Flight decision guidance system for special situation and maneuvering flight of helicopter

Abstract

The invention discloses a flight decision guidance system for special situation and maneuvering flight of a helicopter, and the system comprises a main program which comprises a pre-loading decision tree, a data reading unit, a data analysis unit, a subject judgment unit, a prediction task scheduling unit and a control quantity output unit, serves as a general control core of the system, and is used for carrying out the global scheduling function; the control subprogram comprises a loading prediction unit, a division and judgment flight unit and a calling prediction model unit, is used as a subject special processor and undertakes a service logic processing function; and the prediction subprogram comprises an input data unit, a data preprocessing unit, a prediction manipulated variable unit and a result processing and output manipulated variable unit, is used as an intelligent decision engine of the system, and is responsible for core model reasoning calculation. The flight subjects can be automatically identified through the received data, the flight stages are divided in real time, manual intervention of pilots is not needed, the workload is relieved, and human judgment errors are avoided.

Description

Technical Field

[0001] This invention relates to the technical field of helicopter flight, and more particularly to a flight decision guidance system for helicopter special situations and maneuvering flight. Background Technology

[0002] In the 21st century, with the rapid development of avionics technology, sensor-based flight guidance systems have emerged. Traditional flight decision support systems are primarily based on standard operating procedures in flight manuals and the accumulated experience of individual pilots. While this approach has developed a relatively complete theoretical framework through long-term practice, it has significant limitations: static flight manuals cannot provide real-time, quantitative operational guidance; pilots are prone to decision delays and judgment errors under high-pressure environments; and the system lacks the ability to accurately perceive flight conditions. Although modern avionics systems possess multi-sensor data fusion capabilities, most systems remain at the level of parameter display and basic warnings, failing to achieve true intelligent decision support.

[0003] With the development of artificial intelligence technology, data-driven intelligent decision-making methods have shown application potential in the aviation field. These methods establish a mapping relationship between flight status and control strategies by analyzing historical flight data. However, existing research has significant shortcomings: most studies focus on the routine flight status of fixed-wing aircraft, while research on the maneuvering flight and emergency response needs of helicopters is relatively weak; at the technical level, they focus on single-point breakthroughs and lack systematic solutions; in particular, a complete system has not yet been formed in key technologies such as multi-stage flight recognition and real-time decision-making.

[0004] Furthermore, with the widespread application of helicopters in military, rescue, and other fields, flight missions are becoming increasingly complex, and the requirements for safety and standardization are constantly increasing. Traditional flight decision-making based on personal experience is no longer sufficient to meet the needs of modern flight missions, and there is an urgent need to establish a more scientific flight decision guidance system. However, the existing technology is clearly insufficient in its support capabilities in this area and cannot provide real-time, reliable decision support.

[0005] Therefore, developing an intelligent decision-making system capable of identifying flight phases in real time and providing quantitative operational recommendations not only has theoretical innovation value but also significant practical implications for improving flight safety and standardizing emergency response procedures. Such a system needs to break through traditional limitations, integrate data-driven technologies and domain knowledge, and establish a complete intelligent decision-making system to provide reliable assurance for helicopter flight safety. Summary of the Invention

[0006] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0007] In view of the problems existing in the flight decision guidance system for helicopter special situations and maneuvering flight, the present invention is proposed.

[0008] Therefore, the purpose of this invention is to provide a flight decision guidance system for helicopter special situations and maneuvering flights, which can automatically identify flight subjects and divide flight phases in real time through received data, without the need for manual intervention by the pilot, thereby reducing workload and avoiding human judgment errors.

[0009] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a flight decision guidance system for helicopter special situations and maneuvering flight, comprising: The main program, including the preload decision tree, data reading unit, data analysis unit, subject judgment unit, prediction task scheduling unit, and output control quantity unit, serves as the overall control core of the system and undertakes the global scheduling function. The control subroutine, including loading the prediction unit, dividing and judging the flight unit, and calling the prediction model unit, serves as a subject-specific processor and undertakes the business logic processing function. The prediction subroutine, including the input data unit, data preprocessing unit, prediction manipulation unit, result processing and output manipulation unit, serves as the system's intelligent decision engine and is responsible for the core model inference calculation. The three core programs form a close collaborative relationship through standardized data interfaces and calling protocols: the main program provides data input and result output channels, controls the subprograms to implement business logic processing, and predicts the subprograms to complete core computing tasks.

[0010] As a preferred embodiment of the flight decision guidance system for helicopter special situations and maneuvering flight described in this invention, the following steps are taken: the pre-loading decision tree is executed upon system startup, and all machine learning models, decision trees, and normalizers for the four special situations and maneuvering subjects are loaded into the memory cache through a pre-loading function; the data reading unit receives real-time helicopter flight data streams at high frequency through a data bus interface, and the data packets are encapsulated using a binary protocol, including a message header, data body, and checksum; the data parsing unit performs format parsing and content extraction on the received data packets, including: verifying the integrity of the message header identifier and checksum; decoding key flight parameters; and filtering physically unreasonable data through numerical range checks to ensure the validity of the input data.

[0011] As a preferred embodiment of the flight decision guidance system for helicopter special situations and maneuvering flight described in this invention, the subject judgment unit parses the value of the special situation identifier field and accurately identifies the flight subject type and implements data routing according to the predefined subject identifier; the predictive task scheduling unit is a system in which the main program calls the corresponding subject processing function through a dynamic scheduling mechanism to establish a complete task execution link and maintains global state variables to track the current flight phase of each subject.

[0012] As a preferred embodiment of the flight decision guidance system for helicopter special situations and maneuvering flight described in this invention, the output control unit is a system that receives prediction result data packets returned by each subject processing unit, standardizes and encapsulates the prediction results according to a preset unified data format specification, and transmits the encapsulated data packets to the flight display system through a high-speed data bus to present them to the pilot in a visual manner.

[0013] As a preferred embodiment of the flight decision guidance system for helicopter special situations and maneuvering flight described in this invention, the loading prediction unit controls the subroutine to import the corresponding prediction module during initialization based on the subject type information passed by the main function, establishes a call relationship with the prediction subroutine, and allocates necessary memory resources for data processing and status maintenance.

[0014] As a preferred embodiment of the flight decision guidance system for helicopter special situations and maneuvering flight described in this invention, the following steps are taken: the division and judgment of each subject in the flight unit adopts a dedicated stage judgment mechanism based on its flight characteristics, and the accuracy of stage identification is ensured through multi-parameter analysis; the prediction model unit is called to pass a complete data dictionary containing stage information to the prediction subroutine, triggering parallel prediction calculation, controlling the subroutine to synchronously wait for the prediction to complete, and then returning to the main program.

[0015] As a preferred embodiment of the flight decision guidance system for helicopter special situations and maneuvering flight described in this invention, the input data unit is the flight data divided by the prediction subroutine from the control subroutine receiving stage. First, data integrity verification is performed: checking whether the required fields exist, whether the numerical range is reasonable, and whether the data format meets the requirements. For missing or abnormal data, the default value or the previous valid value is used as a substitute.

[0016] As a preferred embodiment of the flight decision guidance system for helicopter special situations and maneuvering flight described in this invention, the data preprocessing unit first performs feature extraction and preprocessing on the input data: converting the dictionary data format; selecting features according to the predefined feature list for each subject; and performing dynamic feature calculation for the autopilot gliding subject. Secondly, it calls a preloaded normalizer to normalize the feature data: obtaining the normalizer object from the corresponding model cache dictionary; transforming the feature data to eliminate the influence of dimensions; and ensuring that the input features maintain the same distribution characteristics as those during model training.

[0017] As a preferred embodiment of the flight decision guidance system for helicopter special situations and maneuvering flight described in this invention, the predicted control quantity unit first utilizes a global thread pool to achieve parallel prediction of four control quantities: creating a prediction task for each control quantity; submitting tasks through the thread pool, with a maximum concurrency of 4; establishing a mapping relationship between future objects and control targets to facilitate result collection; secondly, performing independent quantile regression prediction for each control quantity: obtaining pre-trained models for the corresponding stage and control quantity from the model cache; calculating the predicted values ​​of the 20th and 80th quantiles as the lower and upper bounds of the control interval, respectively.

[0018] As a preferred embodiment of the flight decision guidance system for helicopter special situations and maneuvering flight described in this invention, the result processing unit performs rounding and format normalization on the original prediction results: retaining 3 decimal places for the prediction value; checking the rationality of the boundaries to ensure that the lower bound is not greater than the upper bound; correcting abnormal prediction results or adopting a default safe range; the output manipulation unit collects the prediction results of all threads, sorts and integrates them according to the predefined manipulation order, generates a structured prediction result dictionary, and returns it to the control subroutine.

[0019] The beneficial effects of this invention are: First, a high degree of intelligence and automation: the system can automatically identify flight subjects and divide flight phases in real time through received data, without the need for manual intervention by the pilot, reducing workload and avoiding human judgment errors; Second, the quantification and precision of decision support: abandoning the traditional qualitative prompts, it adopts quantile regression prediction based on machine learning models to provide pilots with specific numerical range suggestions for key control quantities; Third, modular design: The system adopts a modular design, with each subject equipped with an independent processing unit, and can be flexibly expanded and maintained through standardized interfaces; Fourth, high real-time performance: Through model preloading and parallel computing optimization, the system ensures that it meets the real-time processing requirements of high-frequency data. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a schematic diagram of the main program of the flight decision guidance system for helicopter special situations and maneuvering flight according to the present invention.

[0021] Figure 2 This is a schematic diagram of the control subroutine of the flight decision guidance system for helicopter special situations and maneuvering flight according to the present invention.

[0022] Figure 3 This is a schematic diagram of the prediction subroutine of the flight decision guidance system for helicopter special situations and maneuvering flight according to the present invention. Detailed Implementation

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0026] Secondly, the present invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of the present invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not according to the usual scale. Furthermore, the schematic diagrams are merely examples and should not limit the scope of protection of the present invention. In addition, actual fabrication should include three-dimensional spatial dimensions of length, width, and depth.

[0027] Reference Figures 1-3 A flight decision guidance system for helicopter special situations and maneuvering flight is provided, including: The main program, including the preload decision tree, data reading unit, data analysis unit, subject judgment unit, prediction task scheduling unit, and output control quantity unit, serves as the overall control core of the system and undertakes the global scheduling function. The control subroutine, including loading the prediction unit, dividing and judging the flight unit, and calling the prediction model unit, serves as a subject-specific processor and undertakes the business logic processing function. The prediction subroutine, including the input data unit, data preprocessing unit, prediction manipulation unit, result processing and output manipulation unit, serves as the system's intelligent decision engine and is responsible for the core model inference calculation. The three core programs form a close collaborative relationship through standardized data interfaces and calling protocols: the main program provides data input and result output channels, controls the subprograms to implement business logic processing, and predicts the subprograms to complete core computing tasks.

[0028] The main program, serving as the system's central control core, undertakes global scheduling functions. Its processing flow begins with model preloading during system initialization, followed by a real-time data processing loop: reading flight data streams through the data bus interface, performing data parsing and format verification, and establishing a unified time reference. The system identifies and routes tasks based on special situation identifier fields, using this to call subroutines. Finally, it integrates the prediction results returned by each task and outputs them in a unified format. The main function coordinates the work of each module through the task scheduler, ensuring the orderly and real-time nature of the system's processing flow.

[0029] Control Subroutine: As a subject-specific processor, it undertakes business logic processing functions. Its processing flow includes: loading the prediction module for the corresponding subject, executing subject-specific stage division logic, maintaining stage state, calling the prediction subroutine to perform parallel prediction, and returning the prediction results to the main function. The control subroutine achieves seamless integration with the main function through the subject adapter and establishes efficient communication with the prediction subroutine through the prediction interface.

[0030] The prediction subroutine, serving as the system's intelligent decision engine, is responsible for core model inference computation. Its processing flow encompasses: data validation and integrity checks, data preprocessing and standardization, multi-threaded parallel prediction task allocation, synchronous prediction computation of the four manipulators, and boundary determination and formatting of prediction results. The prediction subroutine ensures input quality through a preprocessing module, improves processing efficiency through a parallel computing engine, and guarantees output standardization through a result formatting module.

[0031] The three core programs work closely together through standardized data interfaces and calling protocols: the main function provides data input and output channels, the control subroutine handles business logic processing, and the prediction subroutine completes core computational tasks. This layered architecture ensures both the modularity and scalability of the system, as well as the efficiency and real-time performance of data processing, forming a complete helicopter flight decision guidance system.

[0032] The main program specifically includes: 1. Preloading Decision Trees: The model preloading phase is executed at system startup. A preloading function loads all machine learning models, decision trees, and normalizers for the four special and maneuver subjects into the memory cache. All models are stored in a global dictionary, with a three-level index structure for fast access.

[0033] 2. Data Reading Unit: Receives real-time helicopter flight data streams at high frequency via the data bus interface. Data packets are encapsulated using a binary protocol, including a header, data body, and checksum. The reading module employs a flow control strategy to ensure no data packets are lost in high-frequency data stream environments.

[0034] 3. Data Parsing Unit: This unit parses the received data packets, extracting their format and content. This includes verifying the integrity of the message header identifier and checksum; decoding key flight parameters; and filtering physically invalid data through numerical range checks to ensure the validity of the input data. It also establishes a unified system time base, handles data interruption situations, and provides accurate time references for subsequent stage judgments.

[0035] 4. Subject Unit Determination: Parse the value of the special situation identifier field and accurately identify the flight subject type and implement data routing based on the predefined subject identifier. Distribute the data to the corresponding processing channel according to the identification result, supporting multi-subject parallel processing capabilities.

[0036] 5. Predictive Task Scheduling Unit: The main control module uses a dynamic scheduling mechanism to call the corresponding subject processing functions, establish a complete task execution chain, and maintain global state variables to track the current flight phase of each subject. The system sets task execution timeout thresholds to ensure real-time requirements.

[0037] 6. Output Control Unit: The system receives the prediction result data packets returned by the processing units of each subject, and standardizes and encapsulates the prediction results according to a preset unified data format specification. The encapsulated data packets are transmitted to the flight display system via a high-speed data bus and presented to the pilot in a visual manner.

[0038] The control subroutine specifically includes: 1. Load Prediction Unit: Based on the subject type information passed by the main function, control the subroutine to import the corresponding prediction module during initialization, establish the call relationship with the prediction subroutine, and allocate necessary memory resources for data processing and state maintenance.

[0039] 2. Classification and identification of flight units: Each subject employs a dedicated phase identification mechanism based on its flight characteristics, ensuring the accuracy of phase identification through multi-parameter analysis. 1) The obstacle course converts GPS coordinates to runway coordinates and makes judgments in 8 stages based on cumulative flight distance and preset thresholds; 2) The emergency pull / pull maneuver exercise uses Y-axis overload value and maneuver duration, combined with time thresholds, to determine six stages; 3) The autogyro descent maneuver involves seven stages of assessment based on radio altitude and estimated time of landing (TTI). 4) The rapid lateral movement subject determines the four stages based on the ground speed threshold and stage instructions.

[0040] The control subroutine simultaneously maintains subject-specific state variables (such as reference point coordinates for slalom skiing subjects and stage start times for pull / push subjects), coordinates the timing relationship between stage judgments and prediction calls, and handles state resets and parameter updates during stage transitions.

[0041] Calling the prediction model unit: Passes the complete data dictionary containing stage information to the prediction subroutine, triggers parallel prediction calculation, controls the subroutine to wait synchronously for the prediction to complete, and then returns it to the main program.

[0042] The prediction subroutine specifically includes: 1. Input Data Unit: The prediction subroutine receives flight data segmented from the control subroutine and first performs data integrity verification: checking the existence of required fields, the reasonableness of numerical ranges, and the compliance of data format requirements. For missing or abnormal data, default values ​​or the previous valid value are used as replacements.

[0043] 2. Data Preprocessing Unit: First, feature extraction and preprocessing are performed on the input data: the dictionary data is converted to a new format; features are selected based on the predefined feature list for each subject; for the autobiography subject, dynamic feature calculation is also performed. Second, the preloaded normalizer is called to normalize the feature data: the normalizer object is retrieved from the corresponding model cache dictionary; the feature data is transformed to eliminate the influence of units; ensuring that the input features maintain the same distribution characteristics as those during model training.

[0044] 3. Manipulation Variable Prediction Unit: First, parallel prediction of four manipulation variables is implemented using a global thread pool: a prediction task is created for each manipulation variable; tasks are submitted through the thread pool, with a maximum concurrency of 4; a mapping relationship between future objects and manipulation targets is established to facilitate result collection. Second, independent quantile regression prediction is performed for each manipulation variable: pre-trained models for the corresponding stage and manipulation variable are retrieved from the model cache; the predicted values ​​of the 20th and 80th quantiles are calculated as the lower and upper bounds of the manipulation interval, respectively; all prediction models use a random forest model and a quantile regression algorithm.

[0045] 4. Result Processing Unit: Rounds and standardizes the format of the original prediction results: retains 3 decimal places for the predicted value; checks the reasonableness of the boundary to ensure that the lower bound is not greater than the upper bound; corrects abnormal prediction results or adopts the default safe range.

[0046] 5. Output Manipulation Unit: Collects the prediction results from all threads, sorts and integrates them according to the predefined manipulation order, generates a structured prediction result dictionary, and returns it to the control subroutine.

[0047] The specific implementation is as follows: Model preloading implementation: During system startup, machine learning models and normalizers for each subject are loaded into the memory cache via a preloading function. The specific implementation process is as follows: First, the model file storage path and subject / stage configuration are defined. Second, normalizer files and model files for each stage are loaded using a predefined library. Finally, the loaded models are stored in a global dictionary, establishing a three-level index structure of "subject-stage-manipulation quantity" to provide fast access for real-time prediction.

[0048] Data reception and routing implementation: The system receives real-time flight data streams at high frequency through a data bus interface, with data packets encapsulated using a binary protocol. The main program implements the core logic of data routing, with the following specific steps: First, data parsing and verification: The received data packets are parsed to extract key flight parameters such as GPS timestamps, latitude and longitude coordinates, attitude angles, velocity parameters, acceleration, and overload values. The physical rationality of the data is verified by checking its numerical range. Second, emergency identification and route allocation: The values ​​of the emergency identification field are parsed, and data is distributed according to predefined subject identifiers. Finally, GPS time information is extracted and a unified system time base is established based on this, identifying the starting time of the emergency and checking the continuity of timestamps to provide accurate time references for subsequent stage judgments.

[0049] Phase judgment logic implementation: The flight phase identification process employs a subject-specific intelligent judgment algorithm. Each processing unit continuously monitors changes in multi-dimensional flight parameters to achieve precise phase division. Based on preset judgment conditions and real-time flight data streams, the system accurately identifies the current flight phase through multi-source information fusion analysis and maintains a complete state mechanism to track the entire phase transition process. This ensures the accuracy and continuity of phase identification and provides reliable phase state information for subsequent prediction modules. In specific implementation, each subject adopts differentiated judgment strategies based on its flight characteristics. For obstacle skiing, the system achieves accurate judgment of multiple phases through spatial coordinate transformation and trajectory analysis, combined with cumulative flight distance calculation. For self-rotating descent special situations, the system adopts a dual-parameter decision model of altitude and time, achieving intelligent identification of multiple phases by monitoring the dynamic changes in flight altitude and expected landing time in real time. In the handling of pull-up / pull-down maneuvers, the system focuses on monitoring the time series characteristics of overload parameters and achieves phase division by combining preset time thresholds and the duration characteristics of maneuver operations. For rapid lateral movement maneuvers, the system constructs a state recognition mechanism based on speed characteristics, achieving phase judgment by monitoring ground speed changes and combining them with flight commands. Each subject processing unit is equipped with a comprehensive status maintenance mechanism, which ensures the integrity and reliability of flight phase information by updating phase parameters in real time and recording status transition times.

[0050] Parallel prediction implementation: Each subject's prediction subroutine adopts a unified multi-threaded parallel prediction architecture. The specific implementation process is as follows: First, feature preprocessing is performed; the input data is converted to the required format, features are selected according to a predefined feature list, and a pre-loaded normalizer is called to normalize the feature data, ensuring that the input features maintain the same distribution characteristics as during model training. Second, multi-threaded parallel prediction is executed; parallel prediction tasks are created for the four manipulation variables using a global thread pool, and parallel computation with a maximum concurrency of 4 is achieved through function modules, establishing a mapping relationship between future objects and manipulation targets for result collection. Third, quantile regression prediction is performed; pre-trained models for the corresponding stage and manipulation variable are obtained from the model cache, and the predicted values ​​of the 20th and 80th quantiles are calculated as the lower and upper bounds of the manipulation interval, respectively. All prediction models use a random forest model and a quantile regression algorithm. Finally, the prediction results are post-processed and integrated. The original prediction results are rounded to three decimal places, the boundary is checked to ensure that the lower bound is not greater than the upper bound, and the results are sorted and integrated according to the predefined manipulation variable order, returning a structured dictionary of prediction results.

[0051] Output results: The main program receives the prediction result data packets returned by each subject processing unit and encapsulates the prediction results in a unified output format. The specific format is: "Subject type|GPS time|[lower bound of longitudinal periodic range, upper bound of longitudinal periodic range],[lower bound of lateral periodic range, upper bound of lateral periodic range 2],[lower bound of total range, upper bound of total range],[lower bound of pedal, upper bound of pedal]|End".

[0052] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A flight decision guidance system for helicopter special situations and maneuvering flight, characterized in that, include: The main program, including the preload decision tree, data reading unit, data analysis unit, subject judgment unit, prediction task scheduling unit, and output control quantity unit, serves as the overall control core of the system and undertakes the global scheduling function. The control subroutine, including loading the prediction unit, dividing and judging the flight unit, and calling the prediction model unit, serves as a subject-specific processor and undertakes the business logic processing function. The prediction subroutine, including the input data unit, data preprocessing unit, prediction manipulation unit, result processing and output manipulation unit, serves as the system's intelligent decision engine and is responsible for the core model inference calculation. The three core programs form a close collaborative relationship through standardized data interfaces and calling protocols: the main program provides data input and result output channels, controls the subprograms to implement business logic processing, and predicts the subprograms to complete core computing tasks.

2. The flight decision guidance system for helicopter special situations and maneuvering flight according to claim 1, characterized in that: The preload decision tree is executed when the system starts up. The preload function loads all machine learning models, decision trees and normalizers for the four special situations and maneuver subjects into the memory cache. The data reading unit receives real-time flight data streams of the helicopter at high frequency through the data bus interface. The data packets are encapsulated using a binary protocol and include a message header, data body and checksum. The data parsing unit performs format parsing and content extraction on the received data packets, including: verifying the integrity of the message header identifier and checksum; Decode key flight parameters; filter physically unreasonable data through numerical range checks to ensure the validity of input data.

3. The flight decision guidance system for helicopter special situations and maneuvering flight according to claim 2, characterized in that: The subject judgment unit parses the value of the special situation identifier field and accurately identifies the flight subject type and implements data routing based on the predefined subject identifier; the predictive task scheduling unit is the main program that calls the corresponding subject processing function through a dynamic scheduling mechanism to establish a complete task execution link and maintains global state variables to track the current flight phase of each subject.

4. The flight decision guidance system for helicopter special situations and maneuvering flight according to claim 3, characterized in that: The output control unit is responsible for receiving the prediction result data packets returned by the subject processing units, standardizing and encapsulating the prediction results according to the preset unified data format specifications, and transmitting the encapsulated data packets to the flight display system through a high-speed data bus to present them to the pilot in a visual manner.

5. The flight decision guidance system for helicopter special situations and maneuvering flight according to claim 1, characterized in that: The loading prediction unit controls the subroutine to import the corresponding prediction module during initialization based on the subject type information passed by the main function, establishes a call relationship with the prediction subroutine, and allocates necessary memory resources for data processing and status maintenance.

6. The flight decision guidance system for helicopter special situations and maneuvering flight according to claim 5, characterized in that: The division and judgment of each subject in the flight unit adopts a dedicated stage judgment mechanism based on its flight characteristics, and ensures the accuracy of stage identification through multi-parameter analysis; the prediction model unit is called to pass the complete data dictionary containing stage information to the prediction subroutine, triggering parallel prediction calculation, controlling the subroutine to wait synchronously for the prediction to be completed, and then returning to the main program.

7. The flight decision guidance system for helicopter special situations and maneuvering flight according to claim 1, characterized in that: The input data unit is the flight data divided by the prediction subroutine from the control subroutine receiving stage. First, data integrity verification is performed: check whether the required fields exist, whether the numerical range is reasonable, and whether the data format meets the requirements. For missing or abnormal data, the default value or the previous valid value is used to replace it.

8. The flight decision guidance system for helicopter special situations and maneuvering flight according to claim 7, characterized in that: The data preprocessing unit first performs feature extraction and preprocessing on the input data: converting the dictionary data into a new format; selecting features based on the predefined feature list for each subject; and performing dynamic feature calculation for the autobiography subject. Secondly, it calls the preloaded normalizer to normalize the feature data: obtaining the normalizer object from the corresponding model cache dictionary. Transform the feature data to eliminate the influence of dimensions; ensure that the input features maintain the same distribution characteristics as those during model training.

9. The flight decision guidance system for helicopter special situations and maneuvering flight according to claim 8, characterized in that: The predictive manipulator unit first utilizes a global thread pool to implement parallel prediction of four manipulators: creating a prediction task for each manipulator. Tasks are submitted via a thread pool, with a maximum concurrency of 4; a mapping relationship is established between future objects and manipulation targets to facilitate result collection. Next, perform independent quantile regression prediction for each manipulation: retrieve the pre-trained model for the corresponding stage and manipulation from the model cache; The predicted values ​​of the 20th and 80th percentiles are calculated as the lower and upper bounds of the manipulation interval, respectively.

10. The flight decision guidance system for helicopter special situations and maneuvering flight according to claim 9, characterized in that: The result processing unit rounds and standardizes the format of the original prediction results: retains 3 decimal places for the predicted value; checks the rationality of the boundaries to ensure that the lower bound is not greater than the upper bound; corrects abnormal prediction results or adopts a default safe range; the output manipulation unit collects the prediction results of all threads, sorts and integrates them according to the predefined manipulation order, generates a structured prediction result dictionary, and returns it to the control subroutine.

Images