modify and resubmit.

[0117] Approval complete: This concludes the complete approval process for the rapid manual confirmation strategy. Throughout the process, reviewers only need to spend tens of seconds browsing the summary information and clicking confirm to complete the approval of a low-risk flight plan, achieving extremely high review efficiency while ensuring review quality.

[0118] Through the aforementioned technical solution of "AI-powered full-process automated review—summary display—rapid manual confirmation—approval completion," this embodiment has at least the following beneficial effects: Significantly improved review efficiency: The rapid confirmation method for low-risk flight plans reduces the time required for traditional manual review from several minutes or even tens of minutes to tens of seconds, significantly reducing the workload of reviewers; Maintaining the final decision-making authority of human judgment: Although AI approval suggestions are provided, the final decision is still made by the reviewer through confirmation, preserving the authority and responsibility boundaries of human review; Providing an anomaly handling channel: Reviewers can view details or enter the modification process at any time when encountering questions, avoiding approval errors caused by AI misjudgment, balancing efficiency and accuracy; Standardized review process: Through fixed steps and interface design, different reviewers can maintain consistency in their handling of low-risk flight plans, reducing inconsistencies caused by differences in personal habits.

[0119] The following detailed explanation of the implementation process of the rapid manual confirmation strategy of this application is based on a specific low-risk flight plan example.

[0120] Suppose a drone aerial photography company submits a flight plan to low-altitude air traffic control, planning to use a DJI Mavic 3 drone to perform landscape aerial photography in an urban park. The specific data for the flight plan are as follows: the maximum takeoff weight of the aircraft is 0.9 kg; the planned takeoff time is 10:00 AM on January 20, 2026; the planned landing time is 10:30 AM on the same day; and the flight objective is landscape aerial photography of the park. Meteorological data shows clear weather, a wind speed of 3 meters per second, and visibility of 15 kilometers. The applicant has a good history with no violations.

[0121] LLM risk assessment process: After obtaining the aforementioned flight plan data, it was input into the large language model along with the risk pre-assessment prompts. The large language model comprehensively evaluated the flight plan from six dimensions: airspace risk, weather risk, aircraft risk, spatiotemporal conflict risk, applicant risk, and mission complexity. The evaluation results showed that the flight plan's route was within the flyable airspace and did not involve any sensitive airspace, controlled airspace, or no-fly zones; the weather conditions were excellent, with wind speed, visibility, and other indicators far exceeding airworthiness limits; the aircraft's qualifications were complete and valid; there were no spatiotemporal conflicts with other approved flight plans; the applicant had a good track record; and the mission was a routine short-haul aerial photography mission with low complexity.

[0122] Based on the above analysis, the large language model outputs a comprehensive risk index of 18 points, which is judged as low risk. The risk analysis description states that "the aircraft is conducting routine aerial photography within the suitable airspace, the weather conditions are excellent, the aircraft has complete qualifications, the applicant has a good history of service, and all inspection items have been passed."

[0123] Human-machine collaborative review strategy selection: After receiving a comprehensive risk index of 18, the risk grading module determines that the flight plan belongs to the low-risk level (0-40 points) according to the preset grading rules, and automatically selects the Level 1 quick manual confirmation strategy.

[0124] Solution implementation process: The collaboration plan execution module initiates a rapid manual confirmation process. The specific execution steps are as follows: AI completed the entire automated review process. It performed airspace compliance checks, weather condition airworthiness checks, aircraft qualification validity checks, spatiotemporal conflict detection, and applicant qualification verification on the flight plan; all checks resulted in a "pass". Display AI review result summary: The review result is displayed in a concise card format, including the AI ​​approval suggestion as "approved", the risk index as 18 points, the risk level as "low risk", the result of each inspection item as "all passed", and the key information summary as airspace type "suitable airspace", weather conditions "clear, wind speed 3m / s, visibility 15km", and aircraft model "DJI Mavic 3". The reviewers quickly reviewed and confirmed the information. After receiving the above summary information on the human review platform, the reviewers quickly reviewed all the content to confirm that the AI's judgment was reasonable and there were no abnormalities. Complete the approval process: The reviewer clicks the "Confirm" button on the interface, responds to the confirmation operation, records the approval result, updates the flight plan status to "Approved", and feeds back the approval result to the low-altitude airspace control.

[0125] In this embodiment, a complete flight plan approval process takes only tens of seconds from submission to completion, with the reviewer's operation time being less than 10 seconds. Compared to the traditional manual item-by-item review method (which typically takes several minutes to tens of minutes), this embodiment significantly improves the approval efficiency of low-risk flight plans. Simultaneously, since AI has completed the automatic review of all checkpoints, and the reviewer has quickly confirmed the information through the summary, the approval quality is effectively guaranteed. This embodiment fully embodies the collaborative concept of "AI as the primary tool and human assistance as a supplement" in low-risk scenarios, achieving a balance between efficiency and safety.

[0126] In some embodiments, if the target audit strategy is a manual review strategy for key nodes, the step "Execute the target audit strategy" may include the following operations: Artificial intelligence is used to perform full-process automated review of flight plan data. The full-process automated review includes at least one of the following: airspace compliance check, weather condition airworthiness check, aircraft qualification validity check, spacetime conflict detection, and applicant qualification verification. During the review process, key risk points are identified based on preset risk point identification criteria. The results of the fully automated review process and key risk points are pushed to the manual review end so that the reviewers can review the key risk points item by item. Receive the review results from auditors for each key risk point, and complete the approval of flight plan data based on the review results.

[0127] For example, please see Figure 4 , Figure 4 This is a flowchart illustrating another intelligent audit dynamic collaboration method based on risk classification provided in an embodiment of this application. Figure 4 The execution flow corresponding to the manual review strategy for key nodes is shown below: AI completes the entire automated review process: First, an artificial intelligence engine is invoked to perform a fully automated review of the flight plan data. This automated review includes at least five core checks: airspace compliance check, weather condition airworthiness check, aircraft qualification validity check, spatiotemporal conflict detection, and applicant qualification verification. The specific execution methods for each check are consistent with the aforementioned rapid manual confirmation strategy and will not be repeated here.

[0128] AI identifies key risk points: During or after the fully automated review process, key risk points requiring manual attention are automatically identified based on preset risk point identification criteria.

[0129] The preset risk node identification conditions can include the following three categories: Inspection items with a single risk score of 50 or above: Based on the risk scores of each dimension output by the large language model, the inspection items corresponding to the dimensions with a score ≥ 50 are marked as key risk nodes; For inspection items whose results are at the boundary conditions: Determine whether the quantitative indicators of each inspection item are close to the preset limit value. For example, if the wind speed reaches more than 80% of the aircraft's maximum wind resistance capacity, the route is less than 500 meters away from the boundary of the controlled airspace, or the airworthiness certificate has less than 30 days remaining, the inspection items that are at the above boundary conditions will be marked as key risk nodes. Check items with potential conflicts or anomalies: Identify check items with potential spatiotemporal conflicts (such as the spatiotemporal interval of the approved plan being close to the safety minimum) or other anomalies, and mark them as key risk nodes; By using the above identification rules, key nodes that require manual intervention can be accurately selected from numerous inspection items, avoiding the workload of auditors facing all inspection items.

[0130] In some embodiments, the key node types may include: airspace risk nodes, meteorological risk nodes, conflict risk nodes, and qualification risk nodes.

[0131] The triggering conditions for airspace risk nodes can be: the flight path is close to the boundary of the controlled airspace, for example, the flight path is less than 50 meters from the boundary of the controlled airspace; the triggering conditions for meteorological risk nodes can be: meteorological conditions are close to the limit value, for example, the wind speed reaches more than 80% of the limit value; the triggering conditions for conflict risk nodes can include: there is a potential spatiotemporal conflict, for example, the spatiotemporal interval with other plans is close to the minimum value; the triggering conditions for qualification risk nodes can include: the qualification is about to expire or there is a violation record, for example, the airworthiness certificate is less than 30 days away.

[0132] Display the review results and highlight key points: After completing the fully automated review process and identifying key risk points, the review results and key point information are pushed to the human review end. The human review interface displays a complete review report, including detailed results for each inspection item and AI analysis explanations. Simultaneously, the marked key risk points are visually differentiated: key risk points are highlighted in orange or red, displaying detailed data and AI analysis explanations; non-key points are displayed in gray, indicating they have passed AI review and have low risk, requiring no special human attention.

[0133] Through the above differentiated display, auditors can easily identify the audit items that need to be focused on, thus concentrating their limited attention on the review of key risk points.

[0134] The auditors reviewed each key point: After receiving the review results, the manual review team conducts a point-by-point review of the marked key risk nodes. Specifically, all key risk nodes are displayed in the form of a review checklist, with each node including its type, specific data, AI judgment result, and review options. Reviewers examine the detailed data for each key node, independently assess the reasonableness of the AI's judgment, and select "Confirm" (agree with the AI's judgment) or "Modify" (disagree with the AI's judgment) for each node. After completing the review of each node, the reviewers check off each item to ensure that all key nodes have undergone manual review.

[0135] Taking a medium-risk flight plan as an example, the review checklist might include the following: the airspace risk node shows "The flight path is 450 meters from the boundary of the controlled airspace, AI judges it as passed (boundary condition)"; the meteorological risk node shows "Wind speed is 8 meters per second, the limit is 10 meters per second, AI judges it as passed (approaching the limit)"; and the applicant risk node shows "There is one minor violation in the past, AI judges it as passed (needs attention)". Reviewers need to review each of the above nodes one by one and make a confirmation or modification judgment.

[0136] The reviewers made the final decision: After reviewing all key risk points item by item, the auditors fill in their review comments, explaining their review conclusions and the basis for their judgments on each key point. Upon receiving the review results and comments submitted by the auditors, the system generates a final approval decision based on the review status of each point. If all key points are confirmed and no modifications are requested, approval is completed according to the AI's approval suggestions; if any points have been modified, the approval is completed based on the human-modified judgment. Finally, the approval results, review comments, and auditor information are recorded, and the approval status is fed back to the low-altitude airspace control, completing the entire approval process.

[0137] Through the aforementioned technical solution of "AI-powered full-process review—key node identification—differentiated display—item-by-item review—approval completion," this embodiment has at least the following beneficial effects: it reduces the scope of reviewers' work from all inspection items to a limited number of key risk nodes, significantly reducing the workload of manual review; at the same time, through preset quantitative identification rules (such as scoring thresholds and boundary condition thresholds), it ensures that the identification of key nodes is objective and consistent, avoiding omissions or misjudgments caused by differences in human judgment; in addition, the differentiated display and review checklist design make the review process clear and intuitive, improving review efficiency and user experience.

[0138] In some embodiments, identifying key risk points may also employ at least one of the following methods: Rule configuration method: Administrators pre-configure key node identification rules; configure key node trigger conditions for each inspection item; the system automatically marks key nodes according to the configured rules; Historical data learning: Learn key nodes based on historical audit data; analyze the inspection items that were manually focused on during historical audits; learn key node identification patterns; Manual designation method: Reviewers can manually designate key nodes; the system provides AI-suggested key nodes; reviewers can add or delete key nodes.

[0139] The following example, using a specific medium-risk flight plan, details the implementation process of the manual review strategy for key nodes in this application.

[0140] Suppose a logistics company submits a flight plan to low-altitude air traffic control, intending to use an industrial-grade drone to perform logistics delivery tasks in an area on the outskirts of a city. The specific data of the flight plan are as follows: the aircraft is an industrial-grade drone, with a maximum takeoff weight of 15 kg; the planned takeoff time is 14:00 on January 20, 2026; the planned landing time is 15:00 on the same day; and the flight purpose is logistics delivery. Meteorological data shows that the weather on that day will be cloudy, with a wind speed of 8 meters per second and a visibility of 8 kilometers. The applicant has one minor violation record.

[0141] The main risks of this flight plan are: the route is located on the outskirts of the city, and some sections of the route are close to the boundary of the controlled airspace; the wind speed in the meteorological conditions reaches 80% of the aircraft's limit value, which is at the boundary; and the applicant has a minor violation record, which needs to be monitored.

[0142] LLM risk assessment process: After obtaining the aforementioned flight plan data, it was input into the large language model along with the risk pre-assessment prompts. The large language model conducted a comprehensive assessment from six dimensions, outputting risk scores for each dimension and a comprehensive risk index. The assessment results showed: the airspace risk score was 55 points, mainly due to the flight path being only 400 meters from the boundary of the controlled airspace; the meteorological risk score was 50 points, mainly due to wind speeds reaching 80% of the aircraft's maximum wind resistance capacity, placing it under boundary conditions; the aircraft risk score was 20 points, indicating that the industrial-grade drone had complete qualifications and good performance; the applicant's risk score was 35 points, stemming from a minor historical violation record; other dimensions showed low risk. Based on the scores of the above dimensions, the large language model calculated a comprehensive risk index of 52 points, classifying it as a medium-risk level (41-70 points range).

[0143] Human-computer collaboration solution selection: After receiving a comprehensive risk index of 52, the risk grading module determines that the flight plan belongs to the medium risk level according to the preset grading rules and automatically selects the Level 2 key node manual review plan.

[0144] Solution implementation process: The collaboration scheme execution module initiates a manual review strategy for key nodes. The specific execution steps are as follows: AI has completed the entire automated review process: the flight plan has undergone airspace compliance checks, meteorological condition airworthiness checks, aircraft qualification validity checks, spatiotemporal conflict detection, and applicant qualification verification. All checks have been completed. AI identifies key risk points: Based on preset key point identification rules, it automatically identifies key points requiring manual review from each inspection item. These include: airspace risk points, triggered when the flight path approaches the controlled airspace boundary (actual distance 400 meters, less than the 500-meter threshold); meteorological risk points, triggered when meteorological conditions approach the limit (wind speed 8 meters per second, reaching 80% of the limit); and applicant risk points, triggered by a prior violation record (one minor violation in the past). Display the AI ​​review results, highlighting key points. Push the complete review report to the human reviewer, highlighting the three key risk points in orange or red, and displaying detailed data and AI analysis explanations for each key point. Non-critical points are displayed in gray, indicating that they have passed AI review and have low risk. The auditors reviewed each key point item by item. The manual review panel displayed three key points in the form of a review checklist, and the auditors reviewed each point in turn: For airspace risk points, the reviewers checked the route coordinates and controlled airspace boundary data, confirmed that the route did not actually enter the controlled airspace, and the 400-meter interval distance met the relevant requirements. Therefore, the AI's "pass" judgment was confirmed. For meteorological risk points, the reviewers checked the wind speed data and weather forecasts and confirmed that the current wind speed of 8 meters per second was within the acceptable range for industrial-grade drones, and the forecast showed that the wind speed did not show any trend of deterioration. Therefore, the AI's "pass" judgment was confirmed. For the applicant's risk points, the reviewers checked the details of the applicant's violation records, confirmed that the violation was minor and had been rectified, and would not affect the approval of this flight application. Therefore, they confirmed the AI's "pass" judgment. The auditor filled in the review comments: After completing the review of each of the three key nodes, the auditor filled in the review comments as follows: "The key risk nodes have been reviewed, the risks are controllable, and approval is recommended." The auditors make the final decision: Based on the above review results, the auditors make the final approval decision of "approval".

[0145] In this embodiment, the approval process for a medium-risk flight plan is efficiently completed through a key node manual review scheme. Specifically, the AI-powered automated review covers all inspection items, but only three key risk nodes that meet preset rules are pushed to the manual review end for further review. Reviewers do not need to review all inspection items one by one, but only need to focus on these three key nodes. This significantly improves review efficiency compared to a full-process manual review scheme (which typically takes 15 to 20 minutes).

[0146] In some embodiments, if the target review strategy is a full-process manual review strategy, the step "Execute the target review strategy" may include the following operations: Artificial intelligence is used to perform full-process automated review of flight plan data. The full-process automated review includes at least one of the following: airspace compliance check, weather condition airworthiness check, aircraft qualification validity check, spacetime conflict detection, and applicant qualification verification. The results of the fully automated review process, along with flight plan data, are pushed to the manual review end so that reviewers can review each item in the review results. The system receives the review results of each item of the inspection by the auditors, compares the review results of each item with the audit results of the artificial intelligence, and obtains the comparison results. The comparison results are pushed to the manual review end so that the reviewers can confirm the final review decision based on the comparison results and complete the review of the flight plan data.

[0147] For example, please see Figure 5 , Figure 5 This is a flowchart illustrating another intelligent audit dynamic collaboration method based on risk classification provided in an embodiment of this application. Figure 5 The execution flow corresponding to the full-process manual review strategy is shown below: AI completes the entire automated review process: First, an artificial intelligence engine is invoked to perform a fully automated review of the flight plan data. This automated review includes core checks such as airspace compliance, weather condition airworthiness, aircraft qualification validity, spatiotemporal conflict detection, and applicant qualification verification. The specific execution methods for each check are consistent with the corresponding steps of the aforementioned rapid manual strategy, and will not be repeated here.

[0148] Show the complete audit report and raw data: After the automated review process is completed, the review results and original flight plan data are pushed to the manual review end. The pushed data includes a complete review report and all original data, specifically: the AI ​​review report (results and analysis of each inspection item), original flight plan data (route, time, mission objective, etc.), detailed airspace data (airspace type, boundaries, usage status), detailed meteorological data (real-time weather and weather forecasts), detailed aircraft data (model, performance parameters, airworthiness information), and detailed applicant data (qualifications, historical records). This data is presented to the manual review end in the form of a full-process review checklist, enabling reviewers to fully understand all aspects of the flight plan and the basis for the AI's judgments.

[0149] The auditors reviewed all inspection items one by one: After receiving the review results, the reviewers conduct a review of all the check items in the results. Specifically, all check items are displayed in the form of a full-process review checklist, and each check item includes at least the check item name, the AI ​​judgment result (pass / fail), the manual review option (confirm / modify), and a field for filling in the review comments.

[0150] The reviewers filled in their review comments: Auditors need to review the detailed data and AI analysis descriptions for each inspection item in turn, conduct independent manual review of each inspection item, select "Confirm" to agree with the AI ​​judgment result, or select "Modify" to correct the AI ​​judgment result, and fill in the review comments for each inspection item, explaining the basis and reasons for the review.

[0151] Taking a high-risk flight plan as an example, the full-process review checklist may include seven items: airspace compliance, weather airworthiness, aircraft qualifications, spatiotemporal conflict detection, applicant qualifications, route rationality, and mission compliance. Reviewers need to review each item one by one and make a judgment, as shown in Table 3 below: Table 3 The system compares human judgment with AI suggestions: After receiving the review results of each item submitted by the auditors, the review results of each item are compared with the original review results of the AI, generating a comparison result. Specifically, for each item, it is determined whether the human review judgment (confirmation or modification) is consistent with the AI's original judgment: if consistent, the item is recorded as "human-machine judgment consistent"; if inconsistent, the item is recorded as "human-machine judgment inconsistent", and the reasons for modification and review comments filled in by the auditors are retained.

[0152] The reviewers confirmed the final decision: The generated comparison results are pushed to the human review stage, allowing reviewers to clearly understand the consistency and differences between human judgment and AI suggestions on each inspection item. If all inspection items are consistent between human and AI judgments, reviewers can directly confirm the final review decision; if there are inconsistencies between human and AI judgments, reviewers are prompted to focus on these discrepancies and are required to confirm the reasonableness of the reasons for the discrepancies. Based on the comparison results and the entire review process, reviewers make the final review decision (approval or rejection), fill in the final approval comments, and complete the flight plan data review process. Subsequently, the complete review process, comparison results, and final decision are recorded to form an audit log for future reference.

[0153] Through the aforementioned technical solution of "AI full-process review—complete data push—item-by-item review—result comparison—difference confirmation—final decision," this embodiment has at least the following beneficial effects: Reviewers conduct item-by-item reviews of all inspection items, ensuring that each review stage of high-risk flight plans is independently judged by humans, maximizing flight safety; by comparing the results of manual review with the AI ​​review results item by item, reviewers can clearly identify inspection items that differ from the AI ​​judgment, facilitating focused review and difference confirmation; the complete data push and review checklist design standardize and normalize the review process, reducing fluctuations in review quality caused by individual differences among reviewers; and the complete review process record provides a reliable basis for subsequent audit traceability and responsibility determination.

[0154] The following example, using a specific medium-risk flight plan, details the implementation process of the full-process manual review strategy for this application.

[0155] Suppose an emergency rescue organization submits a flight plan to low-altitude air traffic control, intending to use a large drone to perform a nighttime emergency rescue mission. The specific data for this flight plan are as follows: the aircraft is a large drone with a maximum takeoff weight of 50 kg; the planned takeoff time is 20:00 on January 20, 2026; the planned landing time is 22:00 on the same day; the flight purpose is emergency rescue. The flight route needs to cross the edge of sensitive airspace and involves multiple controlled airspaces. Meteorological data shows that the weather on that day will be cloudy, with a wind speed of 9 meters per second, visibility of 5 kilometers, and a possibility of rain. The mission is special in nature, time-sensitive, and requires nighttime flight.

[0156] LLM risk assessment results: After obtaining the aforementioned flight plan data, it was input into the large language model along with risk pre-assessment prompts. The large language model conducted a comprehensive assessment from six dimensions, outputting risk scores for each dimension and a comprehensive risk index. The assessment results showed: the airspace risk score was 85 points, mainly due to the flight route involving the edge of sensitive airspace and multiple controlled airspaces; the meteorological risk score was 78 points, mainly due to nighttime flight, weather conditions being at the edge of their range, and the possibility of rainfall; the aircraft risk score was 45 points, as large drones are heavier and have higher requirements for weather conditions; and the mission complexity score was 90 points, indicating that the emergency rescue mission was time-sensitive and required nighttime flight. Based on the scores of the above dimensions, the large language model calculated a comprehensive risk index of 82 points, classifying it as a high-risk level (71-100 points range).

[0157] Human-computer collaboration solution selection: After receiving a comprehensive risk index of 82, the risk grading module determines that the flight plan belongs to the high-risk level according to the preset grading rules and automatically selects the Level 3 full-process manual review strategy.

[0158] Solution implementation process: The collaboration solution execution module initiates a full-process manual review strategy. The specific execution steps are as follows: AI completes the entire automated review process: various checks are performed on the flight plan, and the AI ​​makes a comprehensive judgment and gives an "approval" suggestion; Display the complete review report and raw data: Push the AI ​​review report and all raw data (including detailed airspace data, detailed meteorological data, detailed aircraft data, detailed applicant data, etc.) to the manual review end and display them in the form of a full-process review checklist; The auditors reviewed all inspection items item by item: They reviewed each of the seven items in turn: airspace compliance, weather airworthiness, aircraft qualifications, spatiotemporal conflict detection, applicant qualifications, route rationality, and mission compliance. Regarding airspace compliance, the auditors confirmed that a temporary use permit for sensitive airspace had been obtained and agreed with the AI's "pass" judgment. Regarding weather airworthiness, the auditors, after analysis, determined that the probability of rainfall was 60%, making nighttime flight too risky, and changed the AI's "pass" to "fail." The remaining inspection items were all confirmed to have correct AI judgments after review.

[0159] The reviewers filled in their review comments and explained the basis for the review.

[0160] The review of all inspection items by the auditors can be found in Table 4 below: Table 4 The reviewers made an independent judgment based on the results of the full review process. The reviewers determined that the weather conditions did not meet the safety requirements for nighttime flight and made an independent judgment of "rejection".

[0161] Compare human judgment with AI suggestions. If a discrepancy is detected between the human judgment (rejection) and the AI ​​suggestion (approval), the reviewer is prompted to confirm and explain the reasons for the difference.

[0162] The reviewer confirmed the final decision. The reviewer explained the reason for the discrepancy: "Although the AI's overall assessment was satisfactory, the weather conditions are at a critical juncture and show signs of deterioration. Nighttime flight is too risky. We recommend waiting for weather conditions to improve before reapplying." The reviewer ultimately confirmed the "reject" decision and completed the approval process.

[0163] This embodiment fully embodies the collaborative concept of "human-based and AI-assisted" in high-risk scenarios, ensuring thorough review of high-risk tasks and effectively safeguarding flight safety.

[0164] In some embodiments, the method may further include the following steps: During the execution of the target audit strategy, record the complete audit process and audit results; An audit log is generated based on the audit process and audit results. The audit log includes at least one of the following: unique identifier of flight plan, comprehensive risk index, risk level, human-machine collaboration depth level, human-machine collaboration audit strategy implemented, artificial intelligence approval suggestions and basis, final decision of the auditor, approval comments filled in by the auditor, list of personnel involved in the approval, and operation timestamps for each audit stage.

[0165] In this embodiment, regardless of which human-machine collaborative audit strategy is adopted, the audit process and results will be fully recorded, forming an audit log. This audit log is stored in a trusted storage medium to ensure the authenticity, integrity, and tamper-proof nature of the data.

[0166] Specifically, the contents of the audit log can be found in Table 5 below: Table 5 In this embodiment, the audit log mainly serves the following purposes: Post-event traceability: Audit logs can trace the entire process of each approval decision. When a flight safety incident or violation occurs after approval, managers can use audit logs to trace the entire process of flight plan submission to approval, including raw data, risk assessment results, AI review results, manual review operations, final decision, and timestamps and operator information for each stage. This allows for accurate reconstruction of the decision-making process and provides a basis for accident investigation and liability determination.

[0167] Quality monitoring: Audit logs can be used to analyze approval quality and identify problems. Management can periodically conduct statistical analysis of audit logs to assess the consistency of auditors' decisions, the accuracy of AI audit results, the efficiency of the audit process, and identify abnormal approval behaviors, thereby establishing a quality evaluation system and continuously optimizing the audit process.

[0168] Clear accountability: Audit logs clearly identify the responsible parties at each stage. The logs fully record the user's identity and timestamp for each operational step, forming an undeniable chain of operational evidence. This supports the tracing of responsibility for approval decisions, verification of operational compliance, and prevention of shirking responsibility, thus establishing an effective accountability mechanism.

[0169] Data Accumulation: Audit logs can provide training data for AI model optimization. The manual review results recorded in the logs, especially the "modification" operations and reasons for the reviewers' comments on the AI ​​judgments, constitute high-quality manually labeled data, which can be used for model fine-tuning training, recognition rule optimization, risk dimension weight adjustment, and prompt word optimization, enabling the system to continuously improve itself.

[0170] The audit log mechanism enables the approval process to be traceable, the review quality to be supervised, the responsibilities of each link to be clearly defined, and the AI ​​model to be continuously optimized, thereby improving the system's reliability, standardization, and intelligence.

[0171] This application provides a risk-based intelligent review and dynamic collaboration method, comprising: acquiring flight plan data to be reviewed; performing multi-dimensional risk assessment on the flight plan data to determine a comprehensive risk index, wherein the comprehensive risk index is used to characterize the comprehensive risk magnitude of the flight plan data under multiple dimensions; determining the corresponding human-machine collaboration depth level based on the comprehensive risk index; determining a matching target review strategy from multiple preset human-machine collaboration review strategies based on the human-machine collaboration depth level, wherein different human-machine collaboration review strategies correspond to different human review intensities; and executing the target review strategy to complete the review of the flight plan data, thereby realizing the automated and intelligent operation of risk-based intelligent review and dynamic collaboration, thereby improving the review efficiency and effectiveness of flight plans.

[0172] To facilitate better implementation of the risk-based intelligent review dynamic collaboration method provided in this application, this application also provides a risk-based intelligent review dynamic collaboration device based on the aforementioned risk-based intelligent review dynamic collaboration method. The meanings of the terms used are the same as in the aforementioned risk-based intelligent review dynamic collaboration method, and specific implementation details can be found in the descriptions in the method embodiments.

[0173] Please see Figure 6 , Figure 6A structural block diagram of a risk-based intelligent audit dynamic collaborative device provided in this application embodiment, the device comprising: Acquisition unit 301 is used to acquire flight plan data to be reviewed; Evaluation unit 302 is used to conduct multi-dimensional risk assessment of flight plan data and determine the comprehensive risk index. The comprehensive risk index is used to characterize the comprehensive risk of flight plan data under multiple dimensions. The first determining unit 303 is used to determine the corresponding human-machine collaboration depth level based on the comprehensive risk index; The second determining unit 304 is used to determine a matching target review strategy from multiple preset human-machine collaboration review strategies based on the depth level of human-machine collaboration, wherein different human-machine collaboration review strategies correspond to different human review intensities. Execution unit 305 is used to execute the target review strategy to complete the review of flight plan data.

[0174] In some embodiments, the evaluation unit 302 may include: Construct sub-units to generate risk pre-assessment prompts corresponding to multiple preset risk dimensions; The assessment subunit is used to input risk pre-screening prompts and flight plan data into the risk assessment model, so that the risk assessment model outputs risk scores for the flight plan data under each risk dimension according to the guidance of the risk pre-screening prompts. The processing subunit is used to integrate the risk scores under each risk dimension to obtain a comprehensive risk index.

[0175] In some embodiments, the multiple preset human-machine collaborative review strategies include at least: a rapid human confirmation strategy, a key node human review strategy, and a full-process human review strategy.

[0176] In some embodiments, the target review strategy is a rapid manual confirmation strategy, and the execution unit 305 may include: The execution subunit is used to perform full-process automated review of flight plan data through artificial intelligence and generate review results. The full-process automated review includes at least one of the following: airspace compliance check, meteorological condition airworthiness check, aircraft qualification validity check, spatiotemporal conflict detection and applicant qualification verification. The review results include at least one of the following: artificial intelligence approval suggestions, risk index and risk level, results of each inspection item, and key information summary. The first push sub-unit is used to push the review results to the human review terminal for viewing and confirmation; The confirmation subunit is used to respond to the reviewer's confirmation operation on the review results and complete the approval of flight plan data.

[0177] In some embodiments, the target review strategy is a manual review strategy for key nodes, and the execution unit 305 may include: The execution subunit is used to perform full-process automated review of flight plan data through artificial intelligence. The full-process automated review includes at least one of the following: airspace compliance check, weather condition airworthiness check, aircraft qualification validity check, spatiotemporal conflict detection, and applicant qualification verification. The identification sub-unit is used to identify key risk nodes during the review process based on preset risk node identification conditions; The second push subunit is used to push the audit results and key risk points of the fully automated audit process to the manual audit end, so that the auditors can review the key risk points item by item. The first receiving subunit is used to receive the review results of the auditors for each key risk point, and to complete the approval of flight plan data based on the review results.

[0178] In some embodiments, the target review strategy is a full-process manual review strategy, and the execution unit 305 may include: The execution subunit is used to perform full-process automated review of flight plan data through artificial intelligence. The full-process automated review includes at least one of the following: airspace compliance check, weather condition airworthiness check, aircraft qualification validity check, spatiotemporal conflict detection, and applicant qualification verification. The third push subunit is used to push the audit results of the fully automated audit process, as well as flight plan data, to the manual audit end, so that the auditors can review all the check items in the audit results item by item. The second receiving subunit is used to receive the review results of the auditors for all inspection items, compare the review results of all inspection items with the review results of artificial intelligence, and obtain the comparison results. The fourth push subunit is used to push the comparison results to the manual review end, so that the reviewers can confirm the final review decision based on the comparison results and complete the review of the flight plan data.

[0179] In some embodiments, the first determining unit 303 may include: The mapping subunit is used to map the comprehensive risk index to the corresponding target risk level according to the preset risk index classification rules. The sub-unit is used to determine the depth level of human-machine collaboration from the preset depth levels of human-machine collaboration based on the target risk level. The preset depth levels of human-machine collaboration include at least: rapid human confirmation level, key node human review level, and full-process human review level.

[0180] In some embodiments, the device may further include: The recording unit is used to record the complete audit process and audit results during the execution of the target audit strategy; The generation unit is used to generate audit logs based on the audit process and audit results. The audit logs include at least one of the following: unique identifier of flight plan, comprehensive risk index, risk level, human-machine collaboration depth level, human-machine collaboration audit strategy implemented, artificial intelligence approval suggestions and basis, final decision of the auditors, approval comments filled in by the auditors, list of personnel involved in the approval, and operation timestamps for each audit stage.

[0181] This application discloses an intelligent dynamic collaborative audit device based on risk grading. The device acquires flight plan data to be audited via an acquisition unit 301; an evaluation unit 302 performs a multi-dimensional risk assessment on the flight plan data to determine a comprehensive risk index, which characterizes the overall risk level of the flight plan data across multiple dimensions; a first determination unit 303 determines the corresponding human-machine collaboration depth level based on the comprehensive risk index; a second determination unit 304 determines a matching target audit strategy from multiple preset human-machine collaboration audit strategies based on the human-machine collaboration depth level, wherein different human-machine collaboration audit strategies correspond to different levels of manual audit intensity; and an execution unit 305 executes the target audit strategy to complete the audit of the flight plan data. This device enables automated and intelligent operation of intelligent dynamic collaborative auditing based on risk grading, thereby improving the efficiency and effectiveness of flight plan auditing.

[0182] Accordingly, embodiments of this application also provide an electronic device. For example... Figure 7 As shown, Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device 400 includes a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, and a computer program stored in the memory 402 and executable on the processor. The processor 401 and the memory 402 are electrically connected. Those skilled in the art will understand that... Figure 7 The electronic device structures shown herein do not constitute a limitation on electronic devices and may include, but are not limited to, those shown. Figure 7 It can show more or fewer parts, or combine certain parts, or arrange different parts.

[0183] The processor 401 is the control center of the electronic device 400. It connects various parts of the electronic device 400 through various interfaces and lines. By running or loading software programs and / or modules stored in the memory 402, and calling data stored in the memory 402, it performs various functions of the electronic device 400 and processes data, thereby monitoring the electronic device 400 as a whole.

[0184] In this embodiment, the processor 401 in the electronic device 400 loads the instructions corresponding to the processes of one or more applications into the memory 402 according to the following steps, and the processor 401 runs the applications stored in the memory 402 to realize various functions: Obtain flight plan data pending review; A multi-dimensional risk assessment is conducted on flight plan data to determine a comprehensive risk index. The comprehensive risk index is used to characterize the overall risk level of flight plan data across multiple dimensions. The corresponding level of human-machine collaboration depth is determined based on the comprehensive risk index. Based on the depth level of human-machine collaboration, a matching target review strategy is determined from multiple preset human-machine collaboration review strategies. Different human-machine collaboration review strategies correspond to different human review intensities. Implement the target review strategy to complete the review of flight plan data.

[0185] This application embodiment acquires flight plan data to be reviewed; performs multi-dimensional risk assessment on the flight plan data to determine a comprehensive risk index, which characterizes the overall risk level of the flight plan data across multiple dimensions; determines the corresponding human-machine collaboration depth level based on the comprehensive risk index; based on the human-machine collaboration depth level, selects a matching target review strategy from multiple preset human-machine collaboration review strategies, wherein different human-machine collaboration review strategies correspond to different levels of manual review intensity; and executes the target review strategy to complete the review of the flight plan data. This achieves automated and intelligent operation of risk-level-based intelligent review dynamic collaboration, thereby improving the efficiency and effectiveness of flight plan review.

[0186] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0187] Optional, such as Figure 7 As shown, the electronic device 400 may further include a display 403 and an input unit 404. The processor 401 is electrically connected to both the display 403 and the input unit 404. Those skilled in the art will understand that... Figure 7 The electronic device structure shown does not constitute a limitation on the electronic device and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0188] Display 403 can be used to display a graphical user interface (GUI) and receive operation commands generated by the user interacting with the GUI. Display 403 may include a display panel and a touch panel. The display panel can be used to display information input by the user or information provided to the user, as well as various graphical user interfaces of the electronic device. These graphical user interfaces can be composed of graphics, guidance information, icons, video, and any combination thereof. Optionally, the display panel can be configured using a liquid crystal display (LCD), organic light-emitting diode (OLED), or other similar technologies. The touch panel can be used to collect touch operations performed by the user on or near it (such as operations performed by the user using a finger, stylus, or any suitable object or accessory on or near the touch panel), generate corresponding operation commands, and execute the corresponding program according to the operation commands. Optionally, the touch panel may include a touch detection device and a touch controller.

[0189] The touch detection device detects the user's touch location and the signal generated by the touch operation, transmitting the signal to the touch controller. The touch controller receives touch information from the touch detection device, converts it into touch point coordinates, and sends it to the processor 401. It can also receive and execute commands from the processor 401. The touch panel can cover the display panel. When the touch panel detects a touch operation on or near it, it transmits the information to the processor 401 to determine the type of touch event. Subsequently, the processor 401 provides corresponding visual output on the display panel based on the type of touch event. In this embodiment, the touch panel and display panel can be integrated into the display 403 to achieve input and output functions. However, in some embodiments, the touch panel and display panel can be implemented as two independent components to achieve input and output functions. That is, the display 403 can also be used as part of the input unit 404 to achieve input functions.

[0190] The input unit 404 can be used to receive input numbers, characters, or user characteristic information (such as fingerprints, iris, facial information, etc.), and to generate keyboard, mouse, joystick, optical, or trackball signal inputs related to user settings and function control.

[0191] In some embodiments, the electronic device may further include an audio circuit, which can provide an audio interface between the user and the device control device via a speaker and a microphone. The audio circuit can convert received audio data into electrical signals and transmit them to the speaker, where the speaker converts them into sound signals for output. Conversely, the microphone converts collected sound signals into electrical signals, which are then received by the audio circuit, converted back into audio data, and processed by the processor 401. The audio data is then transmitted via a radio frequency circuit to, for example, another device control device, or output to a memory 402 for further processing. The audio circuit may also include an earphone jack to provide communication between a peripheral headset and the device control device.

[0192] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0193] As can be seen from the above, the electronic device provided in this embodiment can acquire flight plan data to be reviewed; perform multi-dimensional risk assessment on the flight plan data to determine a comprehensive risk index, which is used to characterize the comprehensive risk of the flight plan data in multiple dimensions; determine the corresponding human-machine collaboration depth level based on the comprehensive risk index; determine the matching target review strategy from multiple preset human-machine collaboration review strategies based on the human-machine collaboration depth level, wherein different human-machine collaboration review strategies correspond to different manual review intensities; and execute the target review strategy to complete the review of the flight plan data.

[0194] Therefore, embodiments of this application provide a computer-readable storage medium storing a plurality of computer programs that can be loaded by a digital signal processor to execute the steps of any of the device control methods provided in embodiments of this application. For example, the computer program can execute the following steps: Obtain flight plan data pending review; A multi-dimensional risk assessment is conducted on flight plan data to determine a comprehensive risk index. The comprehensive risk index is used to characterize the overall risk level of flight plan data across multiple dimensions. The corresponding level of human-machine collaboration depth is determined based on the comprehensive risk index. Based on the depth level of human-machine collaboration, a matching target review strategy is determined from multiple preset human-machine collaboration review strategies. Different human-machine collaboration review strategies correspond to different human review intensities. Implement the target review strategy to complete the review of flight plan data.

[0195] This application embodiment acquires flight plan data to be reviewed; performs multi-dimensional risk assessment on the flight plan data to determine a comprehensive risk index, which characterizes the overall risk level of the flight plan data across multiple dimensions; determines the corresponding human-machine collaboration depth level based on the comprehensive risk index; based on the human-machine collaboration depth level, selects a matching target review strategy from multiple preset human-machine collaboration review strategies, wherein different human-machine collaboration review strategies correspond to different levels of manual review intensity; and executes the target review strategy to complete the review of the flight plan data. This achieves automated and intelligent operation of risk-level-based intelligent review dynamic collaboration, thereby improving the efficiency and effectiveness of flight plan review.

[0196] For details on the implementation of each of the above operations, please refer to the previous examples, which will not be repeated here.

[0197] The computer-readable storage medium may include: read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0198] Since the computer program stored in the computer-readable storage medium can execute the steps of any of the device control methods provided in the embodiments of this application, the beneficial effects that any of the device control methods provided in the embodiments of this application can achieve can be realized, as detailed in the preceding embodiments, and will not be repeated here.

[0199] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A dynamic collaborative intelligent auditing method based on risk classification, characterized in that, The method includes: Obtain flight plan data pending review; A multi-dimensional risk assessment is performed on the flight plan data to determine a comprehensive risk index, which is used to characterize the overall risk level of the flight plan data across multiple dimensions. The corresponding human-machine collaboration depth level is determined based on the comprehensive risk index. Based on the depth level of human-machine collaboration, a matching target review strategy is determined from multiple preset human-machine collaboration review strategies, wherein different human-machine collaboration review strategies correspond to different human review intensities. The target audit strategy is executed to complete the audit of the flight plan data.

2. The method according to claim 1, characterized in that, The multi-dimensional risk assessment of the flight plan data to determine the comprehensive risk index includes: Construct risk pre-assessment prompts corresponding to multiple preset risk dimensions; The risk pre-screening prompts and the flight plan data are input into the risk assessment model so that the risk assessment model outputs the risk scores of the flight plan data under each risk dimension according to the guidance of the risk pre-screening prompts. The risk scores under each risk dimension are fused to obtain the comprehensive risk index.

3. The method according to claim 1, characterized in that, The target review strategy is a rapid manual confirmation strategy, and executing the target review strategy includes: The flight plan data is automatically reviewed throughout the entire process using artificial intelligence, and the review results are generated. The fully automated review process includes at least one of the following: airspace compliance check, weather condition airworthiness check, aircraft qualification validity check, spatiotemporal conflict detection, and applicant qualification verification. The review results include at least one of the following: artificial intelligence approval suggestions, risk index and risk level, results of each inspection item, and key information summary. The review results will be sent to a human reviewer for viewing and confirmation. In response to the reviewer's confirmation of the review results, the approval of the flight plan data is completed.

4. The method according to claim 1, characterized in that, The target review strategy is a key node manual review strategy, and the execution of the target review strategy includes: The flight plan data is automatically reviewed throughout the entire process using artificial intelligence. The automatic review process includes at least one of the following: airspace compliance check, weather condition airworthiness check, aircraft qualification validity check, spatiotemporal conflict detection, and applicant qualification verification. During the review process, key risk points are identified based on preset risk point identification criteria. The results of the fully automated review process and key risk points are pushed to the manual review end so that the reviewers can review the key risk points item by item. The system receives the review results from auditors for each of the key risk points and completes the approval of the flight plan data based on the review results.

5. The method according to claim 1, characterized in that, The target review strategy is a full-process manual review strategy. The execution of the target review strategy includes: The flight plan data is automatically reviewed throughout the entire process using artificial intelligence. The automatic review process includes at least one of the following: airspace compliance check, weather condition airworthiness check, aircraft qualification validity check, spatiotemporal conflict detection, and applicant qualification verification. The results of the fully automated review process, along with the flight plan data, are pushed to the manual review end so that the reviewers can review each item in the review results. The system receives the review results of each of the inspection items from the auditors, compares these results with the audit results from the artificial intelligence, and obtains the comparison results. The comparison results are pushed to the manual review terminal so that the reviewers can confirm the final review decision based on the comparison results and complete the review of the flight plan data.

6. The method according to claim 1, characterized in that, The step of determining the corresponding human-machine collaboration depth level based on the comprehensive risk index includes: According to the preset risk index classification rules, the comprehensive risk index is mapped to the corresponding target risk level; Based on the target risk level, the human-machine collaboration depth level is determined from the preset human-machine collaboration depth levels, wherein the preset human-machine collaboration depth levels include at least: rapid manual confirmation level, key node manual review level, and full-process manual review level.

7. The method according to claim 1, characterized in that, The method further includes: During the execution of the target audit strategy, the complete audit process and audit results are recorded; An audit log is generated based on the audit process and the audit results. The audit log includes at least one of the following: unique identifier of flight plan, comprehensive risk index, risk level, human-machine collaboration depth level, human-machine collaboration audit strategy implemented, artificial intelligence approval suggestions and basis, final decision of the auditor, approval opinions filled in by the auditor, list of personnel involved in the approval, and operation timestamps for each audit stage.

8. A risk-based intelligent auditing dynamic collaborative device, characterized in that, The device includes: The acquisition unit is used to acquire flight plan data to be reviewed; An assessment unit is used to perform multi-dimensional risk assessment on the flight plan data and determine a comprehensive risk index, which is used to characterize the comprehensive risk of the flight plan data in multiple dimensions. The first determining unit is used to determine the corresponding human-machine collaboration depth level based on the comprehensive risk index; The second determining unit is used to determine a matching target review strategy from multiple preset human-machine collaboration review strategies based on the human-machine collaboration depth level, wherein different human-machine collaboration review strategies correspond to different human review intensities. An execution unit is used to execute the target review strategy to complete the review of the flight plan data.

9. An electronic device, characterized in that, The electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the program to implement the risk-based intelligent audit dynamic collaborative method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores multiple instructions adapted for loading by a processor to execute the risk-based intelligent audit dynamic collaboration method according to any one of claims 1 to 7.

Images