A method and device for adjusting the environment of a ship control cabin and an electronic device

Abstract

The application discloses a ship control cabin environment adjusting method and device and electronic equipment, and belongs to the ship control technical field. The method comprises the following steps: obtaining activity data of an operator in a ship control cabin and environment information of the control cabin; extracting working state information of the operator from the activity data; inputting the working state information and the environment information into a parameter prediction model to obtain environment adjusting parameters output by the parameter prediction model, the parameter prediction model being obtained through a plurality of training samples, the training samples comprising sample working state information and sample environment information of a sample operator, and environment adjusting parameter labels in a sample control cabin where the sample operator is located; based on the environment adjusting parameters, calling a mirror model of the operator to perform simulation simulation, and setting the control cabin through the environment adjusting parameters in the case that a simulation result of the mirror model reaches a preset simulation condition. The method can fully consider the needs of different operators for the environment.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of ship control, and particularly relates to an environment adjustment method and device for a ship control cabin and an electronic device. BACKGROUND

[0002] The ship operating cabin is usually designed as a semi-closed or controllable closed environment, and there are special types of ships, and the operating cabin is designed as a highly closed or fully closed environment. When performing a navigation task, the operator needs to work in such a closed and limited space for a long time, and therefore the comfort of the environment is crucial to the physical and mental health of the crew, work efficiency and emergency response capability. Since the control cabin of the special type of ship cannot be directly exchanged with the external environment, the adjustment of natural environmental factors such as air circulation, temperature and humidity becomes particularly important. The closed nature of the operating cabin means that changes in oxygen and carbon dioxide concentrations, temperature and humidity in the air will directly affect the physical condition, psychological state and work efficiency of the crew. Once the environmental parameters change, the crew may experience physical discomfort such as fatigue, dizziness and hypothermia, and even affect their judgment and emergency response capability, which may seriously threaten the success of the task and the safety of the crew.

[0003] At present, in order to protect the health and safety of the crew, sensors are generally used to monitor the environmental parameters in the control cabin and the physiological state of the operator, and an automatic control algorithm is introduced to adaptively adjust the temperature and humidity to ensure that the environmental factors remain within an appropriate range and avoid health problems caused by an unsuitable environment.

[0004] However, there are significant individual differences between different operators, and the above adjustment scheme relies on standardized models and cannot fully consider the environmental needs of different operators. SUMMARY

[0005] The present application aims to at least solve one of the technical problems existing in the prior art. To this end, the present application provides an environment adjustment method, device and electronic device for a ship control cabin, which can fully consider the environmental needs of different operators.

[0006] In a first aspect, the present application provides an environment adjustment method for a ship control cabin, which comprises:

[0007] obtaining activity data of an operator in the ship control cabin and environmental information of the control cabin;

[0008] extracting work state information of the operator from the activity data;

[0009] input the working state information and the environment information into a parameter prediction model to obtain an environment adjustment parameter output by the parameter prediction model, the parameter prediction model being obtained through a plurality of training samples, the training samples including sample working state information and sample environment information of a sample operator and an environment adjustment parameter label in a sample control cabin where the sample operator is located;

[0010] based on the environment adjustment parameter, calling a mirror model of the operator to perform simulation, and in a case where a simulation result of the mirror model reaches a preset simulation condition, setting the control cabin through the environment adjustment parameter, the mirror model being constructed based on physiological characteristics, psychological characteristics and behavior characteristics of the operator.

[0011] According to an embodiment of the present application, the working state information of the operator is extracted from the activity data, including:

[0012] aligning the activity data in scale;

[0013] mapping the aligned activity data to a target graph structure, a node in the target graph structure being used to represent a data point in the activity data, and an edge between nodes in the target graph structure being used to represent an adjacent relationship and similarity between data points;

[0014] extracting features of the target graph structure through a graph attention network to obtain the working state information.

[0015] According to an embodiment of the present application, the activity data is aligned in scale, including:

[0016] calling an inference engine to infer the activity data to determine a rule category and a processing agenda;

[0017] based on the rule category, calling a corresponding processing rule from a rule library to match each type of data in the activity data;

[0018] calling an execution engine to process the corresponding activity data through the processing rule based on the processing agenda, and determining sensitive data and non-sensitive data in the activity data;

[0019] extracting risk assessment features from the sensitive data according to a correlation degree of the sensitive data and the working state of the operator;

[0020] aligning the non-sensitive data and the risk assessment features in scale in a covariance matrix.

[0021] According to one embodiment of the present application, the simulation of the mirror model is based on a state transition model, which includes sequentially connected domain invariant feature extractors, a cross-modal adversarial fusion module, and an environment parameter prediction engine.

[0022] The mirror model of the operator is called for simulation based on the environment adjustment parameters, and in a case where a simulation result of the mirror model reaches a preset simulation condition, the control cabin is set by the environment adjustment parameters, including:

[0023] The activity data and the environment information are subjected to multi-scale feature extraction by the domain invariant feature extractor to obtain a feature vector.

[0024] The feature vector is subjected to adversarial training and fusion by the cross-modal adversarial fusion module to generate a fusion feature.

[0025] The mirror model is subjected to forward migration by the environment parameter prediction engine based on the fusion feature, and in a case where the simulation result deviates from the preset simulation condition, reverse migration is triggered to optimize the parameter prediction model to obtain new environment adjustment parameters, until the simulation result meets the preset simulation condition, and the control cabin is configured according to the new environment adjustment parameters.

[0026] According to one embodiment of the present application, the domain invariant feature extractor includes a ResNet-18 network and a gradient reversal layer connected in sequence.

[0027] The ResNet-18 network is configured to extract multi-scale features from the activity data and the environment information to obtain a feature vector.

[0028] The gradient reversal layer is configured to perform bidirectional state migration and feature alignment on the feature vector.

[0029] According to one embodiment of the present application, in a case where the simulation result deviates from the preset simulation condition, reverse migration is triggered to optimize the parameter prediction model, including:

[0030] In a case where the simulation result deviates from the preset simulation condition, an incremental memory replay mechanism is triggered to calculate a historical optimal parameter combination corresponding to the activity data and the environment information by replaying historical data in a core memory bank and an edge memory bank.

[0031] The parameter prediction model is subjected to elastic weight solidification by the optimal parameter combination, and the parameter prediction model is subjected to parameter update.

[0032] According to one embodiment of the present application, before the mirror model of the operator is called for simulation based on the environment adjustment parameter, the method further comprises:

[0033] The state transition model is pre-trained by source domain data;

[0034] The cross-modal fusion module is fine-tuned by target domain data, while the domain-invariant feature extractor and the environment parameter prediction engine are frozen;

[0035] The state transition model is fine-tuned by target domain data, while the domain-invariant feature extractor and the environment parameter prediction engine are frozen;

[0036] According to one embodiment of the present application, the activity data of the operator in the ship control cabin is obtained by:

[0037] The physiological data, psychological data and behavior data of the operator in the control cabin are obtained;

[0038] The signal quality index of the physiological data, psychological data and behavior data is calculated, and the physiological data, psychological data and behavior data are screened to obtain the activity data when the signal quality index is greater than a preset threshold;

[0039] The control cabin is configured with the environment adjustment parameter by an environment adjustment actuator, and after the control cabin is set by the environment adjustment parameter, the method further comprises:

[0040] The activity data of the operator is re-obtained;

[0041] The environment adjustment actuator is monitored by an actuator feedback verification loop, and the control cabin is adjusted by the environment adjustment actuator when the error between the new activity data and the simulation result is greater than an error threshold.

[0042] In a second aspect, the present application provides an environment adjustment device for a ship control cabin, comprising:

[0043] An acquisition module is configured to acquire activity data of an operator in a ship control cabin and environment information of the control cabin;

[0044] A first processing module is configured to extract working state information of the operator from the activity data;

[0045] The second processing module is configured to input the working state information and the environment information into a parameter prediction model to obtain an environment adjustment parameter output by the parameter prediction model, the parameter prediction model being obtained through a plurality of training samples, the training samples including sample working state information and sample environment information of a sample operator and an environment adjustment parameter label in a sample control cabin where the sample operator is located;

[0046] The third processing module is configured to simulate the operator through a mirror model based on the environment adjustment parameter, and set the control cabin through the environment adjustment parameter when a simulation result of the mirror model reaches a preset simulation condition, the mirror model being constructed based on physiological characteristics, psychological characteristics and behavior characteristics of the operator.

[0047] In a third aspect, the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the environment adjustment method of the ship control cabin according to the first aspect.

[0048] In a fourth aspect, the present application provides a non-transitory computer readable storage medium having a computer program stored thereon, and the computer program is executable by a processor to implement the environment adjustment method of the ship control cabin according to the first aspect.

[0049] In a fifth aspect, the present application provides a chip, including a processor and a communication interface, the communication interface and the processor being coupled, and the processor is configured to run a program or an instruction to implement the environment adjustment method of the ship control cabin according to the first aspect.

[0050] In a sixth aspect, the present application provides a computer program product, including a computer program, and the computer program is executable by a processor to implement the environment adjustment method of the ship control cabin according to the first aspect.

[0051] Additional aspects and advantages of the present application will be in part apparent and in part pointed out hereinafter.

[0052] The environment adjustment method, device and electronic device of the ship control cabin provided by the present application have the following beneficial effects compared with the prior art:

[0053] (1) By monitoring the environmental information and the activity data of the operator in real time, the fatigue and distraction of the operator are found in time, the environmental adjustment parameters are verified through the personalized mirror image model of the operator, the reaction of the operator in a specific environment is simulated, and the environmental adjustment parameters are adjusted in real time according to the physiological and psychological state, which can fully consider the needs of different operators to the environment, ensure that the operator can maintain the best working state under long time work, ensure that the environmental adjustment parameters can more accurately meet the needs of each operator, thereby improving the work efficiency and task completion quality of the operator, and the overall operation efficiency of the ship control cabin, further optimizing the environment of the control cabin, not only improving the working state of the operator, but also improving the safety and stability of the ship, preventing errors or accidents caused by the operator not adapting to the environment.

[0054] (2) Through bidirectional state migration, cross-modal fusion and dynamic parameter updating, combined with domain invariant feature extraction, cross-modal adversarial fusion and bidirectional migration mechanism of the environmental parameter prediction engine, the precision, adaptability and stability of the mirror image model in simulation are improved, and the real-time adjustment of the control cabin configuration is optimized. BRIEF DESCRIPTION OF DRAWINGS

[0055] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings, in which:

[0056] Figure 1 is a flowchart of an environmental adjustment method of a ship control cabin provided by an embodiment of the present application;

[0057] Figure 2 is a structural schematic diagram of a state migration model provided by an embodiment of the present application;

[0058] Figure 3 is a structural schematic diagram of an environmental adjustment device of a ship control cabin provided by an embodiment of the present application;

[0059] Figure 4 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0060] The technical solutions in the embodiments of the present application will be described clearly in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art belong to the scope of protection of the present application.

[0061] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0062] The following description, in conjunction with the accompanying drawings, details the environmental conditioning method, environmental conditioning device, electronic equipment, and readable storage medium for the ship control cabin provided in this application, through specific embodiments and application scenarios.

[0063] Among them, the environmental regulation method of the ship's control cabin can be applied to the terminal, and can be executed by the hardware or software in the terminal.

[0064] The terminal includes, but is not limited to, portable communication devices such as mobile phones or tablets with touch-sensitive surfaces (e.g., touchscreen displays and / or touchpads). It should also be understood that, in some embodiments, the terminal may not be a portable communication device, but rather a desktop computer with touch-sensitive surfaces (e.g., touchscreen displays and / or touchpads).

[0065] The following embodiments describe a terminal including a display and a touch-sensitive surface. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, mouse, and joystick.

[0066] The environmental adjustment method for the ship control cabin provided in this application embodiment can be executed by an electronic device or a functional module or entity in an electronic device that can realize the environmental adjustment method for the ship control cabin. The electronic devices mentioned in this application embodiment include, but are not limited to, mobile phones, tablets, computers, cameras and wearable devices. The environmental adjustment method for the ship control cabin provided in this application embodiment will be described below using an electronic device as the execution subject.

[0067] like Figure 1 As shown, the environmental conditioning method for the ship's control cabin includes:

[0068] Step 110: Obtain the activity data of the operators in the ship's control cabin and the environmental information of the control cabin;

[0069] Step 120: Extract the operator's work status information from the activity data;

[0070] Step 130, input the working state information and the environment information into a parameter prediction model to obtain an environment adjustment parameter output by the parameter prediction model, the parameter prediction model being obtained through a plurality of training samples, the training samples including sample working state information and sample environment information of a sample operator, and an environment adjustment parameter label in a sample control cabin where the sample operator is located;

[0071] Step 140, based on the environment adjustment parameter, calling a mirror model of the operator for simulation, and in the case that a simulation result of the mirror model reaches a preset simulation condition, setting the control cabin through the environment adjustment parameter, the mirror model being constructed based on physiological characteristics, psychological characteristics and behavior characteristics of the operator.

[0072] It can be understood that the ship control cabin is a closed working environment, and the operator needs to be active and work in the control cabin for a long time to realize the control of the ship. The environment adjustment parameters in the control cabin need to be adjusted to improve the state of the operator.

[0073] Activity data is physiological data collected by sensors on the operator, and behavior data collected by an operation recording system on the operator, and other parameters.

[0074] Among them, the physiological data includes the heart rate, body temperature, pupil tracking data and brain wave of the operator, which reflects the physiological state of the operator, such as fatigue, attention, etc. The operation behavior data includes operation frequency, device interaction trajectory, instruction response delay, etc. The behavior data of the operator is collected by the operation recording system, such as operation frequency, device interaction trajectory, instruction response delay, etc. These data reflect the behavior state and work efficiency of the operator.

[0075] The environment information is a vector composed of temperature, humidity, oxygen concentration, carbon dioxide concentration, air pressure, air flow condition and other environmental data in the control cabin. Through the environmental data, it can be identified whether the environment inside the control cabin is suitable, whether it is too hot, cold or humid, which affects the physiological and psychological state of the operator.

[0076] The working state information is a quantitative index extracted based on the activity data, including physiological indicators, psychological indicators and work indicators and other information related to the working state of the operator. For example, the fatigue degree is extracted by heart rate variability, the attention concentration degree is extracted by pupil tracking data, and the operation efficiency is extracted by instruction response delay.

[0077] Among them, the pupil tracking data is recorded by an ultra wide band (UWB) positioning system to record the operation trajectory, combined with the attention focus captured by an eye tracker.

[0078] The parameter prediction model is constructed based on a neural network model. In the training process, a large number of training samples are obtained, each sample data including sample operator optimal working state information, sample environment information, and corresponding environment adjustment parameter label. The training samples come from different operators and different situations of the ship control cabin. The training samples are sequentially input to the neural network model for training. The neural network model can calculate the most suitable environment adjustment parameter of the operator according to the input working state information and environment information.

[0079] The environment adjustment parameter is a combination of the set values of multiple devices in the control cabin. Through the cooperative control of multiple set values, the temperature, humidity, oxygen content, lighting intensity, air pressure, etc. in the operation cabin are adjusted. For example, the CO2 concentration linkage mechanism of the ship ventilation control is adopted to optimize the fresh air circulation efficiency, the temperature and humidity sub-domain control algorithm is integrated, the influence of air flow distribution on operation comfort is simulated by Fluent software, local environment data mutation is avoided, and invalid energy consumption is reduced.

[0080] The mirror image model is constructed by the physiological characteristics, psychological characteristics and behavior characteristics of the operator. The physiological characteristics include the change trend of the physiological data such as the body temperature and heart rate of the operator, the psychological characteristics include the emotional state and fatigue degree of the operator, which are evaluated by long-term sleep monitoring, and the behavior characteristics include the activity amount, rest mode, action mode and reaction speed of the operator. The mirror image model also considers individual differences of the operator, such as age, physical fitness, emotional tendency and other difference factors, which can more accurately simulate the reaction of each operator in different environments.

[0081] For example, the physiological data such as heart rate, blood pressure, blood oxygen saturation, body temperature of the operator are collected by wearable devices such as smart bracelets, health bracelets and the like and transmitted to the computing platform in real time. The computing platform efficiently processes and analyzes these data to obtain the physiological characteristics, psychological characteristics and behavior characteristics of the operator, models the operator, accurately maps the difference factors of the operator to the digital twin in the virtual control cabin through virtual mirror technology, and performs personalized training through physiological characteristics, psychological characteristics and behavior characteristics. The mirror image model obtained is used to simulate the state of the operator and the reaction to external stimuli to accurately reflect the physiological state and health condition of the operator. For example, in an emergency state, the operator may be more sensitive to changes in temperature and oxygen concentration, while in a normal working state, the reaction may be relatively stable.

[0082] In addition, a computational fluid dynamics (CFD) simulation module is integrated in the mirror model, which can accurately simulate physical phenomena such as fluid flow and heat transfer, providing real support for the physical laws of digital twins, and verifying the influence of environmental regulation parameters on human-computer interaction of operators in the control cabin through multi-physical field simulation.

[0083] The preset simulation condition is a dynamic threshold for judging the effectiveness of the environmental regulation parameter, for example, the improvement of work efficiency is more than 15%, or the fatigue degree is reduced to the safe interval, the temperature is within the comfort range of the operator, the humidity is moderate, etc. The cycle regulation logic is adopted, and the self-adaptive threshold adjustment mechanism is combined to adjust the preset simulation condition of the operator.

[0084] In addition, a 1:1 control cabin model is established in the Unity3D engine, and the mirror model loads the biomechanical properties of the operator in the control cabin model to simulate the environmental regulation parameters.

[0085] In step 110, a plurality of sensors are arranged in the ship control cabin, including temperature sensors, humidity sensors, air quality sensors, air pressure sensors, etc., to collect environmental data to obtain environmental information. A variety of biological sensors such as heart rate monitors, brain wave sensors, and pupil tracking devices are embedded in the console, seat, and helmet to monitor the physiological data of the operator's heart rate, skin conductance response, etc. The operation recording system collects the behavior data of the operator, and uses edge computing nodes for multi-source data space-time alignment. The combination of physiological data and behavior data is used as activity data.

[0086] In step 120, the working state information of the operator is extracted from the sensor data using machine learning or pattern recognition techniques by analyzing the activity data. For example, by analyzing heart rate and motion data, it can be inferred that the operator's work intensity is mild, moderate, or heavy. The physiological fluctuation entropy value is extracted from the activity data by wavelet transform to determine the fatigue degree. The operation trajectory complexity is extracted from the activity data by fractal tail number calculation to obtain the attention concentration and operation efficiency. The work intensity, fatigue degree, attention concentration, and operation efficiency are used as working state information.

[0087] In step 130, the extracted working state information and environmental information are input into the parameter prediction model, and the most suitable environmental regulation parameter for the operator is calculated and output by the trained parameter prediction model.

[0088] In step 140, the activity data and environmental data of the operator are input into the mirror model to simulate the operator's response under the environmental conditions corresponding to the environmental adjustment parameters, and to ensure that the operator's response meets the preset conditions. For example, when the temperature or humidity in the control cabin is too high, the operator's comfort level may decrease, and the mirror model can reflect this change. When the simulation results of the mirror model meet the preset simulation conditions, the control cabin is set using the environmental adjustment parameters to optimize the control cabin environment, including the adjustment of at least one of the air conditioning, ventilation system, humidification or dehumidification device, air purification system, lighting system, etc., thereby improving the operator's working state and the working efficiency of the ship control cabin, reducing the operator's fatigue and psychological pressure, and maintaining the operator's high alertness and working enthusiasm.

[0089] In addition, the control cabin also has a feedback mechanism. If the environmental conditions change suddenly or the physiological response of the operator does not meet the expectations, the environmental settings are automatically detected and corrected to ensure that the environment in the control cabin always meets the needs of the crew.

[0090] For example, when the operator's working state changes, such as entering a high-load state or a rest state, the activity data of the operator and the environmental information of the control cabin are reacquired, the parameter prediction model is called to recalculate the environmental adjustment parameters based on the new input data, and the mirror model is called to perform simulation to adjust the environment in the ship cabin.

[0091] To improve the environmental adjustment effect of the ship control cabin, the temperature and humidity disturbance data set of the low-pressure test cabin dynamic control is used to simulate the data disturbance under the ship pitching condition, and the adversarial training is performed to cover extreme conditions such as deep diving high-pressure environment. In addition, an inertial measurement unit (IMU) can be added to integrate the attitude compensation algorithm and consider the influence of ship pitching acceleration on the operation.

[0092] A floating anomaly correction mechanism can also be provided, which triggers the parameter locking of the control cabin when the simulated operation failure rate decreases by ≥20%.

[0093] According to the method for adjusting the environment of the ship control cabin provided in the embodiments of the present application, the fatigue and distraction of the operator can be found in time by monitoring the environmental information and the activity data of the operator in real time, the environmental adjustment parameters are verified through the personalized mirror model of the operator, the reaction of the operator in a specific environment is simulated, and the environmental adjustment parameters are adjusted in real time according to the physiological and psychological state, the needs of different operators for the environment can be fully considered, the operator can still maintain the best working state under long-time work, the environmental adjustment parameters can more accurately meet the needs of each operator, and thus the working efficiency and task completion quality of the operator and the overall operation efficiency of the ship control cabin are improved, the environment of the control cabin is further optimized, the working state of the operator is improved, and the safety and stability of the ship are improved, and the mistakes or accidents caused by the inadaptation of the operator to the environment are prevented.

[0094] In some embodiments, the step 120 of extracting the working state information of the operator from the activity data comprises:

[0095] aligning the activity data in scale;

[0096] mapping the aligned activity data to a target graph structure, wherein the nodes in the target graph structure are used to represent the data points in the activity data, and the edges between the nodes in the target graph structure are used to represent the adjacency relationship and similarity between the data points;

[0097] extracting features of the target graph structure through a graph attention network to obtain the working state information.

[0098] In the embodiments, the activity data comprises physiological data, psychological data and behavior data, the behavior data comprises electrocardiogram (ECG), electroencephalogram (EEG), heart rate, blood oxygen, body temperature, etc., the psychological data is used to reflect the data of the psychological state of the operator, such as emotional fluctuation, psychological stress, anxiety level, fatigue degree, etc., and the behavior data comprises the data related to the behavior, action and task execution of the operator, such as mouse click, keyboard input, eye movement track, posture change, operation response time, etc.

[0099] Since the sampling frequencies of the physiological data and the behavior data are different, the physiological data, the psychological data and the behavior data are aligned to the same time point, so that the alignment in time is realized.

[0100] The target graph structure is a graphical data structure used to represent the relationship between different data points, including nodes and edges between nodes, the nodes are used to represent data points, each data point in physiological data, psychological data and behavioral data can be mapped to a node, and the edges and weights are used to represent the adjacency relationship or similarity between the nodes, and the edges are constructed according to the correlation or distance measure between the physiological data, psychological data and behavioral data.

[0101] The physiological indicators are used to describe the characteristics of the operator's physiological state, which are obtained by key feature extraction of physiological data, the psychological indicators are used to describe the characteristics of the operator's psychological state, and the work efficiency is used to evaluate the quality and speed of the operator's task completion. Through the fusion of physiological and behavioral data, the relationship between the operator's work state and work efficiency is quantified.

[0102] In actual execution, the physiological data, psychological data and behavioral data are aligned in time scale by interpolation method. For time series data, edges between adjacent nodes are established according to time sequence, and similarity between nodes is measured by using Euclidean distance or cosine similarity, and corresponding weights are assigned to edges to obtain target graph structure. Graph attention network (GAT) uses self-attention mechanism to learn node features and dynamically adjusts the weights between adjacent nodes.

[0103] Based on time adjacency, similarity calculation or task-based behavior similarity, the edges in the target graph structure are represented by an adjacency matrix, and each element of the adjacency matrix represents the similarity or relationship strength between nodes.

[0104] In the graph attention network, each node dynamically calculates the weights of the neighbor nodes according to the features of the neighbor nodes through the self-attention mechanism, automatically identifies the neighbor nodes that have greater influence on the current node, and effectively extracts the features of the nodes; information is transmitted layer by layer through multiple graph convolutional layers to extract high-order node features, such as frequency domain features extracted from EEG signals or time features extracted from behavioral data.

[0105] Through the extracted features, the work state of the operator is evaluated, and indicators such as physiological load, psychological state and work efficiency are output as work state information.

[0106] In this embodiment, by constructing the target graph structure, the work state information of the operator can be effectively extracted from multi-dimensional data, which provides a basis for decision support, health management and work optimization.

[0107] In some embodiments, the activity data is scaled and aligned, including:

[0108] invoke the inference engine to infer the active data, determine the rule category and the processing agenda;

[0109] based on the rule category, invoke the corresponding processing rule from the rule library to match each type of data in the active data;

[0110] invoke the execution engine to process the corresponding active data based on the processing agenda through the processing rule, determine the sensitive data and non-sensitive data in the active data;

[0111] extract risk assessment features from the sensitive data according to the relevance of the sensitive data and the working state of the operator;

[0112] perform covariance matrix scaling on the non-sensitive data and the risk assessment features.

[0113] It can be understood that the inference engine is used to infer and analyze the active data, determine the rule category corresponding to each type of data of the active data, and generate the processing agenda of the active data in the subsequent classification order. The inference engine can be a rule-based inference engine or a machine learning-based inference engine.

[0114] The rule category is used to represent the behavior pattern of the operator, the change of the environmental parameter, etc. corresponding to each type of data, and the processing agenda is the processing step and process of the active data.

[0115] The rule library contains a variety of processing rules, which is a structured rule set, supports category matching and invocation of specific rules. The processing rule can be classified and defined according to the behavior of the operator, the environmental parameter, etc. to determine the sensitive data which has greater influence on the working state of the operator and other non-sensitive data.

[0116] For example, in the case where the data represents that the operator is in an unauthorized operation, long-term overtime work, etc. abnormal working state, it is determined as sensitive data.

[0117] The risk assessment features are used for subsequent risk analysis and decision-making.

[0118] In actual execution, the active data is input into the inference engine, and the inference engine infers the active data based on the pre-defined rules or trained model, determines the rule category and the processing agenda.

[0119] Based on the determined rule category, the corresponding processing rule is invoked from the rule library to match each type of data in the active data. For example, if the rule category is a certain specific operation behavior of the operator, the rule related to the operation behavior is invoked to match the corresponding data in the active data.

[0120] The execution engine is called to process the corresponding activity data according to the built-in data processing steps and processing logic in the processing agenda, and in the processing process, sensitive data and non-sensitive data are separated.

[0121] According to the correlation between the sensitive data and the working state of the operator, the sensitive data is extracted with a high correlation risk assessment feature. For example, if the sensitive data is the physiological data of the operator, the heart rate, blood pressure and other features can be extracted to assess whether the working state of the operator is normal.

[0122] The non-sensitive data and the risk assessment features are standardized, and the correlation and change degree between the non-sensitive data and the risk assessment features are measured through the calculation of the covariance matrix to realize scale alignment, so that different data can be compared and analyzed on the same scale to provide a basis for subsequent data processing and decision-making.

[0123] In this embodiment, by reasoning, classifying and scaling the activity data, the safe processing of sensitive data and the normal storage of non-sensitive data are ensured. At the same time, the risk assessment features and the covariance matrix alignment provide a basis for subsequent analysis and decision-making.

[0124] In some embodiments, the simulation of the mirror model is based on a state transition model, as shown in Figure 2 The state transition model includes a domain-invariant feature extractor, a cross-modal adversarial fusion module and an environment parameter prediction engine connected in sequence;

[0125] Step 140, based on the environment adjustment parameter, calling the mirror model of the operator for simulation, in the case that the simulation result of the mirror model reaches the preset simulation condition, setting the control cabin through the environment adjustment parameter, including:

[0126] Through the domain-invariant feature extractor, multi-scale feature extraction is performed on the activity data and the environment information to obtain a feature vector;

[0127] Through the cross-modal adversarial fusion module, the feature vector is subjected to adversarial training and fusion to generate a fusion feature;

[0128] Through the environment parameter prediction engine, the fusion feature is subjected to forward migration to the mirror model, and the mirror model is simulated based on the environment adjustment parameter. In the case that the simulation result deviates from the preset simulation condition, reverse migration is triggered to optimize the parameter prediction model to obtain new environment adjustment parameters, until the simulation result meets the preset simulation condition, and the control cabin is configured according to the new environment adjustment parameters.

[0129] The bidirectional state transition includes forward transition and reverse transition.

[0130] It can be understood that the state transition model is a physiological- psychological state transition model (PPSTM) which is trained on a large-scale cross-operator dataset and fine-tuned for each operator, and the obtained state transition model can reflect individual differences of operators.

[0131] In some embodiments, the domain-invariant feature extractor comprises a ResNet-18 network and a gradient reversal layer connected in sequence.

[0132] The ResNet-18 network is configured to perform multi-scale feature extraction on the activity data and the environmental information to obtain a feature vector.

[0133] The gradient reversal layer is configured to perform bidirectional state transition and feature alignment on the feature vector.

[0134] The domain-invariant feature extractor is constructed by adding a gradient reversal layer (GRL) to the output layer of the ResNet-18 on the basis of an adversarial discriminative domain adaptation (ADDA) framework.

[0135] The ResNet-18 network extracts multi-scale features from the input data through multiple convolutional layers and residual block structures to obtain a plurality of scale feature vectors, solves the problems of gradient vanishing and gradient explosion in deep networks while maintaining high performance, and enables the network to more effectively learn feature representations through residual learning and skip connections.

[0136] In the domain adaptation task, the gradient reversal layer (GRL) reverses the gradient sign during backpropagation, so that the domain-invariant feature extractor learns domain-invariant features, and the extracted feature vectors are more consistent between the source domain and the target domain. This mechanism makes it difficult for the domain classifier to distinguish between the features of the source domain and the target domain, thereby improving the generalization ability of the model in different domains.

[0137] In actual implementation, the domain-invariant feature extractor uses the ResNet-18 network to extract multi-scale features, eliminates the differences in data distribution between the laboratory (source domain) and the ship control cabin (target domain), and ensures feature alignment. For example, the physiological data of the operator and the environmental information are extracted to obtain time series and spatial features, respectively, and the maximum mean discrepancy (MMD) loss is used to constrain the consistency of the feature distribution.

[0138] The gradient reversal layer (GRL) is used for adversarial training to align the feature vectors of different modalities in the hidden space, and the attention mechanism is used to dynamically allocate weights. In the cross-modal adversarial fusion module, the adversarial training can effectively alleviate the modal gap, and the strong semantic correlation between the feature vectors of different modalities can improve the fusion effect. For example, the brain wave spectrum features are fused with the time sequence changes of the cabin CO2 concentration.

[0139] The environmental parameter prediction engine performs forward migration based on the fused features to predict the operator state after adjustment of the environmental parameters through the mirror model. In reverse migration, the parameter update amount of the parameter prediction model is back-propagated according to the deviation of the simulation results, forming a closed-loop optimization. For example, when the temperature in the operation cabin is raised and the operation delay rate exceeds a threshold (such as 5%), the optimal temperature control strategy is calculated through reverse migration.

[0140] In forward migration, the activity data and environmental information are mapped to the physiological and psychological state changes of the operator through the fused features, and the causal relationship between the state and the modeling parameters is modeled through the state transition matrix.

[0141] In reverse migration, the environmental inducement that leads to simulation deviation is traced back through counterfactual reasoning, and the weights of the parameter prediction model are updated. For example, the environmental inducement is the over-standard CO2 concentration leading to decision-making errors.

[0142] In this embodiment, through bidirectional state migration, cross-modal fusion and dynamic parameter update, combined with the bidirectional migration mechanism of the domain-invariant feature extraction, cross-modal adversarial fusion and environmental parameter prediction engine, the precision, adaptability and stability of the mirror model in simulation are improved, and the real-time adjustment of the control cabin configuration is optimized.

[0143] In some embodiments, the feature alignment loss function of the state migration model includes an adversarial alignment loss function, a statistical quantity alignment loss function and a semantic preservation loss function;

[0144] The adversarial alignment loss function L adv is:

[0145]

[0146] wherein x s and x t represent the input data of the source domain and the input data of the target domain respectively, D is a domain discriminator, and f is a feature extractor;

[0147] The statistical quantity alignment loss function L stat is:

[0148]

[0149] where MMD is the maximum mean difference, is a feature mapping function, C s and C t are the covariance matrices of the source domain and the target domain respectively, λ1 and λ2 are weight coefficients, and CORAL is the covariance alignment;

[0150] The semantic preservation loss function L task is:

[0151]

[0152] where CE is the cross-entropy loss, used for the source domain classification task; KL is the Kullback-Leibler divergence, used for the target domain knowledge distillation, y s and are the real label and the predicted label of the source domain respectively, p t and p teacher are the output probability distributions of the target domain model and the teacher model respectively.

[0153] In actual execution, the domain-invariant feature extractor is responsible for extracting features from the source domain and the target domain, and the feature alignment loss function matches the features of the source domain and the target domain through the corresponding alignment methods of the adversarial alignment loss function, the statistical quantity alignment loss function and the semantic preservation loss function, so that the extracted features can maintain consistency in cross-domain tasks. The feature alignment loss function helps the feature extractor to learn a feature representation that is effective for both the source domain and the target domain.

[0154] The cross-modal adversarial fusion module further optimizes feature fusion through an adversarial training method, so that the features of the source domain and the target domain can be effectively fused to facilitate subsequent environmental parameter prediction. The feature alignment loss function promotes the consistency of cross-modal features to a certain extent through the combination of adversarial alignment loss, statistical quantity alignment loss and semantic preservation loss, and further promotes the adversarial fusion of the source domain and the target domain.

[0155] The environmental parameter prediction engine uses the aligned features for final task prediction, and the aligned features ensure that the target domain data can also be effectively predicted in the environmental parameter prediction engine.

[0156] where the adversarial alignment loss function L adv maximizes the correct classification probability of the source domain features and the incorrect classification probability of the target domain features by the domain discriminator D, so as to promote the feature extractor to generate domain-invariant features. The gradient reversal layer (GRL) is introduced to dynamically confuse the judgment of the domain discriminator D;

[0157] The statistical quantity alignment loss function L statMaximum Mean Discrepancy (MMD) and Correlation Alignment (CORAL), MMD calculates the distance of distribution in reproducing kernel Hilbert space, CORAL aligns the second-order statistics of covariance matrix, which is used to align the first and second-order statistics of source domain and target domain features.

[0158] Semantic preservation loss function L task Jointly optimize the source domain classification task and the target domain knowledge distillation to ensure that the features extracted by the feature extractor can preserve the semantic information of the data and the consistency of the feature semantics.

[0159] The above three loss functions are jointly optimized to train the domain-invariant feature extractor, so that the feature extractor can extract domain-invariant and semantically consistent features by minimizing the total loss function.

[0160] In this embodiment, the effect of the feature alignment loss function is to optimize the domain-invariant feature extractor to ensure that the features of the source domain and the target domain can be aligned after passing through the feature extractor, and further optimize the feature alignment in the cross-modal adversarial fusion module. Finally, the environment parameter prediction engine provides effective and aligned features, which can achieve better feature representation alignment in cross-domain migration and improve the generalization ability of the state migration model in different domains.

[0161] In some embodiments, when the simulation result deviates from the preset simulation condition, the reverse migration is triggered to optimize the parameter prediction model, including:

[0162] When the simulation result deviates from the preset simulation condition, the incremental memory replay mechanism is triggered to calculate the historical optimal parameter combination corresponding to the activity data and the environment information by replaying the historical data in the core memory bank and the edge memory bank.

[0163] The optimal parameter combination is used to solidify the elastic weight of the parameter prediction model, and the parameter prediction model is updated.

[0164] Wherein, the update of the parameter prediction model is realized by the time series enhanced incremental learning algorithm, and a double memory replay mechanism is designed, including a core memory bank and an edge memory bank.

[0165] In the core memory bank, key samples are selected based on influence score (Influence Score):

[0166]

[0167] Wherein, IS(x i ) is the influence score of x iinfluence score of the i-th input sample; x i is the total number of samples in the dataset; represents the gradient of the loss function with respect to the model parameters θ on the input sample x i is the update direction of the model parameters, L(x i ) is the loss function on the input sample x i is the gradient related to x j , which represents the gradient of the loss function L(x j ) on the j-th sample with respect to the model parameters θ; is used to calculate the square of the gradient difference between x i and x j .

[0168] The most representative samples are retrieved in the core memory bank, and the information of these samples is used to adjust the model parameters. The core memory bank selects key samples through influence sampling (Influence Score) and stores 200 most representative samples.

[0169] In the construction of the edge memory bank, the edge memory bank saves the data of the last 24 hours, and is dynamically updated through the sliding window mechanism. If the Kullback-Leibler divergence of the new data is greater than the threshold value, the old data is triggered to be eliminated, and the model is adjusted according to these real-time data.

[0170] Through the combination of dynamic memory management and the advantages of edge computing, the balance between real-time response and continuous learning is achieved in a resource-constrained environment.

[0171] In actual execution, when the simulation result deviates from the preset condition, the state transition model triggers the incremental memory playback mechanism, and calculates the optimal parameter combination through the historical data stored in the core memory bank and the edge memory bank. The optimal parameter combination is used to represent the configuration of the parameter prediction model under the same activity data and environmental information conditions to achieve better performance.

[0172] In the time-enhanced incremental learning algorithm, data enhancement techniques are used to enhance the performance of time series data, considering the time dependence of historical data. Data smoothing, time series interpolation, time window, etc. can be used to enhance the processing ability of the model for time series data.

[0173] At each incremental learning, the samples in the playback memory bank are used to strengthen the most important part of reinforcement learning. Memory playback and current data can be combined to update the model at each time.

[0174] ​​The loss function calculation needs to consider not only the prediction error at the current moment, but also the influence of historical data playback. By combining the playback data and the current model prediction results, the loss function is optimized.

[0175] In the process of continuous optimization, the Elastic Weight Consolidation (EWC) algorithm is used to avoid catastrophic forgetting and maintain the stability of historical learning.

[0176] Elastic Weight Consolidation (EWC) loss function L total is:

[0177]

[0178] where L new is the current loss function, is the importance weight of the historical parameter updated for the kth time, θ k is the current parameter updated for the kth time, is the historical optimal parameter updated for the kth time, and λ is a hyperparameter.

[0179] EWC loss function L total is used to punish the case where the parameter changes too much, and the importance weight is calculated by combining the Fisher information matrix and the time decay factor. The Fisher information matrix reflects the influence of each parameter on the model output, and the time decay factor considers that long-term non-updated parameters need more punishment.

[0180] Through the parameter update process triggered by reverse transfer, the optimization goal is to minimize the bias and make the model adapt to new input data.

[0181] In this embodiment, the advantages of dynamic memory management and edge computing are combined to achieve the balance between real-time response and continuous learning in resource-constrained environments. When processing the edge memory bank, a sliding window mechanism and an active forgetting strategy are used to maintain the real-time performance of the system, and edge computing resources are used to quickly respond and update the model.

[0182] In some embodiments, before calling the mirror model of the operator for simulation based on the environmental adjustment parameter, the method further comprises:

[0183] Pre-train the state transition model through source domain data;

[0184] Freeze the domain-invariant feature extractor and the environmental parameter prediction engine, and fine-tune the cross-modal adversarial fusion module through target domain data;

[0185] The domain-invariant feature extractor and the environment parameter prediction engine are gradually unfrozen, and the state transition model is parameter optimized by target domain data.

[0186] In actual execution, the state transition model is pre-trained, the domain-invariant feature extractor extracts general features of source domain data, the cross-modal adversarial fusion module learns how to find shared features between different modalities, and the environment parameter prediction engine uses source domain data to preliminarily learn how to predict environment states and related parameters.

[0187] The domain-invariant feature extractor and the environment parameter prediction engine are frozen, and the cross-modal adversarial fusion module is optimized to make the model better perform feature fusion of different modalities under target domain data. It is ensured that the domain-invariant feature extractor and the environment parameter prediction engine of the state transition model are not over-adjusted, and only focus on improving the performance of the cross-modal adversarial fusion module.

[0188] The domain-invariant feature extractor and the environment parameter prediction engine are unfrozen from bottom to top. First, the low-level feature extraction part of the network is unfrozen, and then the high-level structure is gradually unfrozen. The state transition model can gradually adapt to new patterns and feature changes that may exist in target domain data; during the unfreezing process, the domain-invariant feature extractor, the cross-modal adversarial fusion module and the environment parameter prediction engine are updated gradually by training with target domain data, and the parameters and overall performance of the entire state transition model are optimized.

[0189] In addition, the parameters of each module in the state transition model can also be updated jointly by target domain data, so that the state transition model can better adapt to the task requirements of the target domain.

[0190] In this embodiment, by training the state transition model on the source domain data set, learning general features, freezing the domain-invariant feature extractor and the environment parameter prediction engine, and only fine-tuning the cross-modal adversarial fusion module, learning the adversarial fusion features between different modalities through adversarial training, the model can find effective common features between multiple modalities, promote the state transition process, and adapt to the individual needs of the target domain; in the joint optimization of the whole model, a curriculum learning strategy is adopted to gradually unfreeze the network layers, so that the model can adapt to the complexity of the target domain and the requirements of state transition.

[0191] In some embodiments, the activity data of the operator in the ship control cabin is obtained, including:

[0192] The physiological data, psychological data and behavior data of the operator in the control cabin are obtained;

[0193] The signal quality index is calculated for the physiological data, psychological data and behavior data, and the physiological data, psychological data and behavior data are screened when the signal quality index is greater than a preset threshold to obtain the activity data.

[0194] The control cabin is configured with the environmental regulation parameters by an environmental regulation actuator, and after the control cabin is set by the environmental regulation parameters, the method further comprises:

[0195] The activity data of the operator is reacquired.

[0196] The environmental regulation actuator is monitored by a feedback verification loop, and the parameters of the control cabin are adjusted by the environmental regulation actuator when it is determined that the error between the new activity data and the simulation result is greater than an error threshold.

[0197] The double redundancy verification is established by the signal quality index (SQI) real-time monitoring and the actuator feedback verification loop.

[0198] The SQI value is obtained by comprehensively considering the factors such as the validity of data, transmission delay and data fluctuation, and the integrity, accuracy and consistency of the physiological data, psychological data and behavior data are evaluated.

[0199] In order to ensure the matching between the environmental regulation parameters and the actual execution effect, the actuator feedback verification loop is set up to monitor the environmental regulation actuator in the control cabin in real time.

[0200] By comparing the new activity data actually fed back with the expected simulation result, it is verified whether the environmental regulation actuator adjusts the environment of the control cabin according to the set parameters. If the deviation between the activity data and the simulation result is large, the environment of the control cabin is automatically corrected, or an alarm mechanism is triggered, and enters into a safe mode to ensure that the working environment of the operator is stable and reliable.

[0201] In the case of the safe mode, the environmental regulation parameters in the control cabin are returned to a preset safe range, the environmental regulation operation is stopped, and guidance is provided to the operator, and the ship control cabin enters a stable state which does not harm the health of the operator or the safety of the ship.

[0202] In this embodiment, by checking data in stages and monitoring the environmental control actuators in the control cabin in real time, the efficient, safe, and stable operation of the ship's control cabin environmental control system is ensured. The environmental control method for the ship's control cabin provided in this application embodiment can be executed by the ship's control cabin environmental control device. This application embodiment uses the ship's control cabin environmental control device executing the environmental control method as an example to illustrate the ship's control cabin environmental control device provided in this application embodiment.

[0203] This application also provides an environmental control device for a ship's control cabin.

[0204] like Figure 3 As shown, the environmental control system in the ship's control compartment includes:

[0205] The acquisition module 310 is used to acquire the activity data of the operator in the ship's control cabin and the environmental information of the control cabin;

[0206] The first processing module 320 is used to extract the operator's work status information from the activity data;

[0207] The second processing module 330 is used to input the working status information and the environmental information into the parameter prediction model to obtain the environmental adjustment parameters output by the parameter prediction model. The parameter prediction model is obtained through multiple training samples. The training samples include the sample working status information and sample environmental information of the sample operator, as well as the environmental adjustment parameter labels in the sample control cabin where the sample operator is located.

[0208] The third processing module 340 is used to call the operator's mirror model for simulation based on the environmental adjustment parameters. When the simulation result of the mirror model meets the preset simulation conditions, the control cabin is set according to the environmental adjustment parameters. The mirror model is constructed based on the operator's physiological characteristics, psychological characteristics and behavioral characteristics.

[0209] According to the environment adjusting device of the ship control cabin provided in the embodiment of the present application, the fatigue and distraction of the operator can be found in time by monitoring the environmental information and the activity data of the operator in real time, the environmental adjusting parameters are verified by the personalized mirror model of the operator, the reaction of the operator in a specific environment is simulated, and the environmental adjusting parameters are adjusted in real time according to the physiological and psychological state, the needs of different operators for the environment can be fully considered, the operator can still maintain the best working state under long-time work, the environmental adjusting parameters can more accurately meet the needs of each operator, and therefore the working efficiency and the task completion quality of the operator and the overall operation efficiency of the ship control cabin are improved, the environment of the control cabin is further optimized, the working state of the operator is improved, and the safety and stability of the ship are improved, and the mistakes or accidents caused by the inadaptation of the operator to the environment are prevented.

[0210] The environment adjusting device of the ship control cabin in the embodiment of the present application can be an electronic device or a component in the electronic device, for example, an integrated circuit or a chip. The electronic device can be a terminal or other devices except the terminal. For example, the electronic device can be a tablet computer, a notebook computer, a palm computer, a mobile Internet device (MID), an ultra-mobile personal computer (UMPC), a server or a personal computer (PC), and the like, and the embodiment of the present application is not limited in this way.

[0211] The environment adjusting device of the ship control cabin in the embodiment of the present application can be a device with an operating system. The operating system can be an Android operating system, an iOS operating system or other possible operating systems, and the embodiment of the present application is not limited in this way.

[0212] The environment adjusting device of the ship control cabin provided in the embodiment of the present application can realize each process of the environment adjusting method of the ship control cabin in the above embodiment, and details are not described herein again to avoid repetition.

[0213] In some embodiments, as shown in Figure 4 The embodiment of the present application further provides an electronic device 400, which includes a processor 401, a memory 402 and a computer program stored in the memory 402 and executable on the processor 401. When the program is executed by the processor 401, each process of the above environment adjusting method of the ship control cabin is realized, and the same technical effects can be achieved, and details are not described herein again to avoid repetition.

[0214] It should be noted that the electronic device in the embodiments of the present application includes the mobile electronic device and the non-mobile electronic device described above.

[0215] The embodiments of the present application further provide a non-transitory computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement each process of the method for adjusting environment of a ship control cabin and achieve the same technical effects. To avoid repetition, details are not described herein.

[0216] The processor is the processor in the electronic device in the above embodiments. The readable storage medium includes a computer readable storage medium, such as a computer readable only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a magnetic disk or an optical disk, etc.

[0217] The embodiments of the present application further provide a computer program product, which includes a computer program. The computer program is executed by a processor to implement the method for adjusting environment of a ship control cabin.

[0218] The processor is the processor in the electronic device in the above embodiments. The readable storage medium includes a computer readable storage medium, such as a computer readable only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a magnetic disk or an optical disk, etc.

[0219] The embodiments of the present application further provide a chip, which includes a processor and a communication interface. The communication interface is coupled with the processor. The processor is configured to run a program or an instruction to implement each process of the method for adjusting environment of a ship control cabin and achieve the same technical effects. To avoid repetition, details are not described herein.

[0220] It should be understood that the chip mentioned in the embodiments of the present application can also be referred to as a system level chip, a system chip, a chip system or a system on chip, etc.

[0221] It should be noted that, in the present document, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. An element proceeded by "comprises a", "comprising", or "includes a", does not, without more constraints, preclude the existence of additional identical elements in the process, method, article, or apparatus that comprises the element. Additionally, it should be noted that the scope of the methods and apparatus of the present embodiments are not limited by the order of the steps or the sequence for performing the steps, as some steps can occur simultaneously, in other steps can occur sequentially, or in between other steps can occur, as can be readily appreciated by those skilled in the art. Further, the features of the various examples described can be combined, in other examples.

[0222] From the above description of the embodiments, it is apparent that the above-described method of the embodiments can be realized by means of software and the necessary universal hardware platform, of course, also by hardware, but in many cases the former is the better embodiment. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer software product stored in a storage medium (such as ROM / RAM, magnetic disc, optical disc), including a number of instructions for causing a terminal (which can be a mobile phone, computer, server, or network device, etc.) to execute the environmental regulation method of the ship control cabin.

[0223] In the description of the present application, "first feature" and "second feature" can include one or more of the features.

[0224] In the description of the present application, "a plurality of" means two or more.

[0225] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the specific embodiments described above, which are merely illustrative and not restrictive, and those of ordinary skill in the art can make many forms under the inspiration of the present application without departing from the scope of the present application and the scope protected by the claims, which all belong to the protection of the present application.

[0226] In the description of the application, reference has been made to descriptive terms such as "one embodiment", "some embodiments", "an embodiment", "example", "specific example" or "some examples" etc. It is emphasized that each of these terms refers to a specific feature, structure, material or characteristic described in connection with a particular embodiment or example. The descriptive terms are not necessarily referring to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0227] While embodiments of the application have been shown and described, it is to be understood that the application is not limited to the details of the embodiments described, since the scope of the application is defined with respect to the appended claims.

Claims

1. A method for environmental regulation in a ship's control cabin, characterized in that, include: Acquire activity data of operators in the ship's control cabin and environmental information of the control cabin; Extract the operator's work status information from the activity data; The working status information and the environmental information are input into the parameter prediction model to obtain the environmental adjustment parameters output by the parameter prediction model. The parameter prediction model is obtained through multiple training samples. The training samples include the sample working status information and sample environmental information of the sample operator, as well as the environmental adjustment parameter labels in the sample control cabin where the sample operator is located. Based on the environmental adjustment parameters, a simulation is performed using the operator's mirror model. When the simulation results of the mirror model meet the preset simulation conditions, the control cabin is set using the environmental adjustment parameters. The mirror model is constructed based on the operator's physiological, psychological, and behavioral characteristics. The simulation of the mirror model is based on a state transition model, which includes a domain-invariant feature extractor, a cross-modal adversarial fusion module, and an environmental parameter prediction engine connected in sequence. The step of simulating the operation by calling a mirror model of the operator based on the environmental adjustment parameters, and setting the control cabin according to the environmental adjustment parameters when the simulation results of the mirror model meet the preset simulation conditions, includes: The domain-invariant feature extractor is used to perform multi-scale feature extraction on the activity data and the environmental information to obtain a feature vector. The cross-modal adversarial fusion module performs adversarial training and fusion on the feature vectors to generate fused features. The environmental parameter prediction engine forward transfers the fused features to the mirror model, performs simulation on the mirror model based on the environmental adjustment parameters, and triggers reverse transfer when the simulation results deviate from the preset simulation conditions to optimize the parameter prediction model and obtain new environmental adjustment parameters until the simulation results meet the preset simulation conditions. The control cabin is then configured according to the new environmental adjustment parameters.

2. The environmental conditioning method for the ship's control cabin according to claim 1, characterized in that, The step of extracting the operator's work status information from the activity data includes: The activity data is scale-aligned. The aligned activity data is mapped to a target graph structure, where nodes in the target graph structure represent data points in the activity data, and edges between nodes in the target graph structure represent adjacency relationships and similarities between data points. The working state information is obtained by extracting features from the target graph structure using a graph attention network.

3. The environmental conditioning method for the ship's control cabin according to claim 2, characterized in that, Scale alignment of the activity data includes: The inference engine is invoked to reason about the active data, determine the rule categories and processing agenda; Based on the rule category, the corresponding processing rule is called from the rule base to match each type of data in the activity data; The execution engine is invoked to process the corresponding activity data according to the processing agenda and the processing rules, and sensitive and non-sensitive data are identified in the activity data. Based on the correlation between the sensitive data and the operator's work status, risk assessment features are extracted from the sensitive data; The non-sensitive data and the risk assessment features are scaled together using their covariance matrices.

4. The environmental conditioning method for the ship's control cabin according to claim 1, characterized in that, The domain-invariant feature extractor comprises a ResNet-18 network and a gradient inversion layer connected in sequence; The ResNet-18 network is used to perform multi-scale feature extraction on the activity data and environmental information to obtain feature vectors; The gradient inversion layer is used to perform bidirectional state transition and feature alignment on the feature vector.

5. The environmental conditioning method for the ship's control cabin according to claim 1, characterized in that, The step of triggering reverse migration to optimize the parameter prediction model when the simulation results deviate from the preset simulation conditions includes: If the simulation results deviate from the preset simulation conditions, the incremental memory playback mechanism is triggered. By replaying historical data in the core memory bank and the edge memory bank, the historical optimal parameter combination corresponding to the activity data and the environmental information is calculated. The optimal parameter combination is used to solidify the elastic weights of the parameter prediction model and update the parameters of the parameter prediction model.

6. The environmental conditioning method for the ship's control cabin according to claim 1, characterized in that, Before performing simulation by calling the operator's mirror model based on the environmental adjustment parameters, the method further includes: The state transition model is pre-trained using source domain data; Freeze the domain-invariant feature extractor and the environmental parameter prediction engine, and fine-tune the parameters of the cross-modal adversarial fusion module using target domain data; The domain-invariant feature extractor and the environmental parameter prediction engine are gradually unfrozen, and the parameters of the state transition model are optimized using target domain data.

7. The environmental conditioning method for a ship's control cabin according to any one of claims 1-6, characterized in that, The acquisition of operator activity data in the ship's control room includes: Acquire physiological, psychological, and behavioral data of the operators in the control cabin; A signal quality index is calculated for the physiological data, psychological data, and behavioral data. If the signal quality index is greater than a preset threshold, the physiological data, psychological data, and behavioral data are filtered to obtain the activity data. The control cabin configures the environmental control parameters via an environmental control actuator. After setting the control cabin using the environmental control parameters, the method further includes: Reacquire the operator's activity data; By using an actuator feedback verification loop, the environmental control actuator is monitored. If the error between the new activity data and the simulation result is greater than the error threshold, the parameters of the control cabin are adjusted by the environmental control actuator.

8. An environmental control device for a ship's control cabin, using the method as described in any one of claims 1-7, characterized in that, include: The acquisition module is used to acquire the activity data of the operators in the ship's control cabin and the environmental information of the control cabin; The first processing module is used to extract the operator's work status information from the activity data; The second processing module is used to input the working status information and the environmental information into the parameter prediction model to obtain the environmental adjustment parameters output by the parameter prediction model. The parameter prediction model is obtained through multiple training samples. The training samples include the sample working status information and sample environmental information of the sample operator, as well as the environmental adjustment parameter labels in the sample control cabin where the sample operator is located. The third processing module is used to call the operator's mirror model for simulation based on the environmental adjustment parameters. When the simulation result of the mirror model meets the preset simulation conditions, the control cabin is set according to the environmental adjustment parameters. The mirror model is constructed based on the operator's physiological, psychological and behavioral characteristics.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the environmental regulation method for the ship control cabin as described in any one of claims 1-7.

Images