Marine crew adaptability period adjustment method and device based on large language model and storage medium

Abstract

The invention discloses a crew adaptability period adjustment method and device based on a large language model and a storage medium, and the method comprises the steps: collecting the physiological activity data, living environment data and navigation task data of a to-be-adjusted crew through a wearable device in a closed space of an underwater vehicle, a personal state adjusting scheme is generated through a large language model; obtaining hull motion data of the underwater vehicle, and generating a virtual reality balance training scheme in combination with the vestibular function data of the to-be-adjusted sailors; predicting a biological feedback change trend based on the personal state adjustment scheme and the virtual reality balance training scheme; and carrying out balance training and controlling the training intensity based on the biological feedback change trend and the virtual reality balance training scheme so as to realize adaptive cycle adjustment of the sailors. According to the method, the large language model and the virtual reality technology are introduced to regulate and control the balance training of the sailors, so that the sailors can quickly adapt to the navigation environment, the adaptive period of navigation is shortened, and the navigation working efficiency of the sailors is improved.

Description

Technical Field

[0001] The present application relates to the field of ship navigation technology, and in particular to a crew adaptive period adjustment method, device, and storage medium based on a large language model. Background Art

[0002] On modern ships, crew efficiency and adaptability are key factors in ensuring safe and efficient operations. However, when crew members first join a ship or have not sailed for an extended period, they often need time to adapt to the ship's dynamic motion (such as rolling and pitching) and the unique working environment. This adaptation period not only impacts crew efficiency but can also, in certain circumstances, affect navigation safety, particularly in adverse weather conditions.

[0003] Currently, traditional crew training methods rely primarily on theoretical knowledge and simple simulations, which struggle to fully simulate the complex environments and dynamic changes experienced during actual sailing. Consequently, crew members still require a significant amount of time to adapt to the actual sailing experience. Furthermore, traditional training methods lack personalized adjustment plans and are unable to tailor training to individual crew members' differences, further prolonging the crew's adaptation period. Summary of the Invention

[0004] The main purpose of this application is to provide a crew adaptation period adjustment method, device and storage medium based on a large language model, aiming to solve the technical problem that traditional crew training methods are difficult to fully simulate the complex environment and dynamic changes in actual navigation, resulting in crew members still needing a long time to adapt during actual navigation.

[0005] To achieve the above objectives, the present application proposes a crew adaptive period adjustment method based on a large language model, the crew adaptive period adjustment method based on a large language model comprising: In the confined space of the underwater vehicle, wearable devices are used to collect physiological activity data, living environment data, and navigation mission data of the crew members to be adjusted; Generate a personal state adjustment plan based on the physiological activity data, living environment data and navigation task data through a large language model; Acquiring hull motion data of the underwater vehicle, and generating a virtual reality balance training program based on the hull motion data and vestibular function data of the crew member to be adjusted; Predicting a biofeedback change trend of the crew member to be adjusted based on the personal state adjustment program and the virtual reality balance training program; Based on the biofeedback change trend and the virtual reality balance training program, balance training is performed on the crew member to be adjusted and the training intensity of the crew member to be adjusted is controlled to achieve adaptive periodic adjustment of the crew member.

[0006] In one embodiment, before collecting the physiological activity data, living environment data, and navigation mission data of the crew member to be adjusted through the wearable device in the confined space of the underwater vehicle, the method further includes: Conduct static vestibular function tests and dynamic vestibular function tests on crew members to obtain test parameters; Evaluate the vestibular function stability level of crew members under different sailing conditions according to the test parameters; Determining a vestibular function score for each crew member according to the crew member's vestibular function stability level under different sailing conditions; The crew members to be adjusted are determined according to the vestibular function scores of the crew members.

[0007] In one embodiment, the test data includes static test data and dynamic test data. The static vestibular function test and the dynamic vestibular function test are performed on the crew member to obtain test parameters, including: Apply different vestibular stimulations to the crew members through vestibular stimulation equipment and test the crew members' head position-dependent response, vestibular adaptability, and spatial orientation ability; Provide conflict situations between the visual and vestibular systems through virtual reality equipment and test crew members' adaptability and vestibular perception stability in the virtual environment; The dynamic balance test equipment is used to simulate the motion of underwater vehicles in different navigation states and to detect how the crew maintains balance under hull motions of different frequencies and amplitudes; Use a ship motion simulator to simulate different sailing conditions and collect gait data, body position data, and reaction data of crew members in different dynamic environments; taking the head position-dependent response, the vestibular adaptability, the spatial orientation ability, the crew member's adaptability in a virtual environment, and vestibular perception stability as static test data; The crew's balance-maintaining manner under ship motions of different frequencies and amplitudes, as well as the crew's gait data, body position data, and reaction data in different dynamic environments are used as dynamic test data.

[0008] In one embodiment, generating a personal state adjustment plan based on the physiological activity data, the living environment data, and the navigation task data through a large language model includes: Obtaining sleep data, exercise data, and nutritional data based on the physiological activity data; Obtaining noise data, temperature and humidity data, and air quality data based on the living environment data; Based on the sleep data, exercise data, nutrition data, noise data, temperature and humidity data, air quality data, and navigation mission data, the large language model is used to identify the psychological state, physical condition, and work stress level of the crew member to be adjusted; A personal state adjustment plan is generated according to the psychological state, physical condition and work stress level of the crew member to be adjusted, wherein the personal state adjustment plan at least includes a work and rest adjustment plan, an exercise plan, and a nutritional diet guidance plan.

[0009] In one embodiment, obtaining the underwater vehicle's hull motion data and generating a virtual reality balance training program based on the hull motion data and vestibular function data of the crew member to be trained includes: The underwater vehicle's sensors are used to obtain real-time ship motion data, including the ship's sway angle, pitch angle, roll angle, and acceleration data; determining the difficulty and training mode of balance training for the crew member according to the vestibular function score of the crew member to be adjusted and the hull motion data; Based on the balance training difficulty and training mode, a balance training scene matching the motion state of the underwater vehicle is generated by virtual reality technology, wherein the balance training scene includes a static balance training scene and a dynamic balance training scene; A virtual reality balance training program is generated according to the balance training scenario, wherein the virtual reality balance training program includes training time, training times, training content, and a training feedback mechanism.

[0010] In one embodiment, generating a balance training scene that matches the motion state of the underwater vehicle using virtual reality technology based on the balance training difficulty and training mode includes: Use 3D modeling technology to create a virtual reality scene of the underwater vehicle's internal environment, including the internal structure layout, equipment distribution, and lighting conditions of the underwater vehicle; Integrate the actual environmental elements of the underwater vehicle into the virtual reality scene to generate a static balance training scene; Setting balance training tasks of different difficulty levels based on the balance training difficulty and training mode; Generate corresponding dynamic balance training scenarios according to the balance training tasks of different difficulty levels.

[0011] In one embodiment, the predicting of the biofeedback change trend of the crew member to be adjusted based on the personal state adjustment program and the virtual reality balance training program includes: Constructing an initial biofeedback prediction model, wherein the initial biofeedback prediction model is composed of a superposition of several restricted Boltzmann machine networks; Obtain historical biofeedback data, historical personal state adjustment plans, and historical virtual reality balance training plans of the crew members to be adjusted; Constructing a training set based on the historical biofeedback data, historical personal state adjustment program, and historical virtual reality balance training program; Training the initial biofeedback prediction model using the training set to obtain a target biofeedback prediction model; The target biofeedback prediction model is used to predict biofeedback data of the crew member to be adjusted according to the personal state adjustment plan and the virtual reality balance training plan. The biofeedback data at least includes the change trends of heart rate, blood pressure, skin conductivity, respiratory rate and brain wave activity.

[0012] In one embodiment, the training of the initial biofeedback prediction model using the training set to obtain a target biofeedback prediction model includes: generating a corresponding biofeedback tag according to the historical biofeedback data, wherein the biofeedback tag includes a change trend of the historical biofeedback data; Vectorizing the historical personal state adjustment scheme and the historical virtual reality balance training scheme to obtain corresponding feature vectors; Using the corresponding feature vector as a first training sample, and combining the biofeedback label and the corresponding feature vector into a second training sample; training a first restricted Boltzmann machine network of the initial biofeedback prediction model using the first training sample to obtain a reconstruction error of the first restricted Boltzmann machine network; When the reconstruction error of the first restricted Boltzmann machine network is less than a preset error threshold, the output of the first restricted Boltzmann machine network is used as the input of the second restricted Boltzmann machine network, and the second restricted Boltzmann machine network is continued to be trained, and so on, until the training of the last layer of restricted Boltzmann machine network is completed, thereby obtaining an initially trained biofeedback prediction model, wherein the training process of each layer of restricted Boltzmann machine network includes forward propagation and backward adjustment, the forward propagation is used to calculate the output of the current layer of restricted Boltzmann machine network, and the backward adjustment is used to adjust the weight and bias of the current layer of restricted Boltzmann machine network according to the output error; The biofeedback prediction model trained at the initial training is fine-tuned using the second training sample to obtain a target biofeedback prediction model, wherein the fine-tuning training process includes verifying the biofeedback prediction model trained at the initial training using a cross-validation method and adjusting model parameters based on the verification results.

[0013] In addition, to achieve the above-mentioned purpose, the present application also proposes a crew adaptive period adjustment device based on a large language model, the crew adaptive period adjustment device based on a large language model comprising: A collection module, used to collect physiological activity data, living environment data, and navigation mission data of the crew to be adjusted through wearable devices in the confined space of the underwater vehicle; A generation module, configured to generate a personal state adjustment plan based on the physiological activity data, the living environment data, and the navigation task data through a large language model; an acquisition module, configured to acquire hull motion data of the underwater vehicle and generate a virtual reality balance training program based on the hull motion data and vestibular function data of the crew member to be adjusted; A prediction module, configured to predict a biofeedback change trend of the crew member to be adjusted based on the personal state adjustment program and the virtual reality balance training program; A training module is used to perform balance training on the crew member to be adjusted based on the biofeedback change trend and the virtual reality balance training program and control the training intensity of the crew member to be adjusted to achieve adaptive periodic adjustment of the crew member.

[0014] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by the processor, the steps of the crew adaptive period adjustment method based on the large language model as described above are implemented.

[0015] One or more technical solutions proposed in this application include collecting physiological activity data, living environment data, and navigation task data of a crew member to be adjusted within the confined space of an underwater vehicle through a wearable device; generating a personal state adjustment plan based on the physiological activity data, living environment data, and navigation task data using a large language model; obtaining the underwater vehicle's hull motion data, and generating a virtual reality balance training plan based on the hull motion data and the vestibular function data of the crew member to be adjusted; predicting the biofeedback change trend of the crew member to be adjusted based on the personal state adjustment plan and the virtual reality balance training plan; and performing balance training on the crew member to be adjusted based on the biofeedback change trend and the virtual reality balance training plan, and controlling the training intensity of the crew member to be adjusted to achieve adaptive cycle adjustment of the crew member. Through the above-mentioned method, by introducing a large language model and virtual reality technology, combining personal state adjustment with balance training, the crew member's balance training process is accurately regulated, enabling the crew member to quickly adapt to the navigation environment, thereby shortening the navigation adaptation cycle and improving the crew member's navigation work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0018] Figure 1 A flowchart of the first embodiment of the crew adaptive period adjustment method based on a large language model of the present application is provided; Figure 2 A flowchart of Example 2 of the crew adaptive period adjustment method based on a large language model of this application is provided; Figure 3 This is a schematic diagram of the module structure of the crew adaptive period adjustment device based on the large language model in an embodiment of the present application.

[0019] The purpose, features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0020] It should be understood that the specific embodiments described herein are merely used to explain the technical solutions of the present application and are not intended to limit the present application.

[0021] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0022] The main solution of the embodiment of the present application is: collecting physiological activity data, living environment data and navigation task data of the crew member to be adjusted through wearable devices in the confined space of the underwater vehicle; generating a personal state adjustment plan through a large language model based on the physiological activity data, living environment data and navigation task data; obtaining the hull motion data of the underwater vehicle, and generating a virtual reality balance training plan according to the hull motion data and the vestibular function data of the crew member to be adjusted; predicting the biofeedback change trend of the crew member to be adjusted based on the personal state adjustment plan and the virtual reality balance training plan; performing balance training on the crew member to be adjusted based on the biofeedback change trend and the virtual reality balance training plan and controlling the training intensity of the crew member to be adjusted, so as to achieve adaptive periodic adjustment of the crew member.

[0023] Currently, traditional crew training methods rely primarily on theoretical knowledge and simple simulations, which struggle to fully simulate the complex environments and dynamic changes experienced during actual sailing. Consequently, crew members still require a significant amount of time to adapt to the actual sailing experience. Furthermore, traditional training methods lack personalized adjustment plans and are unable to tailor training to individual crew members' differences, further prolonging the crew's adaptation period.

[0024] This application provides a solution that combines personal state adjustment and balance training by introducing a large language model and virtual reality technology to accurately regulate the crew's balance training process, enabling the crew to quickly adapt to the navigation environment, thereby shortening the adaptation period of navigation and improving the crew's navigation work efficiency.

[0025] It should be noted that the execution subject of this embodiment may be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, mobile phone, etc., or an electronic device capable of implementing the above functions, a crew adaptive periodic adjustment device based on a large language model, etc. The following describes this embodiment and the following embodiments using the crew adaptive periodic adjustment device based on a large language model as an example.

[0026] Based on this, the embodiment of the present application provides a crew adaptive cycle adjustment method based on a large language model, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the crew adaptive period adjustment method based on a large language model of this application.

[0027] In this embodiment, the crew adaptability period adjustment method based on the large language model includes steps S10 to S50: Step S10: In the confined space of the underwater vehicle, the physiological activity data, living environment data and navigation mission data of the crew member to be adjusted are collected through the wearable device.

[0028] It should be noted that the crew's adaptive periodic adjustment in this embodiment is applicable to confined spaces such as underwater vehicles. Underwater vehicles refer to submersibles with the ability to navigate and operate underwater, such as submarines and deep-sea submersibles. Their internal spaces are relatively closed, and due to the particularity of the underwater environment, crew members often experience various discomforts when they first enter, such as seasickness and anxiety. Therefore, adaptive periodic adjustment is required to enable the crew to quickly adapt to the confined environment of the underwater vehicle and improve navigation efficiency and safety.

[0029] Wearable devices are devices that can be worn on the human body and use sensors and other technical means to monitor the human body in real time and collect data, such as smart watches, smart bracelets, and smart glasses. In this embodiment, wearable devices are used to collect physiological activity data, living environment data, and navigation mission data of the crew members for subsequent personal status adjustment and balance training.

[0030] It is worth noting that when collecting the physiological activity data, living environment data and navigation mission data of the crew members to be adjusted, wearable devices can monitor the crew members' various physiological indicators in real time, such as sleep data, exercise data, nutritional data, etc., and record the crew members' living environment data, such as cabin temperature, humidity, noise level, etc., as well as the specific content of the navigation mission, such as navigation time, depth, target, etc.

[0031] Specifically, not all crew members experience seasickness and other discomfort. Therefore, when collecting data, it is necessary to assess individual crew members' conditions, such as age, physical condition, and vestibular function, to determine which crew members require adaptive cycle adjustments (i.e., those to be adjusted), as well as the focus and direction of adjustment. For example, for crew members with weak vestibular function, virtual reality balance training can be emphasized to improve their ability to adapt to the ship's motion.

[0032] In a feasible implementation, before step S10, steps A11 to A14 are included: Step A11: Performing a static vestibular function test and a dynamic vestibular function test on the crew member to obtain test parameters.

[0033] It should be noted that static vestibular function testing refers to testing the vestibular organ function of crew members under static conditions to evaluate the crew members' perception and adaptability to changes in head position in a static state, including vestibular stimulation testing and visual-vestibular conflict testing.

[0034] Dynamic vestibular function test refers to the test of the crew's vestibular organ function by simulating the dynamic changes of the hull movement or the external environment, so as to evaluate the crew's perception and adaptability to the hull movement or external stimuli in a dynamic environment, including dynamic balance test and hull motion simulation test.

[0035] It should be understood that the test parameters also include static test parameters and dynamic test parameters, which can fully reflect the vestibular function status of the crew.

[0036] Specifically, static vestibular function testing, through vestibular stimulation testing and visual-vestibular conflict testing, can assess a crew member's ability to perceive and adapt to changes in head position while at rest. For example, the vestibular stimulation test uses stimulation methods such as rotation and swaying to observe the crew member's response and adaptability to stimuli; the visual-vestibular conflict test uses the conflict between visual and vestibular information to assess the crew member's adaptability in conflicting situations.

[0037] Dynamic vestibular function tests assess the crew's vestibular function by simulating ship motion or dynamic changes in the external environment. For example, dynamic balance tests assess a crew's balance and adaptability to dynamic changes by having them stand or walk on an unstable surface. Ship motion simulation tests assess a crew's perception of and adaptability to ship motion by simulating the movement of a ship in complex environments, such as waves and currents.

[0038] By using the parameters of static and dynamic vestibular function tests, we can more accurately understand the vestibular function status of the crew members and then accurately determine the crew members to be adjusted.

[0039] Step A12: Evaluate the vestibular function stability level of the crew member under different sailing conditions according to the test parameters.

[0040] It should be noted that vestibular function stability refers to a crew member's ability to maintain balance and stability under different navigation conditions, such as when the vestibular organs are stimulated. This assessment result will serve as an important basis for the subsequent development of personalized balance training programs.

[0041] Step A13: Determine the vestibular function score of each crew member according to the vestibular function stability level of the crew member under different navigation conditions.

[0042] It should be noted that to ensure efficient resource utilization, adjustments are prioritized for crew members with weaker vestibular function, who may be more susceptible to seasickness and other discomfort. Therefore, crew members are scored to determine which ones require more rigorous adaptation cycles. The vestibular function score directly reflects the strength of a crew member's vestibular function and their ability to adapt to the sailing environment.

[0043] It is understood that the crew's static and dynamic test parameters are compared with the standard vestibular function stability parameters. Based on the comparison results, the vestibular function stability level of each crew member under different navigation conditions is determined, such as excellent, good, fair, poor, or bad. Then, based on the score range corresponding to the stability level, a vestibular function score is determined for each crew member. The higher the score, the stronger the crew member's vestibular function and the better their ability to adapt to the navigation environment.

[0044] Step A14: determining the crew members to be adjusted according to the vestibular function scores of the crew members.

[0045] It should be noted that crew members who need adjustment are those with low vestibular function scores and need to undergo adaptive cycle adjustment before sailing. Vessels that need adaptive cycle adjustment are selected based on vestibular function scores.

[0046] In a feasible embodiment, step A11 may include: applying different vestibular stimulations to the crew through a vestibular stimulation device and detecting the crew's head position dependent reaction, vestibular adaptability, and spatial orientation ability; providing a conflict scenario between the visual and vestibular systems through a virtual reality device and testing the crew's adaptability and vestibular perception stability in the virtual environment; simulating the movement of the underwater vehicle under different navigation states through a dynamic balance test device and detecting the crew's balance maintenance method under hull motions of different frequencies and amplitudes; using a hull motion simulator to simulate different navigation conditions and collect gait data, body position data, and reaction data of the crew in different dynamic environments; using the head position dependent reaction, the vestibular adaptability, the spatial orientation ability, the crew's adaptability in the virtual environment, and vestibular perception stability as static test data; using the crew's balance maintenance method under hull motions of different frequencies and amplitudes and the crew's gait data, body position data, and reaction data in different dynamic environments as dynamic test data.

[0047] It should be noted that vestibular stimulation equipment refers to equipment that can produce stimulation such as rotation and swaying, which is used to simulate the stimulation of the vestibular organs by the movement of the hull or the external environment to evaluate the vestibular function status of the crew. For example, equipment such as rotating chairs, visual stimulation, and tilting platforms apply different vestibular stimulations (such as rotation, rolling, etc.) to detect the crew's head position-dependent reaction, vestibular adaptability, and spatial orientation ability. Among them, head position-dependent reaction refers to the crew's perception and response ability to spatial orientation and balance when the head position changes; vestibular adaptability refers to the degree of adaptation of the crew's vestibular system to stimulation under continuous or repeated vestibular stimulation; spatial orientation ability refers to the crew's ability to accurately judge their own position and movement status in space.

[0048] Virtual reality equipment provides virtual environments and scenarios, simulating ship motion or changes in the external environment to assess crew members' adaptability and vestibular stability. For example, using devices such as virtual reality helmets, they can create conflicting scenarios between the visual and vestibular systems, allowing crew members to observe their reactions and adaptability in the virtual environment.

[0049] Dynamic balance testing equipment simulates the motion of underwater vehicles under different navigational conditions, examining how crew members maintain balance under varying frequencies and amplitudes of hull motion. For example, using balance boards and unstable surfaces, the equipment simulates the motion of the hull in complex environments like waves and currents, observing the crew's balance and ability to adapt to dynamic changes.

[0050] The ship motion simulator can simulate different sailing conditions, such as speed, depth, and heading, collecting gait, body position, and reaction data from the crew in these dynamic environments to comprehensively assess their vestibular function and adaptability. This data will serve as an important basis for developing personalized balance training programs.

[0051] By conducting static and dynamic vestibular function tests on crew members, combined with the test results of vestibular stimulation equipment, virtual reality equipment, dynamic balance testing equipment and hull motion simulators, we can more accurately understand the vestibular function status of crew members and provide a scientific basis for the subsequent formulation of personalized adaptive periodic adjustment plans.

[0052] Step S20: Generate a personal state adjustment plan based on the physiological activity data, living environment data and navigation task data through a large language model.

[0053] It should be noted that individual state adjustment plans refer to the intelligent analysis and prediction of crew members' physiological activity data, living environment data, and navigation mission data through a large language model, providing personalized adjustment suggestions for crew members to improve their physical and mental state and enhance their adaptability to the navigation environment. The large language model can understand and process large amounts of text data, including crew members' physiological indicators, living environment information, and navigation mission content. Through deep learning algorithms, it discovers the correlations and regularities between the data, thereby generating adjustment plans that meet the individual needs of the crew members.

[0054] In a feasible implementation, step S20 may include steps B11 to B14: Step B11: Obtaining sleep data, exercise data, and nutritional data based on the physiological activity data.

[0055] It should be noted that, in this embodiment, physiological activity data includes sleep data, exercise data, and nutritional data. Sleep data refers to the crew's sleep quality, sleep duration and other related data during the voyage. These data can reflect the crew's sleep condition and thus assess their fatigue level and recovery needs; exercise data refers to the crew's physical activity amount, exercise type and other related data during the voyage. These data can reflect the crew's physical fitness and exercise habits, and help to formulate appropriate exercise recommendations; nutritional data refers to the crew's dietary intake during the voyage, including the intake of nutrients such as calories, protein, and vitamins. These data can reflect the crew's nutritional status and provide a basis for formulating personalized dietary recommendations. By analyzing these physiological activity data, we can fully understand the crew's physical and mental state, and provide an important reference for the subsequent formulation of personal state adjustment plans.

[0056] Step B12: Obtain noise data, temperature and humidity data, and air quality data based on the living environment data.

[0057] It should be noted that living environment data refers to data related to the internal environment of the underwater vehicle in which the crew members are located, including noise, temperature, humidity, and air quality data. Noise data reflects the acoustic environment within the cabin; excessive noise levels can adversely affect the crew's hearing and mental health. Temperature and humidity data reflect changes in cabin temperature and humidity; a suitable temperature and humidity environment helps the crew maintain good physical condition. Air quality data, including oxygen content, carbon dioxide concentration, and harmful gas concentrations within the cabin air, are crucial for assessing the comfort and safety of the cabin environment. By analyzing this living environment data, we can understand the crew's comfort level and potential risks, providing an environmental basis for developing individual state adjustment plans.

[0058] Step B13: Based on the sleep data, exercise data, nutrition data, noise data, temperature and humidity data, air quality data, and navigation mission data, the psychological state, physical condition, and work stress level of the crew member to be adjusted are identified through a large language model.

[0059] It should be noted that mission data refers to various information related to the mission, such as the route, mission objectives, voyage time, and depth. This information can reflect the workload and challenges a crew member may face during the voyage, as well as their adaptation to the mission. By analyzing mission data, we can understand the crew's work tasks, work intensity, and work pace.

[0060] In its specific implementation, the large language model comprehensively considers the crew's sleep data, exercise data, nutritional data, noise data, temperature and humidity data, air quality data, and navigation mission data, and uses deep learning algorithms to perform correlation analysis and pattern recognition on this data, thereby accurately assessing the crew's psychological state, physical condition, and work stress level. For example, by analyzing sleep data, it is possible to determine whether the crew's sleep quality is good, whether they suffer from insomnia or lack of sleep, and then assess their fatigue level and emotional stability; by analyzing exercise data and nutritional data, it is possible to understand the crew's physical fitness and nutritional status, providing a basis for formulating appropriate exercise and dietary recommendations; by analyzing noise data, temperature and humidity data, and air quality data, it is possible to assess the comfort of the crew's environment and its impact on their physical and mental health; finally, combined with navigation mission data, it is possible to comprehensively consider the crew's work tasks, work intensity, and work pace, and further assess their work stress and adaptability.

[0061] Step B14: Generate a personal state adjustment plan based on the psychological state, physical condition and work stress level of the crew member to be adjusted, wherein the personal state adjustment plan at least includes a work and rest adjustment plan, an exercise plan, and a nutritional diet guidance plan.

[0062] It should be noted that the work and rest adjustment plan refers to the development of a personalized work and rest schedule based on the crew's sleep data and psychological state, including reasonable sleep and rest time to ensure that the crew can get sufficient rest and recovery. For example, for crew members with poor sleep quality, their work and rest schedule can be appropriately adjusted to avoid resting in noisy or bright environments to improve their sleep quality; for crew members with high work pressure, appropriate relaxation and entertainment time can be arranged to relieve their psychological stress. The exercise plan plan is based on the crew's physical fitness and exercise data, and the plan includes a personalized exercise plan, including exercise type, intensity, and duration, to improve the crew's physical fitness and immunity. The nutritional diet guidance plan is based on the crew's nutritional data and physical condition, and the plan includes a personalized diet plan, including a reasonable meal structure and nutrient intake, to ensure that the crew can get adequate nutrition and energy.

[0063] Through personal state adjustment plans, the physical and mental state of the crew can be effectively improved, and their ability to adapt to the navigation environment can be enhanced, thereby ensuring the smooth progress of the navigation mission.

[0064] Step S30: acquiring the hull motion data of the underwater vehicle, and generating a virtual reality balance training program according to the hull motion data and the vestibular function data of the crew member to be adjusted.

[0065] It's understood that underwater vehicle hull motion data refers to various motion information about the hull during navigation, such as roll amplitude, roll frequency, pitch angle, and roll angle. This data can be collected in real time by onboard sensors and reflects the complexity and variability of the navigation environment. Vestibular function data refers to the crew's vestibular organs' ability to perceive and adapt to hull motion, including the vestibular function score mentioned above.

[0066] It's worth noting that the VR balance training program utilizes virtual reality technology, combined with underwater vehicle motion data and vestibular function data from the crew member being trained, to provide a personalized balance training program for the crew. By simulating a realistic sailing environment, the program aims to help crew members improve their vestibular system's adaptability and balance, alleviate symptoms like seasickness, and enhance their ability to adapt to the sailing environment.

[0067] Optionally, the ship's motion data can be compared and analyzed with the crew's vestibular function data. Using a large language model or specialized algorithms, the crew's vestibular function characteristics and potential problems under different ship motion conditions can be identified. For example, for crew members whose balance ability is poor under large or frequent ship motions, balance training under these specific conditions can be strengthened. Based on these analysis results, a personalized virtual reality balance training program is generated. This program utilizes virtual reality technology to simulate the motion of underwater vehicles under different navigation conditions, providing crew members with a safe and controllable training environment. By adjusting ship motion parameters in the virtual reality scene, such as rolling amplitude, frequency, and pitch angle, crew members can conduct balance training in conditions close to those of a real-world navigation environment, thereby improving their adaptability to the navigation environment. During training, the crew's vestibular function responses and balance status can be monitored in real time, allowing timely adjustments to the training program based on training results to ensure effective and targeted training. By implementing the virtual reality balance training program, crew members' balance ability and adaptability to the navigation environment can be further improved, effectively ensuring the smooth progress of navigation missions.

[0068] Step S40: predicting a biofeedback change trend of the crew member to be adjusted based on the personal state adjustment program and the virtual reality balance training program.

[0069] It should be noted that the biofeedback change trend refers to the changing trend of the crew's physiological and psychological state after they receive the personal state adjustment program and virtual reality balance training program, including changes in biological indicators such as heart rate, blood pressure, muscle tension, and emotional stability.

[0070] By monitoring and analyzing these bioindicators in real time, the effectiveness of conditioning and training programs can be evaluated, allowing adjustments to the intensity and frequency of balance training to ensure that the training is both challenging and not overly burdensome for the crew. For example, if a crew member's heart rate is detected to be abnormally elevated or their mood becomes unstable, the intensity of the ship's movements in the virtual reality scenario can be appropriately reduced to prevent excessive fatigue or stress. Conversely, if the crew member demonstrates good adaptability and progress, the difficulty of the training can be gradually increased to further enhance their balance and adaptability. This dynamic adjustment strategy based on biofeedback ensures a safe and efficient training process, maximizing the crew's adaptability and the successful completion rate of navigation missions.

[0071] In a feasible implementation, step S40 may include steps C11 to C15: Step C11: constructing an initial biofeedback prediction model, wherein the initial biofeedback prediction model is composed of a superposition of several restricted Boltzmann machine networks.

[0072] It should be noted that the restricted Boltzmann machine network is a commonly used deep learning model with powerful feature extraction and pattern recognition capabilities. In this embodiment, a deep biofeedback prediction model is constructed by superimposing multiple restricted Boltzmann machine networks. This model can receive crew members' physiological and psychological status data, such as heart rate, blood pressure, muscle tension, and emotional stability, as input. By learning the complex relationship between this data and biofeedback trends, it predicts the biofeedback trends of crew members after receiving a personal state adjustment program and a virtual reality balance training program.

[0073] Step C12: Obtain historical biofeedback data, historical personal state adjustment plans, and historical virtual reality balance training plans of the crew member to be adjusted.

[0074] It should be noted that historical biofeedback data, historical personal state adjustment plans, and historical virtual reality balance training plans serve as training samples for training and optimizing the initial biofeedback prediction model. By continuously adjusting the model's parameters and structure, it can more accurately predict the biofeedback changes of crew members after undergoing different adjustment and training plans.

[0075] Step C13: constructing a training set based on the historical biofeedback data, historical personal state adjustment program, and historical virtual reality balance training program.

[0076] It should be noted that the training set is the data set used to train and optimize the biofeedback prediction model. By organizing and preprocessing historical biofeedback data, historical personal state adjustment plans, historical virtual reality balance training plans, and other data, a training set containing multiple training samples is constructed. These training samples include physiological and psychological state data of crew members under different adjustment and training plans, as well as corresponding biofeedback change trends, providing rich information for model training. During training, the model learns the characteristics and patterns in these training samples, gradually optimizing its parameters and structure to improve the accuracy and reliability of predictions. By constructing a training set, strong data support is provided for the training of the biofeedback prediction model, ensuring that the model can perform well in practical applications.

[0077] Step C14: training the initial biofeedback prediction model using the training set to obtain a target biofeedback prediction model.

[0078] It should be noted that training the initial biofeedback prediction model using the training set is an iterative and optimization process. During training, the model continuously learns the characteristics and patterns of the training samples, gradually improving the accuracy and reliability of predictions by adjusting its internal parameters and structure. Once training is complete, the resulting target biofeedback prediction model will be able to accept new physiological and psychological state data from crew members and predict changes in their biofeedback after receiving the personalized state adjustment program and virtual reality balance training program.

[0079] Step C15: predicting biofeedback data of the crew member to be adjusted according to the personal state adjustment program and the virtual reality balance training program through the target biofeedback prediction model, wherein the biofeedback data at least includes the changing trends of heart rate, blood pressure, skin conductivity, respiratory rate and brain wave activity.

[0080] It's important to note that the targeted biofeedback prediction model can accurately and in real time predict the physiological and psychological state trends of crew members undergoing specific conditioning and training programs. These biofeedback data, such as trends in heart rate, blood pressure, skin conductance, respiratory rate, and brainwave activity, provide important insights for further optimizing conditioning and training programs. For example, if a crew member's heart rate is predicted to increase abnormally after training, the training program can be adjusted in advance, reducing the intensity or increasing rest periods to prevent excessive fatigue. Conversely, if a crew member is predicted to demonstrate good adaptability and progress, the training difficulty can be appropriately increased to further enhance their balance and adaptability. This dynamic adjustment strategy based on predictions ensures a more personalized, precise, and efficient training process, maximizing the crew member's adaptability and mission completion rate.

[0081] In a feasible embodiment, step C14 may include: generating a corresponding biofeedback label based on the historical biofeedback data, wherein the biofeedback label includes a change trend of the historical biofeedback data; vectorizing the historical personal state adjustment program and the historical virtual reality balance training program to obtain a corresponding feature vector; using the corresponding feature vector as a first training sample, and combining the biofeedback label and the corresponding feature vector into a second training sample; training the first restricted Boltzmann machine network of the initial biofeedback prediction model through the first training sample to obtain a reconstruction error of the first restricted Boltzmann machine network; when the reconstruction error of the first restricted Boltzmann machine network is less than a preset error threshold, outputting the first restricted Boltzmann machine network. As the input of the second restricted Boltzmann machine network, continue to train the second restricted Boltzmann machine network, and so on until the training of the last layer of restricted Boltzmann machine network is completed, and the initially trained biofeedback prediction model is obtained, wherein the training process of each layer of restricted Boltzmann machine network includes forward propagation and reverse adjustment, forward propagation is used to calculate the output of the current layer of restricted Boltzmann machine network, and reverse adjustment is used to adjust the weights and biases of the current layer of restricted Boltzmann machine network according to the output error; the initially trained biofeedback prediction model is fine-tuned through the second training sample to obtain the target biofeedback prediction model, wherein the fine-tuning training process includes using the cross-validation method to verify the initially trained biofeedback prediction model, and adjusting the model parameters according to the verification results.

[0082] It should be noted that the initial biofeedback prediction model is composed of multiple stacked restricted Boltzmann machines (RBMs), which is an important multi-layer network structure in the deep learning network model. Its network training method adopts a semi-supervised learning algorithm: the weights and bias values ​​of the restricted Boltzmann machine network are initialized through a greedy layer-by-layer unsupervised algorithm training, and then these weights and bias values ​​are fine-tuned through a supervised back-propagation algorithm.

[0083] It's understandable that the purpose of vectorizing the historical personal state adjustment plans and historical virtual reality balance training plans is to convert the parameters and features within these plans into numerical form, making them easier to process and calculate by the model. Vectorization is a common technique in machine learning. By converting data into vector form, it's easier to capture relationships and features within the data, improving the model's predictive performance.

[0084] In the specific implementation, historical personal state adjustment plans and historical virtual reality balance training plans are vectorized to obtain corresponding feature vectors. These feature vectors contain key information and features of the plans, providing important input data for the training of the biofeedback prediction model. These feature vectors are combined with the corresponding biofeedback labels to form a second training sample, which is used to fine-tune the initially trained biofeedback prediction model. Fine-tuning training is the process of fine-tuning model parameters. By verifying the initially trained biofeedback prediction model using cross-validation methods and adjusting the model parameters based on the verification results, the model's predictive accuracy and reliability can be further improved. Ultimately, the target biofeedback prediction model obtained after fine-tuning training will be able to more accurately predict the biofeedback change trends of the crew member after receiving the personal state adjustment plan and virtual reality balance training plan, providing more precise and personalized guidance for the crew member's adaptive cycle adjustment.

[0085] It is worth noting that for l The forward propagation process of the restricted Boltzmann machine (RBM) of the layer is:

[0086] in, is the visible layer neuron output of the current layer, is the hidden layer output of the previous layer, is the weight matrix from the previous layer to the current layer, is the bias term of the current layer, As the activation function, sigmoid is usually used.

[0087] The reconstruction error is an important indicator to measure the difference between the output of the current layer of restricted Boltzmann machine network and the real data. During the training process, it is hoped that this error is as small as possible, because it reflects the degree of fit of the model to the data. When the reconstruction error is less than the preset error threshold, it is considered that the restricted Boltzmann machine network of the current layer has learned the characteristics of the data well, and its output can be used as the input of the next layer of network to continue training the next layer of network. This layer-by-layer training method enables the entire deep network to gradually learn the deep-level characteristics of the data, thereby improving the prediction performance of the model. Through the output of the current layer RBM network , calculate the reconstruction error , can be calculated using the mean square error (MSE), and the formula for the reconstruction error is:

[0088] in, is the original input data (e.g. historical biofeedback data), is the reconstruction result output by the current layer, and N is the total number of RBM layers.

[0089] For each layer of RBM, backpropagation is performed to adjust the weights and bias ,The training process of each layer includes forward propagation, error calculation and ,backward propagation until the preset error threshold is reached. If the reconstruction error is less than the threshold, the ,output of the current layer is used as the input of the next layer, and ,training continues on the next layer.

[0090] Fine-tuning training is verified by cross-validation method to obtain the optimal model parameters, as shown below:

[0091] in, is the loss function, usually the mean square error (MSE), For biofeedback tags, is the biofeedback value predicted by the model, is the model parameter, and N is the total number of RBM layers.

[0092] Step S50: performing balance training on the crew member to be adjusted based on the biofeedback change trend and the virtual reality balance training program and controlling the training intensity of the crew member to be adjusted to achieve adaptive periodic adjustment of the crew member.

[0093] It's important to note that based on biofeedback trends and virtual reality balance training programs, personalized balance training can be provided to crew members undergoing adjustments. Biofeedback trends reflect changes in a crew member's physiological and psychological state after adjustment and training, and serve as an important basis for evaluating training effectiveness and adjusting training intensity. Real-time monitoring and analysis of this biofeedback data allows us to understand the crew member's adaptation and progress, allowing for timely adjustments to the training program to ensure both effectiveness and safety.

[0094] In practice, the difficulty and intensity of VR balance training programs can be dynamically adjusted based on biofeedback trends. For example, if biofeedback data indicates improved adaptability in a particular area, the training difficulty can be appropriately increased to further enhance balance and adaptability. Conversely, if biofeedback data indicates excessive fatigue or decreased adaptability, the training intensity should be promptly reduced or rest periods increased to avoid adverse effects on the crew's physical and mental health.

[0095] It's understandable that balance training can effectively improve crew members' sense of balance and spatial cognition, enhancing their adaptability and stability in the navigation environment. During balance training, virtual reality technology provides crew members with a safe, controllable, and challenging training environment, allowing them to practice in simulated navigation scenarios, thereby improving their performance in real-world navigation missions. Furthermore, combined with the changing trends of biofeedback data, a more accurate assessment of the crew members' training effectiveness and adaptability can be achieved, not only improving the efficiency and effectiveness of training but also ensuring their physical and mental health and navigation safety.

[0096] This embodiment provides a crew member adaptation cycle adjustment method based on a large language model. The method comprises collecting physiological activity data, living environment data, and navigation task data of the crew member to be adjusted through a wearable device in the confined space of an underwater vehicle; generating a personal state adjustment plan based on the physiological activity data, living environment data, and navigation task data through a large language model; obtaining the underwater vehicle's hull motion data, and generating a virtual reality balance training plan based on the hull motion data and the vestibular function data of the crew member to be adjusted; predicting the biofeedback change trend of the crew member to be adjusted based on the personal state adjustment plan and the virtual reality balance training plan; performing balance training on the crew member to be adjusted based on the biofeedback change trend and the virtual reality balance training plan and controlling the training intensity of the crew member to be adjusted to achieve crew member adaptation cycle adjustment. In this way, by introducing a large language model and virtual reality technology, combining personal state adjustment with balance training, accurately regulating the crew member's balance training process, enabling the crew member to quickly adapt to the navigation environment, thereby shortening the navigation adaptation cycle and improving the crew member's navigation work efficiency.

[0097] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above embodiment 1 can be referred to the above introduction and will not be described in detail later. Figure 2 , step S30 includes steps S301 to S304: Step S301: The underwater vehicle's sensors are used to obtain real-time motion data of the hull, including the hull's sway angle, pitch angle, roll angle, and acceleration data.

[0098] It's important to note that underwater vehicle sensors can include gyroscopes, accelerometers, and magnetometers. The roll angle, the tilt of the hull relative to the horizontal plane, describes the ship's side-to-side swaying. The pitch angle, the fore-aft tilt of the hull, describes its pitching. The roll angle, the angle of rotation about its longitudinal axis, describes its roll. Acceleration data reflects the changes in the hull's acceleration in different directions. Together, these data provide a comprehensive description of the hull's motion. By acquiring this real-time motion data, the dynamic state of the underwater vehicle can be accurately determined.

[0099] Step S302: Determine the balance training difficulty and training mode of the crew member according to the vestibular function score of the crew member to be adjusted and the hull motion data.

[0100] It should be noted that the vestibular function score is obtained by conducting a vestibular function test on the crew. This test can evaluate the health status of the crew in terms of the function of the inner ear vestibular organs, including their adaptability and reaction speed to different motion stimuli.

[0101] The difficulty and training mode of balance training are determined based on individual differences and vestibular function status of crew members to ensure that the training is both challenging and safe and feasible. For example, for crew members with good vestibular function, more difficult training modes and more complex motion stimulation can be selected to further improve their balance ability; while for crew members with poor vestibular function, relatively simple training modes and moderate motion stimulation can be selected to avoid causing discomfort or aggravating symptoms. Through personalized balance training programs, crew members' balance ability and adaptability can be more effectively improved.

[0102] In a specific implementation, determining the balance training difficulty and training mode for the crew member to be adjusted based on the crew member's vestibular function score and hull motion data specifically includes: comparing the hull motion data with a preset threshold range to determine the intensity of the hull motion. If the hull motion data exceeds the preset threshold range, indicating that the hull motion is relatively intense, the balance training difficulty should be appropriately reduced to avoid excessive physical strain on the crew member; conversely, if the hull motion data is within the preset threshold range, the balance training difficulty can be appropriately increased to better train the crew member's balance ability. Secondly, the balance training difficulty and training mode are further refined based on the crew member's vestibular function score. For example, for crew members with high vestibular function scores, even if the hull motion is relatively intense, a relatively high training difficulty and complex training mode can be considered to challenge their balance ability to the limit; while for crew members with lower vestibular function scores, greater emphasis should be placed on training safety and adaptability, with a simpler training mode and moderate difficulty.

[0103] Step S303: Based on the balance training difficulty and training mode, a balance training scene matching the motion state of the underwater vehicle is generated through virtual reality technology, wherein the balance training scene includes a static balance training scene and a dynamic balance training scene.

[0104] It should be noted that the balance training scenario is dynamically generated based on the crew's balance training difficulty and training mode, and is designed to simulate the environment of underwater vehicles in different motion states to help crew members adapt to and improve their balance ability. The static balance training scenario mainly simulates the environment of underwater vehicles in a relatively stable state, such as slow navigation or mooring. At this time, the crew needs to maintain a stable standing posture in the virtual environment to exercise their static balance ability. The dynamic balance training scenario simulates the environment of underwater vehicles in complex and changeable motion states, such as fast navigation, sharp turns or encountering water currents. At this time, the crew needs to quickly adjust their body posture in the virtual environment to maintain balance, thereby exercising their dynamic balance ability. By alternating between these two balance training scenarios, the crew's balance ability and adaptability can be comprehensively improved.

[0105] In specific implementation, virtual reality technology uses high-precision sensors and algorithms to capture the crew's body posture and movement trajectory in real time, and dynamically adjusts the visual and auditory stimulation in the virtual environment according to the requirements of balancing training difficulty and training mode to provide a more real and realistic training experience.

[0106] In a feasible implementation, step S303 may include: constructing a virtual reality scene of the internal environment of the underwater vehicle through 3D modeling technology, including the structural layout, equipment distribution and lighting conditions inside the underwater vehicle; integrating the actual environmental elements of the underwater vehicle into the virtual reality scene to generate a static balance training scene; setting balance training tasks of different difficulty levels based on the balance training difficulty and training mode; and generating corresponding dynamic balance training scenes according to the balance training tasks of different difficulty levels.

[0107] It should be noted that 3D modeling technology was used to precisely measure and model each area of ​​the underwater vehicle to ensure a high degree of consistency between the virtual reality scene and the actual environment. During the modeling process, special attention was paid to key areas where crew members may conduct balance training, such as the control room and lounge, to ensure accurate detail in these areas. Furthermore, given the unique lighting conditions within the underwater vehicle, professional lighting simulation technology was employed to replicate the lighting and shadow effects of a real environment, enhancing the realism of the virtual reality scene.

[0108] Specifically, when integrating the actual environmental elements of underwater vehicles into virtual reality scenes, emphasis is placed on simulating the unique underwater sound and vibration effects, such as the sound of flowing water, the sound of equipment running, and the vibration of the hull shaking. These elements can further enhance the crew's sense of immersion and training effect in the virtual environment.

[0109] When setting balance training tasks of varying difficulty, we design diverse training tasks based on the crew's balance training difficulty and training mode, such as standing on one leg, walking with eyes closed, and rapid turns, to challenge the crew's balance ability. At the same time, we gradually increase the training intensity according to the difficulty of the task to ensure that the crew can gradually improve their balance ability during the training process.

[0110] When generating dynamic balance training scenarios, dynamic generation technology is used to dynamically adjust the movement state and difficulty level in the virtual environment based on the crew's real-time body posture and movement trajectory, providing a more personalized and adaptive training experience. For example, when a crew member demonstrates high balance ability in the virtual environment, the movement speed and difficulty level can be automatically increased to challenge their limits; while when a crew member shows signs of fatigue or discomfort, the movement speed and difficulty level can be appropriately reduced to ensure the safety and effectiveness of the training.

[0111] It's worth noting that when generating dynamic balance training scenarios, it's also possible to consider incorporating random elements, such as sudden changes in water flow direction, speed, or intensity, to simulate the motion of underwater vehicles encountering emergencies, further training crew members' adaptability and balance adjustment abilities. Furthermore, virtual reality technology can be combined with sound and tactile feedback, such as the sound of simulated water flow and the tactile sensation of hull vibration, to enhance the realism and immersion of training, helping crew members better adapt to the motion of underwater vehicles.

[0112] Optionally, in a static balance training scenario, stable visual reference points and balance challenge tasks are set, requiring crew members to complete specific actions or tasks while maintaining balance; in a dynamic balance training scenario, hull motions of different frequencies and amplitudes are simulated according to the hull motion data, requiring crew members to maintain balance under the simulated hull motion state and adapt to different dynamic environments; through the real-time feedback mechanism in the virtual reality environment, the crew members' performance data and training effect evaluation during the balance training process are provided, so as to further optimize the personal state adjustment plan and balance training plan.

[0113] Step S304: generating a virtual reality balance training program according to the balance training scenario, wherein the virtual reality balance training program includes training time, number of training sessions, training content, and a training feedback mechanism.

[0114] It's important to note that the VR balance training program is developed based on individual crew differences, vestibular function scores, ship motion data, and the desired balance training difficulty and training mode. Training schedules must take into account crew members' daily work and rest schedules to ensure that training doesn't interfere with their normal work and life rhythms. The frequency of training sessions is adjusted based on the crew's balance improvement rate and training effectiveness, ensuring that they are neither too frequent, which can lead to fatigue, nor too infrequent, which can compromise training effectiveness.

[0115] The training program encompasses both static and dynamic balance training, aiming to comprehensively enhance the crew's balance and adaptability. Static balance training primarily focuses on improving the crew's ability to control their balance in relatively stable conditions, such as maintaining specific postures or performing specific movements to enhance their sense of balance. Dynamic balance training, on the other hand, emphasizes the crew's reaction speed and ability to adjust to complex and changing motions, challenging their balance limits by simulating the various motions of underwater vehicles.

[0116] The training feedback mechanism is a crucial component of the VR balance training program. It evaluates training effectiveness by monitoring crew members' performance data during training in real time, such as body posture, movement trajectory, and balance maintenance time. Furthermore, based on this feedback, the training program can be dynamically adjusted to better suit the crew members' individual needs and training progress. For example, if a crew member's balance ability is found to be weak in a particular area, the difficulty and frequency of the relevant training content can be increased to strengthen targeted training.

[0117] The VR balance training program is a personalized approach that comprehensively considers multiple factors, including individual crew differences, vestibular function scores, vessel motion data, and training difficulty and mode. Through scientific training design and real-time feedback mechanisms, it effectively improves crew balance and adaptability, providing a strong guarantee for the safe navigation of underwater vehicles.

[0118] In this embodiment, the balance training difficulty and training mode are determined by the vestibular function score of the crew member to be adjusted and the hull motion data, and then the corresponding balance training scene is matched and a corresponding virtual reality balance training program is generated. Through scientific training design and real-time feedback mechanism, the crew member's balance ability and adaptability can be effectively improved.

[0119] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the crew adaptive periodic adjustment method based on the large language model of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0120] This application also provides a crew adaptive period adjustment device based on a large language model, please refer to Figure 3, the crew adaptive period adjustment device based on the large language model includes: The collection module 10 is used to collect physiological activity data, living environment data and navigation mission data of the crew member to be adjusted through wearable devices in the confined space of the underwater vehicle.

[0121] The generating module 20 is configured to generate a personal state adjustment plan based on the physiological activity data, the living environment data and the navigation task data through a large language model.

[0122] The acquisition module 30 is used to acquire the hull motion data of the underwater vehicle and generate a virtual reality balance training program based on the hull motion data and the vestibular function data of the crew member to be adjusted.

[0123] The prediction module 40 is used to predict the biofeedback change trend of the crew member to be adjusted based on the personal state adjustment program and the virtual reality balance training program.

[0124] The training module 50 is used to perform balance training on the crew member to be adjusted based on the biofeedback change trend and the virtual reality balance training program and control the training intensity of the crew member to be adjusted to achieve adaptive periodic adjustment of the crew member.

[0125] The crew adaptive periodic adjustment device based on a large language model provided in this application adopts the crew adaptive periodic adjustment method based on a large language model in the above-mentioned embodiment, which can solve the technical problem that traditional crew training methods are difficult to fully simulate the complex environment and dynamic changes in actual navigation, resulting in crew members still needing a long time to adapt during actual navigation. Compared with the existing technology, the beneficial effects of the crew adaptive periodic adjustment device based on a large language model provided in this application are the same as the beneficial effects of the crew adaptive periodic adjustment method based on a large language model provided in the above-mentioned embodiment, and the other technical features of the crew adaptive periodic adjustment device based on a large language model are the same as the features disclosed in the above-mentioned embodiment method, and will not be repeated here.

[0126] The present application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, and the computer-readable program instructions are used to execute the crew adaptive period adjustment method based on a large language model in the above-mentioned embodiment.

[0127] The computer-readable storage medium provided herein may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems, or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more wires, a portable computer disk, a hard disk, RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory or Flash memory), optical fiber, CD-ROM (CD-Read Only Memory), optical storage device, magnetic storage device, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including, but not limited to, wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0128] The computer-readable storage medium may be included in the crew adaptive periodic adjustment device based on a large language model; or it may exist independently without being assembled into the crew adaptive periodic adjustment device based on a large language model.

[0129] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by the crew adaptive periodic adjustment device based on the large language model, the crew adaptive periodic adjustment device based on the large language model: collects physiological activity data, living environment data and navigation task data of the crew to be adjusted through wearable devices in the confined space of the underwater vehicle; generates a personal state adjustment plan through the large language model based on the physiological activity data, living environment data and navigation task data; obtains the hull motion data of the underwater vehicle, and generates a virtual reality balance training plan according to the hull motion data and the vestibular function data of the crew to be adjusted; predicts the biofeedback change trend of the crew to be adjusted based on the personal state adjustment plan and the virtual reality balance training plan; performs balance training on the crew to be adjusted based on the biofeedback change trend and the virtual reality balance training plan and controls the training intensity of the crew to be adjusted, so as to achieve adaptive periodic adjustment of the crew.

[0130] The computer program code for performing the operations of the present application may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a LAN (Local Area Network) or a WAN (Wide Area Network), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0131] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a module, program segment or a part of code, and the module, program segment or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flow chart, and the combination of the boxes in the block diagram and / or flow chart can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0132] The modules described in the embodiments of the present application may be implemented in software or hardware, wherein the name of a module does not necessarily limit the unit itself.

[0133] The computer-readable storage medium provided in this application is a computer-readable storage medium storing computer-readable program instructions (i.e., a computer program) for executing the aforementioned large language model-based crew adaptation period adjustment method. This computer-readable storage medium can address the technical issue that traditional crew training methods struggle to fully simulate the complex environments and dynamic changes experienced during actual voyages, resulting in crew members requiring a significant amount of time to adapt during actual voyages. Compared to the prior art, the beneficial effects of the computer-readable storage medium provided in this application are similar to those of the large language model-based crew adaptation period adjustment method provided in the aforementioned embodiments and are not further elaborated upon here.

[0134] The above description is only part of the embodiments of the present application and does not limit the patent scope of the present application. All equivalent structural transformations made by using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A crew adaptive period adjustment method based on a large language model, characterized in that: The method comprises: In the confined space of the underwater vehicle, wearable devices are used to collect physiological activity data, living environment data, and navigation mission data of the crew members to be adjusted; Generate a personal state adjustment plan based on the physiological activity data, living environment data and navigation task data through a large language model; Acquiring hull motion data of the underwater vehicle, and generating a virtual reality balance training program based on the hull motion data and vestibular function data of the crew member to be adjusted; Predicting a biofeedback change trend of the crew member to be adjusted based on the personal state adjustment program and the virtual reality balance training program; Based on the biofeedback change trend and the virtual reality balance training program, balance training is performed on the crew member to be adjusted and the training intensity of the crew member to be adjusted is controlled to achieve adaptive periodic adjustment of the crew member.

2. The method according to claim 1, wherein Before collecting the physiological activity data, living environment data, and navigation mission data of the crew member to be adjusted by the wearable device in the confined space of the underwater vehicle, the method further includes: Conduct static vestibular function tests and dynamic vestibular function tests on crew members to obtain test parameters; Evaluate the vestibular function stability level of crew members under different sailing conditions according to the test parameters; Determining a vestibular function score for each crew member according to the crew member's vestibular function stability level under different sailing conditions; The crew members to be adjusted are determined according to the vestibular function scores of the crew members.

3. The method according to claim 2, wherein The test data includes static test data and dynamic test data. The static vestibular function test and the dynamic vestibular function test are performed on the crew to obtain test parameters, including: Apply different vestibular stimulations to the crew members through vestibular stimulation equipment and test the crew members' head position-dependent response, vestibular adaptability, and spatial orientation ability; Provide conflict situations between the visual and vestibular systems through virtual reality equipment and test crew members' adaptability and vestibular perception stability in the virtual environment; The dynamic balance test equipment is used to simulate the motion of underwater vehicles in different navigation states and to detect how the crew maintains balance under hull motions of different frequencies and amplitudes; Use a ship motion simulator to simulate different sailing conditions and collect gait data, body position data, and reaction data of crew members in different dynamic environments; taking the head position-dependent response, the vestibular adaptability, the spatial orientation ability, the crew member's adaptability in a virtual environment, and vestibular perception stability as static test data; The crew's balance-maintaining manner under ship motions of different frequencies and amplitudes, as well as the crew's gait data, body position data, and reaction data in different dynamic environments are used as dynamic test data.

4. The method according to claim 1, wherein The generating of a personal state adjustment plan based on the physiological activity data, the living environment data, and the navigation task data through a large language model includes: Obtaining sleep data, exercise data, and nutritional data based on the physiological activity data; Obtaining noise data, temperature and humidity data, and air quality data based on the living environment data; Based on the sleep data, exercise data, nutrition data, noise data, temperature and humidity data, air quality data, and navigation mission data, the large language model is used to identify the psychological state, physical condition, and work stress level of the crew member to be adjusted; A personal state adjustment plan is generated according to the psychological state, physical condition and work stress level of the crew member to be adjusted, wherein the personal state adjustment plan at least includes a work and rest adjustment plan, an exercise plan, and a nutritional diet guidance plan.

5. The method according to claim 1, wherein The step of obtaining the hull motion data of the underwater vehicle and generating a virtual reality balance training program based on the hull motion data and the vestibular function data of the crew member to be adjusted includes: The underwater vehicle's sensors are used to obtain real-time ship motion data, including the ship's sway angle, pitch angle, roll angle, and acceleration data; determining the difficulty and training mode of balance training for the crew member according to the vestibular function score of the crew member to be adjusted and the hull motion data; Based on the balance training difficulty and training mode, a balance training scene matching the motion state of the underwater vehicle is generated by virtual reality technology, wherein the balance training scene includes a static balance training scene and a dynamic balance training scene; A virtual reality balance training program is generated according to the balance training scenario, wherein the virtual reality balance training program includes training time, training times, training content, and a training feedback mechanism.

6. The method according to claim 5, wherein The method of generating a balance training scene that matches the motion state of the underwater vehicle based on the balance training difficulty and training mode by using virtual reality technology includes: Use 3D modeling technology to create a virtual reality scene of the underwater vehicle's internal environment, including the internal structure layout, equipment distribution, and lighting conditions of the underwater vehicle; Integrate the actual environmental elements of the underwater vehicle into the virtual reality scene to generate a static balance training scene; Setting balance training tasks of different difficulty levels based on the balance training difficulty and training mode; Generate corresponding dynamic balance training scenarios according to the balance training tasks of different difficulty levels.

7. The method according to claim 1, wherein The predicting of the biofeedback change trend of the crew member to be adjusted based on the personal state adjustment program and the virtual reality balance training program includes: Constructing an initial biofeedback prediction model, wherein the initial biofeedback prediction model is composed of a superposition of several restricted Boltzmann machine networks; Obtain historical biofeedback data, historical personal state adjustment plans, and historical virtual reality balance training plans of the crew members to be adjusted; Constructing a training set based on the historical biofeedback data, historical personal state adjustment program, and historical virtual reality balance training program; Training the initial biofeedback prediction model using the training set to obtain a target biofeedback prediction model; The target biofeedback prediction model is used to predict biofeedback data of the crew member to be adjusted according to the personal state adjustment plan and the virtual reality balance training plan. The biofeedback data at least includes the change trends of heart rate, blood pressure, skin conductivity, respiratory rate and brain wave activity.

8. The method according to claim 7, wherein The initial biofeedback prediction model is trained using the training set to obtain a target biofeedback prediction model, including: generating a corresponding biofeedback tag according to the historical biofeedback data, wherein the biofeedback tag includes a change trend of the historical biofeedback data; Vectorizing the historical personal state adjustment scheme and the historical virtual reality balance training scheme to obtain corresponding feature vectors; Using the corresponding feature vector as a first training sample, and combining the biofeedback label and the corresponding feature vector into a second training sample; training a first restricted Boltzmann machine network of the initial biofeedback prediction model using the first training sample to obtain a reconstruction error of the first restricted Boltzmann machine network; When the reconstruction error of the first restricted Boltzmann machine network is less than a preset error threshold, the output of the first restricted Boltzmann machine network is used as the input of the second restricted Boltzmann machine network, and the second restricted Boltzmann machine network is continued to be trained, and so on, until the training of the last layer of restricted Boltzmann machine network is completed, thereby obtaining an initially trained biofeedback prediction model, wherein the training process of each layer of restricted Boltzmann machine network includes forward propagation and backward adjustment, the forward propagation is used to calculate the output of the current layer of restricted Boltzmann machine network, and the backward adjustment is used to adjust the weight and bias of the current layer of restricted Boltzmann machine network according to the output error; The biofeedback prediction model trained at the initial training is fine-tuned using the second training sample to obtain a target biofeedback prediction model, wherein the fine-tuning training process includes verifying the biofeedback prediction model trained at the initial training using a cross-validation method and adjusting model parameters based on the verification results.

9. A crew adaptive adjustment device based on a large language model, characterized in that: The crew adaptive adjustment device based on the large language model includes: A collection module, used to collect physiological activity data, living environment data, and navigation mission data of the crew to be adjusted through wearable devices in the confined space of the underwater vehicle; A generation module, configured to generate a personal state adjustment plan based on the physiological activity data, the living environment data, and the navigation task data through a large language model; an acquisition module, configured to acquire hull motion data of the underwater vehicle and generate a virtual reality balance training program based on the hull motion data and vestibular function data of the crew member to be adjusted; A prediction module, configured to predict a biofeedback change trend of the crew member to be adjusted based on the personal state adjustment program and the virtual reality balance training program; A training module is used to perform balance training on the crew member to be adjusted based on the biofeedback change trend and the virtual reality balance training program and control the training intensity of the crew member to be adjusted to achieve adaptive periodic adjustment of the crew member.

10. A storage medium, characterized in that: The storage medium stores a crew adaptability adjustment program based on a large language model, and when the crew adaptability adjustment program based on a large language model is executed by a processor, the crew adaptability adjustment method based on a large language model as described in any one of claims 1 to 8 is implemented.

Images