Flight state training method, device, equipment and medium

Abstract

The invention discloses a flight state training method and device, equipment and a medium. The method comprises the following steps: acquiring flight data of a target flight device at a plurality of moments in a plurality of takeoff processes, and driving data of a driving target driving the target flight device at a plurality of corresponding moments; obtaining a plurality of state data groups; inputting the state data set into the flight monitoring neural network model to be trained to obtain a plurality of predicted driving results; adjusting neural network parameters of the flight monitoring neural network model to be trained; and obtaining a target training flight monitoring neural network model. The invention belongs to the field of flight monitoring. According to the method, the to-be-trained flight monitoring neural network model is trained by using the data under a plurality of conditions, so that driving guidance under various take-off conditions can be provided for the to-be-trained pilot, the training efficiency is improved, and the labor cost is greatly reduced.

Description

Technical Field

[0001] The present invention relates to the field of flight monitoring, and in particular to a flight status training method, device, equipment and medium. Background Art

[0002] An airplane is a vehicle that can fly through the air by generating lift through wings and other aerodynamic surfaces. An airplane is a type of aircraft that relies on the lift generated by fixed wings (wings) to support its weight and glide or fly in the air. The design of an airplane usually includes: fuselage, wings, tail, engine, landing gear, and control system. Aircraft can be divided into many types according to their use, including commercial airliners, military fighter jets, private small aircraft, cargo planes, etc.

[0003] Flight instruction for new drivers (pilots) usually includes: theoretical learning, pre-flight preparation, flight operation, emergency handling, and recording and feedback. For new drivers, they mainly rely on the guidance of experienced instructors and conduct a lot of practical flight training. For training in the take-off process, its efficiency is low and it is difficult to promote and apply it in an all-round way. Therefore, it is urgent to propose a training method for "new" drivers. Summary of the invention

[0004] The embodiments of the present invention solve the technical problem of low efficiency of take-off training process in the prior art by providing a flight status training method, device, equipment and medium, and achieve the technical effect of improving training efficiency.

[0005] In a first aspect, the present invention provides a flight status training method, the method comprising: Acquire flight data of the target flying device at several moments during several takeoffs, and driving data of a driving target driving the target flying device at corresponding several moments, wherein the driving data at each moment includes a driving posture; A state data group is formed according to the flight data and a plurality of driving data corresponding to the flight data, and a plurality of state data groups are obtained in total; Inputting a state data group into the flight monitoring neural network model to be trained to obtain a number of predicted driving results; According to a number of predicted driving results, adjusting the neural network parameters of the flight monitoring neural network model to be trained; When the preset training requirements are met, the latest neural network parameters are saved, and the flight monitoring neural network model to be trained corresponding to the latest neural network parameters is used as the target training flight monitoring neural network model.

[0006] Furthermore, a state data group is formed according to the flight data and a plurality of driving data corresponding to the flight data, and a plurality of state data groups are obtained, including: Determine target flight data at target time during target takeoff; Determine several target driving data at target times during several takeoffs, A target state data group is formed based on the target flight data and a number of target driving data.

[0007] Further, according to a number of predicted driving results, the neural network parameters of the flight monitoring neural network model to be trained are adjusted, including: Determine a predicted driving result corresponding to each state data group, the predicted driving result including a correct match and an incorrect match; When the predicted driving result is a wrong match, the neural network parameters of the flight monitoring neural network model to be trained are adjusted.

[0008] Furthermore, when the preset training requirements are met, the latest neural network parameters are saved, including: When the preset maximum number of iterations is reached, the latest neural network parameters are saved; or, When the preset minimum number of iterations is reached and the accuracy of the predicted driving results is greater than the preset threshold, the latest neural network parameters are saved.

[0009] Furthermore, the loss function of the flight monitoring neural network model to be trained includes: , in, is the total number of status data groups, is the preset stage weight, For the Accurate driving data, For the Predicted driving data, is the total average loss, is the regularization coefficient, Monitor all learnable parameters in the neural network model for the flight to be trained.

[0010] Further, including: The flight data includes flight device data and flight environment data. The flight device data includes yaw angle, pitch angle, roll angle, longitude and latitude, relative pressure altitude, engine speed, airspeed and vertical acceleration. The flight environment data includes outdoor temperature, outdoor air pressure and weather conditions.

[0011] Furthermore, neural network parameters include: Weights, biases, learning rate, batch size, number of layers, and number of nodes in each layer.

[0012] In a second aspect, the present invention provides a flight status training device, the device comprising: An acquisition module, used to acquire flight data of the target flying device at several moments during several take-offs, and driving data of a driving target driving the target flying device at corresponding several moments, wherein the driving data at each moment includes a driving posture; A forming module, used to form a state data group according to the flight data and a plurality of driving data corresponding to the flight data, so as to obtain a plurality of state data groups in total; An input module, used to input a state data group into the flight monitoring neural network model to be trained to obtain a number of predicted driving results; An adjustment module, used for adjusting the neural network parameters of the flight monitoring neural network model to be trained according to a number of predicted driving results; The output module is used to save the latest neural network parameters when the preset training requirements are met, and use the flight monitoring neural network model to be trained corresponding to the latest neural network parameters as the target training flight monitoring neural network model.

[0013] In a third aspect, the present invention provides an electronic device, comprising: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to execute to implement a flight status training method as provided in the first aspect.

[0014] In a fourth aspect, the present invention provides a non-temporary computer-readable storage medium, characterized in that when the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute a flight status training method as provided in the first aspect.

[0015] One or more technical solutions provided in the embodiments of the present invention have at least the following technical effects or advantages: The present invention provides a flight state training method, which includes: obtaining flight data of a target flight device at several moments during several take-offs, and driving data of a driving target driving the target flight device at several corresponding moments, wherein the driving data at each moment includes a driving posture; forming a state data group according to the flight data and several driving data corresponding to the flight data, and obtaining several state data groups in total; inputting the state data group into a flight monitoring neural network model to be trained to obtain several predicted driving results; adjusting the neural network parameters of the flight monitoring neural network model to be trained according to the several predicted driving results; when the preset training requirements are met, saving the latest neural network parameters, and using the flight monitoring neural network model to be trained corresponding to the latest neural network parameters as the target training flight monitoring neural network model. The present invention can accurately train the flight monitoring neural network model to be trained through the data state group, and because the present invention compares the driving data at the same moment, the data comparison range is greatly reduced, and the training efficiency is improved. When the method of "saving the latest neural network parameters when the preset minimum number of iterations is reached and the accuracy of the predicted driving result is greater than the preset threshold value" is adopted, the model training effect can be guaranteed while saving computing resources.

[0016] In the loss function provided by the present invention, by adding a regularization term to the loss function, the high complexity of the model can be penalized, thereby avoiding overfitting. By minimizing the loss function, the neural network parameters of the flight monitoring neural network model to be trained can be adjusted so that the model can better fit the training data.

[0017] Since the present invention collects data under several take-off conditions and uses the data under several conditions to train the flight monitoring neural network model to be trained, it can provide flying guidance for various take-off conditions for the pilots to be trained, improve the training efficiency, greatly reduce the cost of human guidance, and standardize the training. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 A schematic diagram of a flow chart of a flight status training method provided by the present invention; Figure 2 A schematic structural diagram of a flight status training device provided by the present invention; Figure 3The present invention provides a schematic structural diagram of an electronic device. DETAILED DESCRIPTION

[0020] The embodiment of the present invention solves the technical problem of low efficiency of take-off training process in the prior art by providing a flight status training method.

[0021] The technical solution of the embodiment of the present invention is to solve the above technical problems, and the overall idea is as follows: A flight status training method comprises: obtaining flight data of a target flight device at several moments during several take-offs, and driving data of a driving target driving the target flight device at several corresponding moments, wherein the driving data at each moment comprises a driving posture; forming a state data group according to the flight data and several driving data corresponding to the flight data, and obtaining several state data groups in total; inputting the state data group into a flight monitoring neural network model to be trained, and obtaining several predicted driving results; adjusting the neural network parameters of the flight monitoring neural network model to be trained according to the several predicted driving results; and when the preset training requirements are met, saving the latest neural network parameters, and using the flight monitoring neural network model to be trained corresponding to the latest neural network parameters as the target training flight monitoring neural network model.

[0022] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0023] First of all, the term "and / or" in this article is only a description of the association relationship of the associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0024] The present invention provides Figure 1 A flight status training method shown includes steps S11-S15: Regarding step S11, the flight data of the target flying device at several moments during several take-offs, and the driving data of the driving target driving the target flying device at corresponding several moments are obtained, and the driving data at each moment includes the driving posture.

[0025] The target flight device refers to a flight device that needs to be driven by a pilot (usually refers to an aircraft). The target flight device may refer to a certain type of aircraft. For example, when flying the same type of aircraft, the same driving action of the pilot has the same effect on the same type of aircraft.

[0026] The flight data of the target flying device at several moments during several takeoffs can be obtained. It is understandable that in order to ensure the training effect of the following neural network model, it is necessary to collect flight data and driving data at different moments under different takeoff conditions for multiple times.

[0027] The take-off process of the flying device refers to the process from a stationary state or a ground gliding state to a successful climb to a predetermined flight trajectory (i.e., entering a horizontal flight state). The duration of the take-off process is usually within 10 minutes. Therefore, in the present invention, the take-off duration of the take-off process of the flying device is set to 10 minutes (in other words, even if during a certain take-off process, the flying device has entered a horizontal flight state from a climbing state within 10 minutes, it is still considered to be in the take-off process before 10 minutes).

[0028] The driving data includes driving posture, wherein the driving posture includes the driving action and body posture of the driver at that moment.

[0029] It is understandable that the purpose of the present invention is to standardize the actions of "new" drivers and guide their driving actions. Preferably, in the data collection stage of step S11, only the data of experienced "old" drivers can be collected, and the data of experienced "old" drivers can be regarded as standard data.

[0030] Flight data includes flight device data and flight environment data. Flight device data includes yaw angle, pitch angle, roll angle, longitude and latitude, relative pressure altitude, engine speed, airspeed and vertical acceleration. Flight environment data includes outdoor temperature, outdoor air pressure and weather conditions. Flight environment data refers to the outdoor temperature, air pressure and weather conditions of the flight device during flight.

[0031] Regarding step S12, a state data group is formed according to the flight data and a plurality of driving data corresponding to the flight data, and a total of a plurality of state data groups are obtained.

[0032] Specifically, it includes: Determine target flight data at a target time during a target takeoff; determine a number of target driving data at target times during a number of takeoffs, and form a target state data group based on the target flight data and the number of target driving data.

[0033] For example, if a total of 100 takeoffs are collected, the flight data during the first takeoff is used as the target flight data, and the collection frequency can be set to 0.1s for collection. The total takeoff duration is 10 minutes. In other words, in the 100 takeoffs, each takeoff will correspond to 6,000 driving data and 6,000 flight data.

[0034] The flight data at the 10th second during the first takeoff is used as the target flight data, and the driving data at the 10th second during each takeoff is used as the target driving data. In other words, corresponding to 100 target driving data, the above 1 target flight data and 100 target driving data are collectively used as 1 target state data group.

[0035] Regarding step S13, a state data group is input into the flight monitoring neural network model to be trained to obtain a number of predicted driving results.

[0036] Specifically, the types of driving results include accurate identification and incorrect identification. Accurate identification means that the flight data at a certain moment in a takeoff process can be accurately matched with the driving data at that moment in the takeoff process.

[0037] For example, the flight data at the 10th second during the first takeoff process can be accurately matched with the driving data at the 10th second during the first takeoff process from 100 driving data.

[0038] The present invention can accurately train the flight monitoring neural network model to be trained through the data state group, and because the present invention compares the driving data at the same time, the data comparison range is greatly narrowed and the training efficiency is improved.

[0039] The loss function of the flight monitoring neural network model to be trained is used to measure the difference between the model prediction output and the actual target, including: , in, is the total number of status data groups, is the preset stage weight, For the Accurate driving data, For the Predicted driving data, is the total average loss, is the regularization coefficient, Monitor all learnable parameters in the neural network model for the flight to be trained.

[0040] The loss function can be expressed in various forms. The above is only one embodiment provided by the present invention, which is used to predict the difference between the output and the actual target. In the loss function provided by the present invention, by adding a regularization term to the loss function, the high complexity of the model can be penalized to avoid overfitting. By minimizing the loss function, the neural network parameters of the flight monitoring neural network model to be trained can be adjusted so that the model can better fit the training data.

[0041] Regarding step S14, according to a number of predicted driving results, the neural network parameters of the flight monitoring neural network model to be trained are adjusted, wherein the neural network parameters include: weight, bias, learning rate, batch size, number of layers and number of nodes in each layer.

[0042] Specifically, it includes: determining the predicted driving result corresponding to each state data group, the predicted driving result includes accurate matching and wrong matching; when the predicted driving result is a wrong matching, adjusting the neural network parameters of the flight monitoring neural network model to be trained.

[0043] For each state data set (including flight data and driving data), the flight monitoring neural network model to be trained will generate a predicted driving result. The predicted driving result includes accurate matching and incorrect matching.

[0044] If the predicted driving data of the flight monitoring neural network model to be trained matches the actual driving data (ie, the prediction is accurate), then it means that the flight monitoring neural network model to be trained can achieve accurate matching in the neural network parameters.

[0045] If the predicted driving data of the flight monitoring neural network model to be trained does not match the actual driving data (i.e., the prediction is wrong), it means that the neural network parameters of the flight monitoring neural network model to be trained cannot be accurately matched, and the neural network parameters of the flight monitoring neural network model to be trained need to be adjusted.

[0046] If the model's predictions do not match the actual data (i.e., a mismatch occurs), the model's parameters (weights, biases, etc.) can be adjusted through the back-propagation algorithm. The goal of the adjustment is to reduce the difference between the predicted value and the actual value, that is, to reduce the value of the loss function.

[0047] Regarding step S15, when the preset training requirements are met, the latest neural network parameters are saved, and the flight monitoring neural network model to be trained corresponding to the latest neural network parameters is used as the target training flight monitoring neural network model.

[0048] When the preset maximum number of iterations is reached, the latest neural network parameters are saved; or when the preset minimum number of iterations is reached and the accuracy of the predicted driving result is greater than a preset threshold, the latest neural network parameters are saved. The preset threshold can be determined according to actual conditions, for example, 99.99%, the preset maximum number of iterations can be determined according to actual conditions, for example, 30,000 times; and the preset minimum number of iterations can be determined according to actual conditions, for example, 10,000 times.

[0049] It should be noted that when the policy of "saving the latest neural network parameters when the preset minimum number of iterations is reached and the accuracy of the predicted driving results is greater than the preset threshold" is adopted, the model training effect can be guaranteed while saving computing resources. When the preset minimum number of iterations and the preset threshold are met at the same time, it means that the model has maintained a good prediction accuracy in a certain number of training times, and the model can be discontinued.

[0050] In summary, the present invention provides a flight state training method, which includes: obtaining flight data of a target flight device at several moments during several take-offs, and driving data of a driving target of the target flight device at several corresponding moments, wherein the driving data at each moment includes a driving posture; forming a state data group according to the flight data and several driving data corresponding to the flight data, and obtaining several state data groups in total; inputting the state data group into a flight monitoring neural network model to be trained to obtain several predicted driving results; adjusting the neural network parameters of the flight monitoring neural network model to be trained according to the several predicted driving results; when the preset training requirements are met, saving the latest neural network parameters, and using the flight monitoring neural network model to be trained corresponding to the latest neural network parameters as the target training flight monitoring neural network model. The present invention can accurately train the flight monitoring neural network model to be trained through the data state group, and because the present invention compares the driving data at the same moment, the data comparison range is greatly reduced, and the training efficiency is improved. When the method of "saving the latest neural network parameters when the preset minimum number of iterations is reached and the accuracy of the predicted driving result is greater than the preset threshold value" is adopted, the model training effect can be guaranteed while saving computing resources.

[0051] In the loss function provided by the present invention, by adding a regularization term to the loss function, the high complexity of the model can be penalized, thereby avoiding overfitting. By minimizing the loss function, the neural network parameters of the flight monitoring neural network model to be trained can be adjusted so that the model can better fit the training data.

[0052] Since the present invention collects data under several take-off conditions and uses the data under several conditions to train the flight monitoring neural network model to be trained, it can provide flying guidance for various take-off conditions for the pilots to be trained, improve training efficiency, greatly reduce labor costs, and standardize training.

[0053] Based on the same inventive concept, the present invention provides Figure 2 A flight status training device is shown, the device comprising: The acquisition module 21 is used to acquire the flight data of the target flying device at several moments during several take-offs, and the driving data of the driving target driving the target flying device at several corresponding moments, wherein the driving data at each moment includes the driving posture; A forming module 22 is used to form a state data group according to the flight data and a plurality of driving data corresponding to the flight data, and obtain a plurality of state data groups in total; An input module 23, used to input a state data group into the flight monitoring neural network model to be trained to obtain a number of predicted driving results; An adjustment module 24, used for adjusting the neural network parameters of the flight monitoring neural network model to be trained according to a number of predicted driving results; The output module 25 is used to save the latest neural network parameters when the preset training requirements are met, and use the flight monitoring neural network model to be trained corresponding to the latest neural network parameters as the target training flight monitoring neural network model.

[0054] Based on the same inventive concept, the present invention also provides Figure 3 An electronic device as shown includes: Processor 31; A memory 32 for storing instructions executable by the processor 31; The processor 31 is configured to execute to implement a flight status training method as provided above.

[0055] Based on the same inventive concept, the present invention also provides a non-temporary computer-readable storage medium. When the instructions in the storage medium are executed by the processor 31 of the electronic device, the electronic device can execute a flight status training device as provided above.

[0056] Since the electronic device introduced in this embodiment is an electronic device used to implement the information processing method in the embodiment of the present invention, based on the information processing method introduced in the embodiment of the present invention, a person skilled in the art can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of the present invention is not described in detail here. As long as the electronic device used by a person skilled in the art to implement the information processing method in the embodiment of the present invention, it belongs to the scope of protection of the present invention.

[0057] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0058] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0059] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0060] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0061] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0062] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A flight status training method, characterized in that: The method comprises: Acquire flight data of a target flying device at a plurality of times during a plurality of take-offs, and driving data of a driving target driving the target flying device at a plurality of corresponding times, wherein the driving data at each time includes a driving posture; A state data group is formed according to the flight data and a plurality of driving data corresponding to the flight data, and a plurality of state data groups are obtained in total; Inputting a state data group into the flight monitoring neural network model to be trained to obtain a number of predicted driving results; According to a number of predicted driving results, adjusting the neural network parameters of the flight monitoring neural network model to be trained; When the preset training requirements are met, the latest neural network parameters are saved, and the flight monitoring neural network model to be trained corresponding to the latest neural network parameters is used as the target training flight monitoring neural network model.

2. A flight status training method as claimed in claim 1, characterized in that: The flight data and a plurality of driving data corresponding to the flight data form a state data group, and a total of a plurality of state data groups are obtained, including: Determine target flight data at target time during target takeoff; Determine several target driving data at target times during several takeoffs, A target state data group is formed according to the target flight data and a plurality of target driving data.

3. A flight status training method as claimed in claim 1, characterized in that: The step of adjusting the neural network parameters of the flight monitoring neural network model to be trained according to the plurality of predicted driving results includes: Determine a predicted driving result corresponding to each state data group, wherein the predicted driving result includes an accurate match and an incorrect match; When the predicted driving result is a wrong match, the neural network parameters of the flight monitoring neural network model to be trained are adjusted.

4. A flight status training method as claimed in claim 1, characterized in that: When the preset training requirements are met, the latest neural network parameters are saved, including: When the preset maximum number of iterations is reached, the latest neural network parameters are saved; or, When the preset minimum number of iterations is reached and the accuracy of the predicted driving results is greater than the preset threshold, the latest neural network parameters are saved.

5. A flight status training method as claimed in claim 1, characterized in that: The loss function of the flight monitoring neural network model to be trained includes: ， in, is the total number of status data groups, is the preset stage weight, For the Accurate driving data, For the Predicted driving data, is the total average loss, is the regularization coefficient, Monitor all learnable parameters in the neural network model for the flight to be trained.

6. A flight status training method as claimed in claim 1, characterized in that: include: The flight data includes flight device data and flight environment data. The flight device data includes yaw angle, pitch angle, roll angle, longitude and latitude, relative pressure altitude, engine speed, airspeed and vertical acceleration. The flight environment data includes outdoor temperature, outdoor air pressure and weather conditions.

7. A flight status training method as claimed in claim 1, characterized in that: Neural network parameters, including: Weights, biases, learning rate, batch size, number of layers, and number of nodes in each layer.

8. A flight status training device, characterized in that: The device comprises: An acquisition module, used to acquire flight data of a target flying device at several moments during several take-offs, and driving data of a driving target driving the target flying device at corresponding several moments, wherein the driving data at each moment includes a driving posture; A forming module, used to form a state data group according to the flight data and a plurality of driving data corresponding to the flight data, so as to obtain a plurality of state data groups in total; An input module, used to input a state data group into the flight monitoring neural network model to be trained to obtain a number of predicted driving results; An adjustment module, used for adjusting the neural network parameters of the flight monitoring neural network model to be trained according to a number of predicted driving results; The output module is used to save the latest neural network parameters when the preset training requirements are met, and use the flight monitoring neural network model to be trained corresponding to the latest neural network parameters as the target training flight monitoring neural network model.

9. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; Wherein, the processor is configured to execute to implement a flight status training method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium, characterized in that: When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to implement a flight status training method as described in any one of claims 1 to 7.

Images