Low-altitude pilot AI model training application system

Abstract

The invention, which relates to the technical field of aviation training, discloses a low-altitude pilot AI model training application system comprising a data module, an intelligent decision-making module, a flight simulation module, a communication module and an interface module. The data module comprises an aircraft type database, a flight operation knowledge base, a meteorological data interface and an airspace and airport database; and the aircraft type database is used for storing eVTOL detailed information of various aircrafts. According to the low-altitude pilot AI model training application system provided by the invention, by integrating the aircraft type database, the flight operation knowledge base, the real-time meteorological data and the airspace and airport database, a highly-simulated training environment is provided for a pilot, so that the coping ability of the pilot under the conditions of complex weather and special airspace can be effectively improved, and the coping ability of the pilot under the conditions of complex weather and special airspace can be improved. The training cost can be remarkably reduced, the training period is shortened, the flight state can be monitored and evaluated in real time through the built-in intelligent decision module, and the authenticity of flight simulation is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of aviation training, and specifically to an AI model training application system for low-altitude pilots. Background Art

[0002] With the accelerated layout of the urban air mobility (UAM) industry, as a representative of the new generation of green aircraft, eVTOL aircraft have entered the commercial implementation stage globally. As of 2025, more than 20 cities around the world have launched air traffic pilot projects. The six-rotor compound-wing model independently developed by Chinese eVTOL manufacturers has completed a cumulative 100,000 flight hours of testing. It is expected that the urban low-altitude travel market scale will exceed 100 billion yuan by 2030. This technological innovation poses new requirements for the pilot training system: The unique vertical takeoff and landing characteristics, distributed electric propulsion system, and full fly-by-wire control mode of eVTOL make it challenging for traditional fixed-wing or helicopter pilots to transfer their skills. At the same time, low-altitude urban flight needs to deal with complex scenarios such as three-dimensional obstacle avoidance in building clusters, drone swarm cooperation, and dynamic airspace management, which puts higher requirements on the multi-modal interaction decision-making ability of pilots.

[0003] There is room for improvement in the existing training system to meet new demands: Traditional flight training relies on a combination of physical aircraft flight and basic simulators, and the training cost for a single pilot is relatively high. The actual flight proportion of special situation handling training is about 15%. Most mainstream flight simulators construct training scenarios based on preset flight routes and standard meteorological conditions, and there are technical bottlenecks in complex special situation simulation and dynamic scenario construction. In addition, there is still room for improvement in the personalized training mechanism and the in-depth utilization of training data in the existing system, and it is necessary to construct a more perfect correlation model between pilot operation characteristics and training effectiveness. In response to this, we propose an AI model training application system for low-altitude pilots. Summary of the Invention

[0004] To solve the above technical problems and provide an AI model training application system for low-altitude pilots, the present technical solution solves the above problems.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An AI model training application system for low-altitude pilots, including: a data module, an intelligent decision module, a flight simulation module, a communication module, and an interface module; The data module includes: an aircraft type database, a flight operation knowledge base, a meteorological data interface, and an airspace and airport database; the aircraft type database is used to store detailed information of various eVTOL aircraft, and the detailed information includes: model, performance parameters, and operating characteristics; the flight operation knowledge base is used to store flight operation procedures and skills; the meteorological data interface is used to connect to an external meteorological source to obtain meteorological information; the airspace and airport database is used to store airspace and airport-related data; The intelligent decision-making module includes: a flight status evaluation unit, a mission planning unit, and a decision-making unit; the flight status evaluation unit is used to monitor the state of the simulated aircraft and compare it with the standard; the mission planning unit is used to formulate a mission plan based on the knowledge base and the environment; the decision-making unit is used to formulate operation decisions through artificial intelligence algorithms; The flight simulation module is used to simulate the physical movement of the aircraft; The communication module is used to achieve communication with the simulated air traffic control environment and the pilot of the aircraft; The interface module is used to provide an interface for integration with flight simulation training equipment and instructor configuration.

[0006] Preferably, in the data module, the formula for calculating the maximum speed of the aircraft performance parameters in the aircraft type database is: , In the formula, is the maximum speed of the aircraft, is the maximum thrust of the aircraft, is the air density, is the wing area; The cruise speed is determined according to the aircraft engine efficiency curve and flight altitude factors, and is within the speed range that satisfies the engine battery power consumption rate at a preset altitude; The handling characteristics are measured by the force feedback coefficient according to the sensitivity of the joystick. When the pilot applies a force , the deflection angle of the control surface is . The values of different aircraft models are different, and are determined according to the mechanical structure and hydraulic / electrical transmission system design of the aircraft; , In the formula, is the acceleration of the aircraft, is the engine thrust, is the aircraft drag, is the mass of the aircraft; When the speed reaches the takeoff decision speed , it is judged whether the takeoff condition is satisfied according to the relationship between the lift coefficient of the aircraft and the speed, and then the wheel-raising operation is performed. The expression of the relationship between the lift coefficient of the aircraft and the speed is: , In the formula, is the lift, is the airspeed, is the lift coefficient; And communicate with air traffic control at a specified frequency throughout the process to report the aircraft status; The landing operation procedure involves determining the glide path angle based on the airport runway conditions and meteorological information , guiding the aircraft through the instrument landing system, and precisely controlling the sink rate at the moment of landing , satisfying: , In the formula, is the sink rate at the moment of landing, is the aircraft weight, is the glide path angle, ensuring a smooth landing of the aircraft and providing comprehensive and accurate operation references for the pilot; Among the meteorological information obtained by the meteorological data interface, the influence of wind speed on the aircraft flight trajectory is calculated through the aircraft ground speed calculation formula: , where is the ground speed, is the airspeed, is the wind speed, is the angle between the wind direction and the aircraft heading; Air temperature affects air density, thereby changing the aircraft performance. According to the ideal gas state equation: , where is the air pressure, is the gas constant, is the air temperature. At different air temperatures, the thrust and lift parameters of the aircraft engine need to be adjusted accordingly. The low-altitude pilot AI model adjusts flight operation decisions in real time according to changes in meteorological factors; In the airspace and airport database, for airspace structure information, the division of control areas is based on the standards of the International Civil Aviation Organization; Above a specific altitude is the high-altitude control area, and below is the low-altitude control area. Different control areas have different communication frequencies and control rules; The flight altitude layer allocation rule follows the standard pressure altitude layer setting and is divided at intervals of 300 meters; In terms of airport data, the runway length and direction determine the aircraft takeoff and landing performance requirements and operation methods.

[0007] Preferably, the method steps for the flight state evaluation unit to monitor the simulated aircraft state and compare it with the standard are as follows: Obtain the real-time state data of the simulated aircraft based on the flight simulation module, and extract the standard state data corresponding to the flight phase and mission from the flight operation knowledge base; Based on the collected data, preprocess the data. The preprocessing includes: removing noise, outliers, and missing values, and standardizing the cleaned data; Compare the pre - processed real - time status data with the standard status data parameter by parameter. For each status parameter, calculate the difference between it and the standard value. According to the comparison results of each parameter, use the weighted comprehensive evaluation method to evaluate the overall flight status of the aircraft; Generate a detailed flight status report according to the evaluation results. The flight status report includes: the comparison results of each status parameter and the comprehensive evaluation index, and feedback the flight status report to the intelligent decision - making module and the interface module.

[0008] Preferably, when the flight status evaluation unit of the intelligent decision - making module monitors the aircraft's power, according to the aircraft engine battery power consumption rate and flight time , calculate the remaining power: , where, is the remaining power, is the initial power; When the power is lower than the safety threshold, issue an alarm in time and plan to land at the nearest suitable airport for an alternate landing; When evaluating the aircraft's attitude stability, by measuring the change rates of the aircraft's pitch angle, roll angle, and yaw angle, if the change rates exceed the preset stable range, combine the aircraft dynamics model to judge whether there is a risk of out - of - control, and take corresponding adjustment measures.

[0009] Preferably, when the task planning unit of the intelligent decision - making module plans the optimal route, use Dijkstra's algorithm to find the shortest path from the starting point to the destination; Regard the navigation points in the airspace as the nodes of the graph, and the distance between nodes as the flight cost. Comprehensively consider meteorological conditions, airspace restrictions, and other aircraft dynamic factors, and weight the path cost; In areas with strong headwinds, increase the path cost of this section. When avoiding busy airspace, set a high - cost obstacle area, so as to plan a safe and efficient flight route to meet the requirements of different task scenarios.

[0010] Preferably, the decision - making unit formulates operation decisions through artificial intelligence algorithms. The specific method is: Collect flight status data, task plan data, and environmental data, and normalize the data; Identify the core variables affecting the decision through principal component analysis technology. The core variables include, but are not limited to: the sensitivity of speed to heading deviation, task priority, and the dynamic association of the flight path, and construct a multi - dimensional feature vector containing time - series data and static parameters; The model network architecture design uses a multi - layer feed - forward neural network. The input layer corresponds to the feature dimension, the number of neurons in the hidden layer is determined by trial - and - error method, and the output layer is mapped to the decision instruction; Define a composite loss function to comprehensively consider flight safety and mission efficiency, use an adaptive optimization algorithm combined with backpropagation to adjust the weights, and adopt a cross-validation strategy to train the model; Based on the trained algorithm model, input real-time data into the trained model, output operation instructions, and combine a rule engine to verify the compliance of the decisions. Continuously collect feedback on flight results through a reinforcement learning algorithm, and update the network weights to achieve online learning.

[0011] Preferably, when the decision-making unit uses the reinforcement learning algorithm for training, set the reward function: , where is the flight safety reward, and a positive reward is given when the aircraft maintains a speed below the speed limit, does not fly below the stall speed, and maintains a safe distance from other aircraft; is the mission completion reward, and rewards are given for arriving at the destination on time and accurately executing air traffic control instructions, etc.; is the operation efficiency reward, and rewards are given if the aircraft flies at an economical battery power consumption rate and selects the optimal flight altitude layer; Through continuous iterative training, the model makes optimal decisions in complex flight scenarios, improving the rationality and accuracy of simulating aircraft behavior.

[0012] Preferably, the flight simulation module simulating the physical motion of the aircraft specifically includes: S1: Model decomposition: Decompose the physical motion of the eVTOL aircraft into a rigid body dynamics, electric propulsion system power / torque calculation, and atmospheric environment simulation subsystems, and use Modelica for non-causal modeling; S2: Differential equation solving: Parse the differential equation solving logic for multi-motor coordinated control through an XML file, and establish a six-degree-of-freedom nonlinear equation set based on the aircraft mass, inertia, and battery-motor response; S3: Environmental impact calculation: Combine the urban boundary layer meteorological model to dynamically calculate the impact of low-altitude turbulence and the heat island effect on lift; S4: Real-time calculation and interaction: Design a real-time solver engine, interact with the flight control module through a shared memory bus, and integrate the building complex wake flow field and urban canyon wind shear algorithms to correct the aerodynamic parameters; S5: Multi-modal switching: Activate the rotor power coupling model during the vertical takeoff and landing phase, switch to the ground effect compensation algorithm during landing, and determine the stability through the motor torque and fuselage swing angle; S6: Power redundancy and safety: Set a power redundancy threshold, and use a model predictive controller to adjust the rotor speed to maintain attitude stability; S7: Data comparison and parameter optimization: Compare flight test and simulation data, switch the dynamic parameters of different eVTOL configurations through a modular configuration library, optimize the power distribution strategy using a generative adversarial network, and iteratively update the turbulence compensation coefficient.

[0013] Preferably, in the flight simulation module, when calculating the motion of the aircraft under various forces and moments, for the lift force, it is calculated through the expression of the relationship between the aircraft lift coefficient and the speed; The values under different states are obtained by fitting the wind tunnel test data and flight test data; Drag: , where the drag coefficient includes the parasite drag coefficient and the induced drag coefficient, which are respectively related to the aircraft shape and wing aspect ratio factors.

[0014] Preferably, during the communication process between the communication module and the SATCE communication unit, based on the ADS-B data transmission protocol, it sends the aircraft position information to the air traffic control system at a frequency of 1 Hz, with a position accuracy within 10 meters, a speed accuracy of 0.5 knots, and an altitude accuracy of 50 feet. At the same time, it receives air traffic control instructions, and the instruction transmission delay is less than 0.5 seconds.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The low-altitude pilot AI model training application system proposed by the present invention provides a highly simulated training environment for pilots by integrating the aircraft type database, flight operation knowledge base, real-time meteorological data, and airspace and airport databases. This can not only effectively improve the pilots' response capabilities under complex meteorological and special airspace conditions, but also significantly reduce the training cost and shorten the training cycle. Through the built-in intelligent decision-making module, it can monitor and evaluate the flight status in real time, intelligently plan the flight route and make decisions according to the environment and mission requirements, which greatly improves the authenticity and challenge of flight simulation. Through continuous iterative training with the reinforcement learning algorithm, the AI model can make optimal decisions in complex flight scenarios, thereby improving the rationality and accuracy of the simulated aircraft behavior, and can evaluate the cognitive load and stress response in real time according to the biometric characteristics of the trainees, so as to develop personalized training plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a module framework diagram of the low-altitude pilot AI model training application system.

[0017] Figure 2 It is an overall architecture diagram of the low-altitude pilot AI model training application system. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments in the following description are only examples, and those skilled in the art can think of other obvious variants.

[0019] Referring to Figure 1 - Figure 2 As shown, the low-altitude pilot AI model training application system includes: a data module, an intelligent decision-making module, a flight simulation module, a communication module, and an interface module; Data module The aircraft type database stores detailed information of various eVTOL aircraft, such as models, performance parameters, and handling characteristics, etc. By classifying and storing different aircraft, it provides basic data support for subsequent flight simulation and decision-making.

[0020] The flight operation knowledge base covers rich flight operation procedures and skills, including operation steps for key links such as takeoff and landing. In the takeoff operation, the initial acceleration stage on the runway is set according to the aircraft type, and the acceleration is calculated based on factors such as engine thrust, aircraft drag, and mass. When the speed reaches the takeoff decision speed, it is judged whether the takeoff condition is met according to the relationship between the aircraft lift coefficient and speed, and then the wheel-raising operation is performed. And throughout the process, communication with air traffic control needs to be maintained at a specified frequency to report the aircraft status. For the landing operation, the glide path angle is determined according to the airport runway conditions and meteorological information, and the instrument landing system is used to guide the aircraft, accurately controlling the sink rate at the moment of landing to ensure the aircraft lands smoothly, providing comprehensive and accurate operation references for pilots.

[0021] The meteorological data interface connects to an external meteorological source to obtain meteorological information. Among them, the wind speed affects the flight trajectory of the aircraft, and the ground speed is calculated through a specific formula to reflect this influence; the air temperature affects the air density and then changes the aircraft performance. The system will adjust the aircraft engine thrust, lift parameters, etc. accordingly at different air temperatures according to the ideal gas state equation, so that the low-altitude pilot AI model can adjust flight operation decisions in real time according to meteorological factor changes.

[0022] The airspace and airport database stores airspace and airport-related data. In terms of airspace structure information, the control area division is based on the standards of the International Civil Aviation Organization. Above a specific altitude layer is the high-altitude control area, and below is the low-altitude control area. Different control areas have different communication frequencies and control rules, and the flight altitude layer allocation rules follow the standard pressure altitude layer setting, which is divided at intervals of 300 meters. Aircraft need to fly at the specified altitude layer when planning flight routes to avoid conflicts. Among the airport data, the runway length and direction determine the aircraft takeoff and landing performance requirements and operation methods.

[0023] Intelligent decision-making module The flight status evaluation unit obtains the real-time status data of the simulated aircraft based on the flight simulation module and extracts the standard status data corresponding to the flight phase and mission from the flight operation knowledge base. First, the collected data is preprocessed, and after removing noise, outliers, and missing values, it is standardized. Then, the preprocessed real-time status data is compared with the standard status data parameter by parameter, the difference between each status parameter and the standard value is calculated, and the overall flight status of the aircraft is evaluated using a weighted comprehensive evaluation method. For example, when monitoring the aircraft's power, the remaining power is calculated based on the power consumption rate of the aircraft engine battery and the flight time. When the power is below the safety threshold, an alarm is issued in a timely manner and the nearest suitable airport is planned for an emergency landing; when evaluating the aircraft's attitude stability, by measuring the change rates of the aircraft's pitch angle, roll angle, and yaw angle, if it exceeds the preset stable range, combined with the aircraft dynamics model, it is judged whether there is a risk of out-of-control and corresponding adjustment measures are taken. Finally, a detailed flight status report is generated and fed back to the intelligent decision-making module and the interface module.

[0024] When planning the optimal flight route, the mission planning unit uses the Dijkstra algorithm to find the shortest path from the starting point to the destination. The navigation points in the airspace are regarded as the nodes of the graph, and the distance between the nodes is the flight cost. The path cost is weighted by comprehensively considering meteorological conditions, airspace restrictions, and the dynamic factors of other aircraft. For example, the cost of this section of the path is increased in a strong headwind area, and a high-cost obstacle area is set when avoiding busy airspace, so as to plan a safe and efficient flight route to meet the requirements of different mission scenarios.

[0025] The decision-making unit first collects flight status data, mission plan data, and environmental data and performs normalization processing. Then, through principal component analysis technology, the core variables affecting decision-making are identified, and a multi-dimensional feature vector containing time series data and static parameters is constructed. The model network architecture uses a multi-layer feedforward neural network, and the number of neurons in the hidden layer is determined by the trial-and-error method. The output layer is mapped to decision instructions. A composite loss function is defined to comprehensively consider flight safety and mission efficiency. The adaptive optimization algorithm is used in combination with backpropagation to adjust the weights, and a cross-validation strategy is used to train the model. After training is completed, real-time data is input into the model to output operation instructions, and the compliance of the decision is verified in combination with the rule engine. In addition, the reinforcement learning algorithm is used to continuously collect flight result feedback and update the network weights to achieve online learning. When training using the reinforcement learning algorithm, a reward function including flight safety rewards, mission completion rewards, and operation efficiency rewards is set. Through continuous iterative training, the model makes optimal decisions in complex flight scenarios, improving the rationality and accuracy of the behavior of the simulated aircraft.

[0026] Flight simulation module First, perform model decomposition. Divide the physical motion of the eVTOL aircraft into a rigid body dynamics subsystem, an electric propulsion system dynamics / torque calculation subsystem, and an atmospheric environment simulation subsystem, and use Modelica for non-causal modeling. Then, parse the differential equation solving logic of multi-motor cooperative control through an XML file, and establish a six-degree-of-freedom nonlinear equation set based on the aircraft mass, inertia, and battery-motor response. Next, combine with the urban boundary layer meteorological model to dynamically calculate the effects of low-altitude turbulence and the heat island effect on lift. Design a real-time calculation engine to interact with the flight control module through a shared memory bus, and integrate the building complex flow field and urban canyon wind shear algorithms to correct the aerodynamic parameters. Activate the rotor power coupling model during the vertical takeoff and landing phase, switch to the ground effect compensation algorithm during landing, and determine the stability by the motor torque and the fuselage swing angle. Set the power redundancy threshold, and use the model predictive controller to adjust the rotor speed to maintain attitude stability. Finally, compare the flight test and simulation data, realize the switching of the dynamic parameters of different eVTOL configurations through the modular configuration library, use the generative adversarial network to optimize the power distribution strategy, and iteratively update the turbulence compensation coefficient to effectively simulate the physical motion of the aircraft.

[0027] Communication module During the communication process between the communication module and the SATCE communication unit, based on the ADS-B data transmission protocol, it sends the aircraft position information to the air traffic control system at a frequency of 1 Hz. The position accuracy is within 10 meters, the speed accuracy is 0.5 knots, and the altitude accuracy is 50 feet. At the same time, it receives air traffic control instructions, and the instruction transmission delay is less than 0.5 seconds. This can ensure that the simulated aircraft can respond to air traffic control instructions in a timely manner, communicate and coordinate with other aircraft and the air traffic control system, comply with the actual air traffic control rules and procedures, provide a real airspace traffic scenario for the pilot, and effectively improve the communication and coordination capabilities of the pilot in actual flight.

[0028] Interface module The interface module is mainly used to provide interfaces for integration with flight simulation training equipment and instructor configuration, facilitating the connection of the system with external devices and the instructor's setting and management of the training process, so that the entire training system can better adapt to different training needs and scenarios.

[0029] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection required by the present invention is defined by the appended claims and their equivalents.

Claims

1. Low-altitude pilot AI model training application system, characterized by: include: Data module, intelligent decision module, flight simulation module, communication module and interface module; The data module includes: aircraft type database, flight operation knowledge base, meteorological data interface and airspace and airport database; the aircraft type database is used to store detailed information of various types of aircraft eVTOL, including: model, performance parameters and handling characteristics; the flight operation knowledge base is used to store flight operation procedures and skills; the meteorological data interface is used to connect to external meteorological sources to obtain meteorological information; the airspace and airport database is used to store airspace and airport related data; The intelligent decision-making module includes: a flight status evaluation unit, a mission planning unit and a decision-making unit; the flight status evaluation unit is used to monitor the status of the simulated aircraft and compare it with the standard; the mission planning unit is used to formulate a mission plan based on the knowledge base and the environment; the decision-making unit is used to make operational decisions through artificial intelligence algorithms; The flight simulation module is used to simulate the physical movement of the aircraft; The communication module is used to realize communication with the simulated control environment and the pilot of the aircraft; The interface module is used to provide integration with flight simulation training equipment and instructor configuration interface.

2. The low-altitude pilot AI model training application system according to claim 1 is characterized in that: In the data module, the maximum speed calculation formula for aircraft performance parameters in the aircraft type database is: ， In the formula, is the maximum speed of the aircraft, is the maximum thrust of the aircraft, is the air density, is the wing area; The cruising speed is determined based on the aircraft engine efficiency curve and flight altitude factors, and is the speed range that meets the engine battery power consumption rate at a preset altitude; The control characteristics are determined by the joystick sensitivity through the force feedback coefficient To measure when the pilot applies force When the control surface deflection angle , different aircraft models The values ​​vary, depending on the aircraft's mechanical structure and hydraulic / electrical drive system design; In the flight operation knowledge base, the takeoff operation procedure includes an initial acceleration phase on the runway set according to the aircraft type. The acceleration calculation formula is: ， In the formula, is the aircraft acceleration, is the engine thrust, is the aircraft drag, is the mass of the aircraft; When the speed reaches the takeoff decision speed Then, according to the aircraft lift coefficient The relationship between the aircraft lift coefficient and the speed determines whether the take-off conditions are met, and then the wheel rotation operation is performed. The expression related to speed is: ， In the formula, For lift, is the airspeed, is the lift coefficient; And during the whole process, the aircraft must maintain communication with air traffic control at the specified frequency to report the aircraft status; Landing procedures involve determining the glide path angle based on airport runway conditions and weather information. , guide the aircraft through the instrument landing system, and accurately control the sinking rate at the moment of landing ,satisfy: ， In the formula, is the sinking rate at the moment of landing, is the aircraft weight, The glide path angle ensures a smooth landing of the aircraft and provides a comprehensive and accurate operational reference for the pilot; In the meteorological information obtained by the meteorological data interface, the impact of wind speed on the aircraft flight trajectory is calculated using the aircraft ground speed calculation formula: ， in is the ground speed, is the airspeed, is the wind speed, is the angle between wind direction and aircraft heading; The air temperature affects the air density, which in turn changes the aircraft performance. According to the ideal gas state equation: ， in is the air pressure, is the gas constant, The thrust and lift parameters of the aircraft engine need to be adjusted accordingly at different temperatures. The low-altitude pilot AI model adjusts flight operation decisions in real time based on changes in meteorological factors. In the airspace and airport database, for airspace structure information, the control area division is based on ICAO standards; Above a certain altitude is the high-altitude control zone, and below is the low-altitude control zone. Different control zones have different communication frequencies and control rules; The flight level allocation rules follow the standard pressure level settings and are divided into intervals of 300 meters; In terms of airport data, the runway length and direction determine the aircraft take-off and landing performance requirements and operating methods.

3. The low-altitude pilot AI model training application system according to claim 1 is characterized in that: The method steps of the flight status evaluation unit monitoring the simulated aircraft status and comparing it with the standard are as follows: Based on the flight simulation module, the real-time status data of the simulated aircraft is obtained, and the standard status data corresponding to the flight phase and task is extracted from the flight operation knowledge base; Based on the collected data, preprocess the data, the preprocessing includes: removing noise, outliers and missing values, and standardizing the cleaned data; The pre-processed real-time status data is compared with the standard status data parameter by parameter. For each status parameter, the difference between it and the standard value is calculated. Based on the comparison results of each parameter, the overall flight status of the aircraft is evaluated using a weighted comprehensive evaluation method. A detailed flight status report is generated based on the evaluation results, wherein the flight status report includes: comparison results of various status parameters and comprehensive evaluation indicators, and the flight status report is fed back to the intelligent decision-making module and the interface module.

4. The low-altitude pilot AI model training application system according to claim 3 is characterized in that: The flight status evaluation unit of the intelligent decision module monitors the aircraft power consumption rate according to the aircraft engine battery power consumption rate. and flight time , calculate the remaining power: ， in, is the remaining power, is the initial charge; When the battery power is lower than the safety threshold, an alarm will be issued in time and the nearest suitable airport will be planned for alternate landing; When evaluating the aircraft's attitude stability, the rate of change of the aircraft's pitch, roll and yaw angles is measured. If the rate of change exceeds the preset stability range, the aircraft dynamics model is used to determine whether there is a risk of loss of control and take corresponding adjustment measures.

5. The low-altitude pilot AI model training application system according to claim 1 is characterized in that: The mission planning unit of the intelligent decision-making module uses the Dijkstra algorithm to find the shortest path from the starting point to the destination when planning the optimal route; The navigation points in the airspace are regarded as nodes of the graph, and the distance between nodes is the flight cost. The path cost is weighted by comprehensively considering the meteorological conditions, airspace restrictions and other aircraft dynamic factors; In areas with strong headwinds, the cost of this path is increased. When avoiding busy airspace, high-cost obstacle areas are set up to plan safe and efficient flight routes to meet the needs of different mission scenarios.

6. The low-altitude pilot AI model training application system according to claim 1 is characterized in that: The specific method by which the decision-making unit makes operational decisions through artificial intelligence algorithms is as follows: Collect flight status data, mission planning data and environmental data, and normalize the data; Identify the core variables that affect decision-making through principal component analysis technology, including but not limited to: sensitivity of speed to heading deviation, task priority and dynamic association of flight path, and construct a multidimensional feature vector containing time series data and static parameters; The model network architecture design adopts a multi-layer feedforward neural network, where the input layer corresponds to the feature dimension, the hidden layer determines the number of neurons by trial and error, and the output layer is mapped to the decision instructions; Define a composite loss function to integrate flight safety and mission efficiency, use an adaptive optimization algorithm combined with back propagation to adjust weights, and adopt a cross-validation strategy to train the model; Based on the trained algorithm model, real-time data is input into the trained model, operation instructions are output, and the decision compliance is verified in combination with the rule engine. Flight result feedback is continuously collected through the reinforcement learning algorithm, and the network weights are updated to achieve online learning.

7. The low-altitude pilot AI model training application system according to claim 6 is characterized in that: When the decision-making unit is trained using a reinforcement learning algorithm, the reward function is set: ， in As a reward for flight safety, a positive reward is given when the aircraft does not exceed the speed limit, does not fall below the stalling speed, and maintains a safe distance from other aircraft; Rewards are given for mission completion, such as arriving at the destination on time and accurately following air traffic control instructions; To reward operational efficiency, if the aircraft flies at an economical battery consumption rate and selects the optimal flight altitude level, rewards will be given; Through continuous iterative training, the model can make optimal decisions in complex flight scenarios and improve the rationality and accuracy of simulated aircraft behavior.

8. The low-altitude pilot AI model training application system according to claim 1 is characterized in that: The flight simulation module simulates the physical movement of the aircraft, including: S1: Model decomposition: The physical motion of the eVTOL aircraft is divided into rigid body dynamics, electric propulsion system power / torque calculation and atmospheric environment simulation subsystems, and Modelica is used for non-causal modeling; S2: Differential equation solving: The differential equation solving logic of multi-motor coordinated control is parsed through XML files, and a six-degree-of-freedom nonlinear equation group is established based on the aircraft mass, inertia, and battery-motor response; S3: Environmental impact calculation: Combined with the urban boundary layer meteorological model, dynamically solve the impact of low-altitude turbulence and heat island effect on lift; S4: Real-time calculation and interaction: Design a real-time calculation engine, exchange data with the flight control module through a shared memory bus, and integrate the flow field around buildings and the urban canyon wind shear algorithm to correct aerodynamic parameters; S5: Multi-mode switching: Activate the rotor power coupling model during vertical takeoff and landing, switch the ground effect compensation algorithm during landing, and determine stability through motor torque and fuselage swing angle; S6: Power redundancy and safety: Set the power redundancy threshold and use the model predictive controller to adjust the rotor speed to maintain attitude stability; S7: Data comparison and parameter optimization: Compare the test flight data with the simulation data, realize the switching of dynamic parameters of different eVTOL configurations through the modular configuration library, optimize the power distribution strategy using the adversarial generative network, and iteratively update the turbulence compensation coefficient.

9. The low-altitude pilot AI model training application system according to claim 1 is characterized in that: In the flight simulation module, the aircraft dynamics model calculates the movement of the aircraft under various forces and moments. For lift, the aircraft lift coefficient is used Calculation of expressions related to speed; The wind tunnel test data and flight test data were fitted to obtain the Value; Resistance: ， Among them, the resistance coefficient It includes the parasite drag coefficient and the induced drag coefficient, which are related to the aircraft shape and the wing aspect ratio respectively.

10. The low-altitude pilot AI model training application system according to claim 1 is characterized in that: During the communication process, the communication module and the SATCE communication unit use the ADS-B data transmission protocol to send aircraft position information to the control system at a frequency of 1Hz. The position accuracy is within 10 meters, the speed accuracy is 0.5 knots, and the altitude accuracy is 50 feet. At the same time, they receive control instructions, and the instruction transmission delay is less than 0.5 seconds.

Images