Vehicle cabin environment control method and device, computer device and storage medium

Abstract

The present disclosure relates to a vehicle cabin environment control method, device, computer device and storage medium. Involve vehicle cabin air treatment technology, solve the problems of scene adaptation limitation and poor adjustment effect. The method comprises: acquiring physiological state information and environment information of a driver and a passenger; performing feature fusion on the physiological state information and the environment information to obtain a thermal comfort coefficient; generating an air conditioner control instruction according to the thermal comfort coefficient, wherein the air conditioner control instruction comprises a temperature control parameter of an automobile air conditioner. The technical scheme provided by the present disclosure is suitable for a vehicle driving and riding scene, and realizes intelligent and accurate matching of air conditioner control requirements of passengers.

Description

Technical Field

[0001] This disclosure relates to vehicle cabin air treatment technology, and more particularly to a vehicle cabin environment control method, apparatus, computer device, and storage medium. Background Technology

[0002] Automotive air conditioning systems mostly rely on environmental parameters such as temperature, humidity, and sunlight intensity inside and outside the cabin for automatic adjustment. Their control logic is relatively simple and is essentially a coarse-grained temperature maintenance strategy based on limited environmental feedback. It can only achieve rough control of the temperature inside the vehicle and is difficult to accurately respond to the actual thermal comfort needs of the occupants.

[0003] With the continuous development of smart cockpit technology, some existing solutions have introduced occupant identification functions, which can make preliminary adaptive adjustments to the air conditioning based on preset personalized parameters or basic environmental information. However, the composition of the occupants, their status, and the environment inside and outside the vehicle are constantly changing. Relying solely on static preset parameters or limited environmental feedback information cannot perceive changes in the thermal perception status of occupants in real time and comprehensively, thus making it difficult to meet the comfort needs of diverse travel scenarios. Summary of the Invention

[0004] To overcome the problems existing in related technologies, this disclosure provides a method, apparatus, computer device, and storage medium for controlling the in-vehicle environment. By acquiring physiological and environmental information of the occupants and performing feature fusion to obtain a thermal comfort coefficient, an air conditioning control command containing temperature control parameters is generated. This allows for real-time perception of the dynamic thermal perception state of the occupants and environmental changes, enabling the air conditioning control command to accurately match the real-time needs of the occupants. This solves the problems of limited scene adaptation and poor adjustment effect, significantly improving the intelligence level and driving comfort of the vehicle air conditioning system.

[0005] According to a first aspect of the embodiments of this disclosure, a method for controlling the environment inside a vehicle cabin is provided, comprising: Obtain physiological and environmental information of drivers and passengers; The physiological state information and the environmental information are fused to obtain the thermal comfort coefficient; Based on the thermal comfort coefficient, an air conditioning control command is generated, which includes the temperature control parameters of the vehicle's air conditioning system.

[0006] Furthermore, the physiological state information includes at least one or more of the following multi-source data: Individual characteristic data, numerical physiological data, image feature data, quantitative auxiliary data, and audio data. The environmental information includes at least one or more of the following data: Vehicle interior and exterior environmental data, third-party environmental data, The steps for obtaining the physiological state information and environmental information of the driver and passengers include: The physiological state information is extracted using in-vehicle sensor arrays and a multimodal large model; The environmental information is extracted through the in-vehicle sensor group and communication module.

[0007] Furthermore, the step of fusing features of the physiological state information and the environmental information to obtain the thermal comfort coefficient includes: The physiological state information and the environmental information are preprocessed and quantized to obtain the raw data vector; Feature extraction is performed on the original data vector to obtain a first feature vector; The first feature vector is weighted to obtain the second feature vector; The second feature vector is mapped to the thermal comfort coefficient.

[0008] Furthermore, before the step of extracting features from the original data vector to obtain the first feature vector, the method further includes: The original data vector is standardized to unify the data dimensions.

[0009] Furthermore, the step of extracting features from the original data vector to obtain the first feature vector includes: The temporal and correlation features of the original data vector are extracted by a bidirectional gated loop unit, and the first feature vector is output.

[0010] Furthermore, the step of weighting the first feature vector to obtain the second feature vector includes: Based on the channel attention mechanism, the first feature vector is weighted to obtain the weighted second feature vector.

[0011] Furthermore, the step of mapping the second feature vector to the thermal comfort coefficient includes: A multilayer perceptron is used to map the second feature vector to the thermal comfort coefficient, which is a quantitative value obtained after quantizing the second feature vector and used to characterize the temperature perception state of the driver and passengers.

[0012] Furthermore, the step of generating air conditioning control commands based on the thermal comfort coefficient includes: Calculate the overall comfort deviation based on the thermal comfort coefficient; The temperature control parameters of the car air conditioner are dynamically adjusted based on the current environmental information. The air conditioning control command is generated based on the corrected temperature control parameters.

[0013] Furthermore, the step of calculating the overall comfort deviation based on the thermal comfort coefficient includes: The overall comfort deviation is calculated using the following expression: , in, The overall comfort deviation is... For facial comfort deviation, This is the facial weighting coefficient. For back comfort deviation, The back weighting coefficient, For the comfort deviation of the buttocks, This is the weighting factor for the buttocks. For foot comfort deviation, This represents the foot weighting coefficient.

[0014] Furthermore, the temperature control parameters include at least the target set temperature of the air conditioner, and the step of dynamically correcting the temperature control parameters of the car air conditioner based on the current environmental information includes: The target set temperature of the air conditioner is calculated according to the following expression: , in, Set the target temperature for the air conditioner. To collect ambient temperature inside the vehicle in real time, This is the temperature regulation sensitivity coefficient.

[0015] Furthermore, the temperature control parameters include the target set temperature of the air conditioner and / or the airflow setting information of each air outlet. The step of generating the air conditioner control command based on the corrected temperature control parameters includes: Obtain the airflow setting information for each air outlet; The air conditioning control command is generated based on the target set temperature and the air volume setting information.

[0016] Furthermore, the method also includes: The system sends air conditioning control commands to the air conditioning control module and obtains the feedback operating parameters.

[0017] According to a second aspect of the embodiments of this disclosure, a vehicle cabin environment control device is provided, comprising: The information collection module is used to acquire physiological and environmental information of drivers and passengers; The feature fusion module is used to perform feature fusion on the physiological state information and the environmental information to obtain the thermal comfort coefficient. The instruction generation module is used to generate air conditioning control instructions based on the thermal comfort coefficient, the air conditioning control instructions including the temperature control parameters of the car air conditioning.

[0018] According to a third aspect of the embodiments of this disclosure, a computer apparatus is provided, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to execute the vehicle cabin environment control method of the embodiments of this disclosure.

[0019] According to a fourth aspect of the embodiments of the present disclosure, a non-volatile computer-readable storage medium is provided, wherein when instructions in the storage medium are executed by a computer processor, the computer is able to perform the vehicle cabin environment control method provided by the embodiments of the present disclosure.

[0020] The technical solutions provided by the embodiments of this disclosure can include the following beneficial effects: acquiring physiological state information and environmental information of drivers and passengers, performing feature fusion on the physiological state information and the environmental information to obtain a thermal comfort coefficient, and then generating air conditioning control commands based on the thermal comfort coefficient, wherein the air conditioning control commands include temperature control parameters of the vehicle air conditioning. This enables real-time perception of the dynamic thermal perception state and environmental changes of drivers and passengers, allowing the air conditioning control commands to accurately match the real-time needs of the occupants, solving the problems of limited scene adaptation and poor adjustment effect, and significantly improving the intelligence level and driving comfort of the vehicle air conditioning.

[0021] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0022] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0023] Figure 1 This is a flowchart illustrating a method for controlling the environment inside a vehicle cabin according to an exemplary embodiment.

[0024] Figure 2 This is a schematic diagram of the structure of a multi-source data fusion neural network according to an exemplary embodiment.

[0025] Figure 3 This is a flowchart illustrating yet another method for controlling the environment inside a vehicle cabin, according to an exemplary embodiment.

[0026] Figure 4 This is a flowchart illustrating yet another method for controlling the environment inside a vehicle cabin, according to an exemplary embodiment.

[0027] Figure 5 This is a flowchart illustrating yet another method for controlling the environment inside a vehicle cabin, according to an exemplary embodiment.

[0028] Figure 6 This is a block diagram illustrating a vehicle cabin environment control device according to an exemplary embodiment. Detailed Implementation

[0029] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0030] With the continuous development of smart cockpit technology, some solutions have introduced occupant identification functions, which can make preliminary adaptive adjustments to the air conditioning based on preset personalized parameters or basic environmental information. However, the composition of passengers, their status, and the environment inside and outside the vehicle are constantly changing. Relying solely on static preset parameters or limited environmental feedback information cannot perceive changes in the thermal perception status of passengers in real time and comprehensively, thus making it difficult to meet the comfort needs of changing travel scenarios.

[0031] To address the aforementioned issues, embodiments of this disclosure provide a vehicle cabin environment control method, apparatus, computer device, and storage medium. By acquiring physiological and environmental information of occupants and performing feature fusion to obtain a thermal comfort coefficient, an air conditioning control command containing temperature control parameters is generated. This method enables real-time perception of the dynamic thermal state of occupants and environmental changes, allowing the air conditioning control command to accurately match the real-time needs of the occupants. It solves the problems of limited scene adaptation and poor adjustment effects, significantly improving the intelligence level and driving comfort of the vehicle air conditioning system.

[0032] An exemplary embodiment of this disclosure provides a method for controlling the in-vehicle environment. This method can intelligently generate air conditioning control commands based on the real-time physiological state of the occupants and environmental information inside and outside the vehicle, enabling adaptive and personalized control of the vehicle's air conditioning system, improving driving comfort and optimizing energy consumption. The specific process is as follows: Figure 1 As shown, it includes: Step 101: Obtain physiological and environmental information of drivers and passengers.

[0033] In this step, multi-source heterogeneous data related to the thermal comfort of drivers and passengers are comprehensively collected through various sensors, cameras, microphones, and external data interfaces, providing sufficient information support for subsequent feature fusion and decision-making.

[0034] The physiological state information includes at least one or more of the following multi-source data: Individual characteristic data, numerical physiological data, image feature data, quantitative auxiliary data, and audio data. The environmental information includes at least one or more of the following data: Vehicle interior and exterior environmental data, third-party environmental data, This step specifically includes: 1. Extract the physiological state information using in-vehicle sensor arrays and multimodal large model.

[0035] According to one exemplary embodiment, the in-vehicle sensor group includes an in-vehicle camera, an infrared sensor, a microphone, a health detection module, etc. The in-vehicle camera includes an AVM (Around View Monitor) camera, an OMS (Occupant Monitoring System) camera, etc. A multimodal large model refers to a pre-trained deep learning model capable of simultaneously processing multiple modalities of data such as images, text, and speech. In this embodiment, it is used to extract individual characteristic data of the driver and passengers, such as age, gender, and ethnicity, as well as image feature data such as sweating status, clothing coverage, and facial expressions, from camera images.

[0036] Specifically, the system acquires image data of the driver and passengers before they get into the vehicle using an AVM camera, and image data of the driver and passengers after they get into the vehicle using an OMS camera. This image data is then input into a pre-trained multimodal large model and visual algorithms (such as the YOLOv8-nano object detection algorithm) to extract the age, gender, and ethnicity of the driver and passengers. Furthermore, based on this information and the clothing features in the images, the system obtains the clothing index of the driver and passengers (for example, quantifying the thickness of the clothing into a value between 0 and 1, with a larger value indicating thicker clothing).

[0037] For numerical physiological data, infrared sensors are used to collect the surface temperature of drivers and passengers. Data collection sites include key areas such as the forehead, neck, and hands (number of sites N≥3). The surface temperature values ​​are normalized to the 0-1 range, forming a 1×N vector. Simultaneously, the infrared sensors also collect the ambient temperature gradient around the drivers and passengers to reflect the temperature distribution of the microenvironment surrounding the human body.

[0038] Regarding image feature data, a multimodal large model is used to identify images acquired by the vehicle-mounted OMS camera, extracting the sweating status of drivers and passengers (e.g., categorized as no sweat, slight sweat, significant sweating, etc.), clothing coverage (e.g., the proportion of body area covered by clothing), and facial expressions (e.g., whether they exhibit expressions of heat or cold). Furthermore, the vehicle-mounted health monitoring module analyzes the OMS video stream in real time, employing photoplethysmography (PPG) or similar non-contact physiological signal extraction techniques to obtain physiological indicators such as respiratory rate, heart rate, blood pressure, blood oxygen saturation, and heart rate variability, and then normalizes these indicators. For example, the normalization rule for respiratory rate is: below 20 breaths / minute is mapped to 0, 20–30 breaths / minute to 0.25, 30–40 breaths / minute to 0.5, and 40–60 breaths / minute to 1.

[0039] For audio data, audio data from drivers and passengers is collected via in-vehicle microphones. A lightweight audio classification model (such as a model based on MobileNet or a similar lightweight convolutional neural network structure) is used to identify the audio, determine whether the drivers and passengers are coughing or sneezing, and count the frequency. The frequency normalization rule is: more than 10 times per minute is mapped to 1, and less than 2 times per minute is mapped to 0.

[0040] In terms of quantifying auxiliary data, this includes individual feature encoding, health status encoding, and temperature satisfaction values ​​output by the large language model. Individual feature encoding encodes features such as gender, age, and body type, forming a 1×8 vector; health status encoding assigns values ​​based on the health status of the driver and passengers: healthy = 0, suspected cold = 0.5, cold = 1, with a dimension of 1×1; temperature satisfaction values ​​are generated by the large language model based on the driver and passengers' voice interactions or historical behavior data, normalized to the 0-1 range, with a dimension of 1×1. 2. The environmental information is extracted through the in-vehicle sensor group and communication module.

[0041] 2. The environmental information is extracted through the in-vehicle sensor group and communication module.

[0042] According to one exemplary embodiment, the in-vehicle sensor group includes in-vehicle and out-of-vehicle temperature sensors, humidity sensors, radiation temperature sensors, solar radiation intensity sensors, etc. The communication module includes an on-board communication unit for accessing third-party weather service APIs via a wireless network.

[0043] Specifically, the system collects the interior ambient temperature via an in-vehicle temperature sensor and the interior humidity via an in-vehicle humidity sensor; it also collects the exterior ambient temperature and humidity via exterior temperature and humidity sensors; it collects the interior radiant temperature (i.e., the thermal radiation effect of surrounding object surface temperatures on the human body) via a radiant temperature sensor; and it collects the interior solar radiation intensity via a solar radiation intensity sensor. All of these environmental data are normalized to the 0-1 range, forming a 1×4 vector. Simultaneously, the system obtains real-time weather conditions, outdoor temperature, and air quality information for the vehicle's location via a weather API, serving as supplementary data for external environmental correction.

[0044] For example, all the collected physiological and environmental information is preprocessed and then concatenated into a 1×141-dimensional raw data vector, which serves as the input for subsequent feature fusion.

[0045] Step 102: Perform feature fusion on the physiological state information and the environmental information to obtain the thermal comfort coefficient.

[0046] In this step, the multi-source heterogeneous, high-dimensional raw data obtained in step 101 is deeply fused using a designed neural network model to extract a quantitative indicator that represents the current thermal comfort feeling of the driver and passengers. The thermal comfort coefficient is a value ranging from 0 to 10 points. The higher the score, the hotter the driver and passengers feel, and the lower the score, the colder they feel. The range of 5 to 6 points represents a comfortable and suitable temperature.

[0047] According to one exemplary implementation, the spliced ​​1×141-dimensional original data vector is input into a pre-built multi-source data fusion neural network.

[0048] The multi-source data fusion neural network comprises an input layer, a feature extraction layer, an attention fusion layer, and an output layer. The feature extraction layer and the attention fusion layer can form multiple hidden layers. Figure 2 As shown, the functions of each layer are as follows: 1. The input layer performs secondary standardization on the raw data to eliminate the influence of dimensions.

[0049] 2. The feature extraction layer employs a 64-unit bidirectional gated recurrent unit (Bi-GRU) to extract temporal and correlation features of the data. For example, features such as "increased body surface temperature, increased sweating and decreased temperature satisfaction", or features such as "cold state, slightly higher body surface temperature and lower temperature satisfaction".

[0050] 3. The attention fusion layer introduces a channel attention mechanism (SE-Net) to assign weights to the 128-dimensional feature vector output by the feature extraction layer, highlighting core influencing factors (such as body surface temperature, sweating status, etc.) and suppressing irrelevant features to obtain a weighted fused feature vector.

[0051] 4. The output layer adopts a multilayer perceptron (MLP), which contains three fully connected layers (the first layer is 128→64, ReLU activation; the second layer is 64→32, ReLU activation; the third layer is 32→1, Sigmoid activation). The fused feature vector is mapped to a value in the range of 0 to 1, and then the value is mapped to the range of 0 to 10 to obtain the thermal comfort coefficient.

[0052] Step 103: Generate air conditioning control commands based on the thermal comfort coefficient.

[0053] The air conditioning control commands include the temperature control parameters of the car's air conditioning system.

[0054] In this step, the thermal comfort coefficient is converted into a specific air conditioning control strategy, generating control commands that can be executed by the vehicle air conditioning system to achieve precise adjustment of temperature, air volume, and air direction.

[0055] For example, first use the center value of the optimal range for human comfort. Based on this, the comfort deviation of each body part (face, back, buttocks, and feet) is calculated, and then weighted and fused using preset physiological weight coefficients to obtain the overall comfort deviation across the entire body. .

[0056] The physiological weighting coefficient reflects the sensitivity of different parts of the human body to temperature changes. Specifically: Facial weight (Most sensitive), back weight Buttock weight Foot weight (least sensitive).

[0057] The formulas for calculating the comfort deviation of each part are as follows: , Where k = face, back, hip, foot. This represents the thermal comfort coefficient for the corresponding part.

[0058] The formula for calculating the overall comfort deviation is:

[0059] Then, combined with the real-time ambient temperature collected inside the vehicle The target set temperature of the air conditioner is calculated according to the following expression: , in, Set the target temperature for the air conditioner. To collect ambient temperature inside the vehicle in real time, This is the temperature regulation sensitivity coefficient. In this embodiment of the disclosure, it is taken as... It should be noted that, (These are vehicle calibration constants, which can be fine-tuned according to different vehicle models).

[0060] Simultaneously, based on the comfort deviations of different areas, the airflow distribution for each air outlet was determined. The airflow for the facial air outlet was set to... The airflow of the air outlet is set to... The airflow of the seat ventilation vents is set to... The airflow of the foot-blowing vent is set to... .in, This is the maximum airflow. In addition, the airflow direction can be dynamically adjusted based on the positive or negative value of the comfort deviation of each part (positive indicates hotter, negative indicates colder), so that the cold or hot air is directed at the hotter parts first and avoids the colder parts.

[0061] Finally, based on the calculated target set temperature and airflow settings of each air outlet, an air conditioning control command is generated. The command includes the temperature set value, airflow level of each air outlet, airflow direction adjustment parameters, etc., and is sent to the actuator of the vehicle air conditioning system via the CAN bus.

[0062] An exemplary embodiment of this disclosure provides yet another method for controlling the in-vehicle environment, the process of obtaining the thermal comfort coefficient using this method being as follows: Figure 3 As shown, it includes: Step 301: Preprocess and quantize the physiological state information and the environmental information to obtain the original data vector.

[0063] In this step, the various heterogeneous data (such as image features, numerical sensor data, audio classification results, coded data, etc.) collected in step 101 are preprocessed and quantized in a unified manner so that they can be input into the neural network in the form of numerical vectors.

[0064] For example, individual characteristic data (age, gender, ethnicity) are encoded using one-hot encoding to form a 1×8 vector; numerical physiological data (body surface temperature, respiratory rate, etc.) are scaled according to preset normalization rules; image feature data (sweating status, clothing coverage, facial expressions, etc.) are extracted into feature vectors using a multimodal large model; environmental data (temperature, humidity, radiation temperature, solar radiation intensity) are normalized to the 0-1 range; audio data (coughing / sneezing frequency) is normalized; and quantitative auxiliary data (health status coding, temperature satisfaction) are assigned values ​​according to predetermined rules. All data are concatenated into a 1×141-dimensional original data vector.

[0065] Step 302: Standardize the original data vector to unify the data dimensions.

[0066] In this step, the differences in dimensions and numerical ranges between different features are eliminated, so that each feature has a similar scale during the training and inference of the neural network, avoiding the dominance of the model by some features with excessively large values, which would affect the fusion effect.

[0067] Using the example from step 301, the data after standardization in this step retains the same dimensionality, remaining at 1×141 dimensions.

[0068] Step 303: Extract features from the original data vector to obtain the first feature vector.

[0069] In this step, a recurrent neural network is used to extract temporal dependencies and deep correlations between features in the input data. Specifically, a bidirectional gated recurrent unit (Bi-GRU) can be used to extract the temporal and correlation features of the original data vector and output the first feature vector.

[0070] In this embodiment, the Bi-GRU contains 64 hidden units and processes the input sequence simultaneously through gated recurrent units in both forward and reverse directions, enabling it to capture contextual information before and after the current moment. It should be noted that the input data in this embodiment is not a strictly time series, but the Bi-GRU can still perform correlation modeling on the dimensions of the feature vector, extracting combined feature patterns such as "increased body surface temperature, intensified sweating, and decreased temperature satisfaction." The output of the Bi-GRU is a 128-dimensional feature vector, namely the first feature vector.

[0071] Step 304: Perform weight allocation on the first feature vector to obtain the second feature vector.

[0072] In this step, the importance of each feature is automatically learned through the attention mechanism, highlighting the features that have the most influence on the determination of thermal comfort (such as body surface temperature, sweating status, etc.) and suppressing redundant or irrelevant features (such as certain environmental noise).

[0073] Specifically, the first feature vector can be weighted based on the channel attention mechanism (SE-Net, Squeeze-and-Excitation Network) to obtain the weighted second feature vector.

[0074] Following the example in step 301, the SE-Net module first performs a squeezing operation on the first feature vector, compressing the 128-dimensional feature into a 128-channel global descriptor through global average pooling. Then, it performs an excitation operation, generating 128 channel weight coefficients through two fully connected layers (the intermediate layer uses ReLU activation, and the output layer uses Sigmoid activation). Each weight coefficient is between 0 and 1, representing the importance of the corresponding feature channel. Finally, the first feature vector is multiplied element-wise by these weight coefficients to obtain a weighted second feature vector, still with a dimension of 1×128.

[0075] Step 305: Map the second feature vector to the thermal comfort coefficient.

[0076] In this step, the weighted and fused high-dimensional feature vector is transformed into an intuitive and quantifiable comfort score. This score directly reflects the current thermal perception state of the driver and passengers, providing a clear decision-making basis for subsequent air conditioning control.

[0077] Specifically, a multilayer perceptron (MLP) can be used to map the second feature vector to the thermal comfort coefficient, which is a quantized value obtained after quantizing the second feature vector and used to characterize the temperature perception state of the driver and passengers.

[0078] The structure of the MLP is as follows: the first fully connected layer has 128-dimensional input and 64-dimensional output, with ReLU activation function; the second fully connected layer has 64-dimensional input and 32-dimensional output, with ReLU activation function; the third fully connected layer has 32-dimensional input and 1-dimensional output, with Sigmoid activation function. The Sigmoid function compresses the output value to the range of 0 to 1, then multiplies the value by 10 and rounds it to the nearest integer to obtain a thermal comfort coefficient of 0 to 10. The specific mapping rules are: 0 to 2 points correspond to "slightly cold", 3 to 4 points correspond to "slightly cool", 5 to 6 points correspond to "suitable", 7 to 8 points correspond to "slightly hot", and 9 to 10 points correspond to "extremely hot".

[0079] An exemplary embodiment of this disclosure also provides a method for controlling the in-vehicle environment, the process of generating air conditioning control commands using this method is as follows: Figure 4 As shown, it includes: Step 401: Calculate the overall comfort deviation based on the thermal comfort coefficient.

[0080] In this step, the thermal comfort coefficients of each part are converted into deviations relative to a comfort baseline, and a comprehensive comfort deviation is obtained through weighted fusion. The magnitude and direction (positive or negative) of the comprehensive comfort deviation determine whether the air conditioner needs to heat up or cool down, and the extent of adjustment.

[0081] Thermal comfort coefficients were obtained for the face, back, buttocks, and feet of drivers and passengers. , , , Using the center value of the optimal range for human comfort. Based on the baseline, the comfort deviation for each part is calculated according to the following expression: , Where k = face, back, hip, foot. This represents the thermal comfort coefficient for the corresponding part.

[0082] Then, the overall comfort deviation is calculated according to the following expression: , in, The overall comfort deviation is... For facial comfort deviation, This is the facial weighting coefficient. For back comfort deviation, The back weighting coefficient, For the comfort deviation of the buttocks, This is the weighting factor for the buttocks. For foot comfort deviation, This represents the foot weighting coefficient.

[0083] For example, the weights for each part are set as follows: Facial weight (Most sensitive), back weight Buttock weight Foot weight (Least sensitive). These weighting coefficients reflect the differences in sensitivity of different parts of the human body to temperature changes, with the face being the most sensitive and the feet the least sensitive. They also support fine-tuning of vehicle calibration to adapt to the preferences of different vehicle models or user groups.

[0084] Step 402: Dynamically adjust the temperature control parameters of the car air conditioner based on the current environmental information.

[0085] In this step, the overall comfort deviation is combined with the current actual temperature inside the vehicle to calculate the target temperature that the air conditioner needs to be adjusted to, thus achieving closed-loop feedback regulation that balances comfort and energy consumption.

[0086] The temperature control parameters include at least the target set temperature of the air conditioner.

[0087] The target set temperature of the air conditioner is calculated according to the following expression: , in, Set the target temperature for the air conditioner. To collect ambient temperature inside the vehicle in real time, This is the temperature regulation sensitivity coefficient.

[0088] For example, when When the value is positive (indicating overall heat), Below Air conditioning needs to cool down; when When the value is negative (overall temperature is low), Higher than Air conditioners need to heat up; when When approaching 0, near The air conditioner is maintaining the current temperature.

[0089] Step 403: Generate the air conditioning control command based on the corrected temperature control parameters.

[0090] In this step, the revised target temperature and airflow distribution strategy is converted into specific control commands so that the actuators of the vehicle air conditioning system can execute them accurately.

[0091] Furthermore, the temperature control parameters include the target set temperature of the air conditioner and / or the airflow setting information of each air outlet.

[0092] In this step, the airflow settings for each air outlet are first obtained. Specifically, based on the comfort deviation and physiological weighting coefficient for each area, the airflow for each air outlet is determined as follows: Face air outlet airflow set The airflow of the air outlet is set to... The airflow of the seat ventilation vents is set to... The airflow of the foot-blowing vent is set to... .

[0093] in, The maximum airflow can be set to a baseline value, such as the highest setting out of 10. The actual airflow can be calibrated and limited according to the vehicle's air conditioning hardware capabilities.

[0094] In addition, the wind direction adjustment strategy can be determined based on the sign of the comfort deviation of each part: If the deviation in a certain area is positive (too hot), adjust the air guide vane of the corresponding air outlet to direct more airflow towards that area; if the deviation is negative (too cold), direct the airflow away from that area or reduce the airflow volume in that area. For example, if the deviation in the face is positive (face is hot), increase the airflow volume to the face and adjust the airflow direction to be aimed at the face area; if the deviation in the feet is negative (feet are cold), reduce the airflow volume to the feet or adjust the airflow direction so that it does not blow directly on the feet.

[0095] Then, based on the target set temperature and the airflow setting information, the air conditioning control command is generated. The command is encapsulated in the form of a CAN message, containing parameters such as the target temperature value, the airflow level of each air outlet, and the airflow direction adjustment angle, and is sent to the air conditioning controller via the vehicle's CAN bus.

[0096] An exemplary embodiment of this disclosure also provides a method for controlling the in-vehicle environment, wherein the method is used to generate air conditioning control commands and execute the control process as follows: Figure 5 As shown, it includes: Step 501: Obtain physiological and environmental information of drivers and passengers.

[0097] Step 502: Perform feature fusion on the physiological state information and the environmental information to obtain the thermal comfort coefficient.

[0098] Step 503: Generate air conditioning control instructions based on the thermal comfort coefficient, wherein the air conditioning control instructions include the temperature control parameters of the car air conditioning.

[0099] The implementation principles of steps 501 to 503 are the same as those of steps 101 to 103, and will not be repeated here.

[0100] Step 504: Send the air conditioning control command to the air conditioning control module and obtain the feedback operating parameters.

[0101] This step achieves a closed loop in the execution of control commands. By acquiring the actual operating parameters of the air conditioner after execution (such as actual outlet air temperature, air volume, compressor power, etc.), the control effect is evaluated, and a basis is provided for parameter adjustment in the next control cycle.

[0102] Specifically, air conditioning control commands are sent to the vehicle's electronic control unit (ECU) via the CAN bus. The ECU drives actuators (such as the compressor inverter, damper motor, blower, etc.) to operate according to the commands. Simultaneously, the air conditioning control module feeds back actual operating parameters (such as actual airflow speed at the vents) to the algorithm processing module. The algorithm processing module can then determine whether the commands have been executed correctly based on the feedback parameters.

[0103] According to one exemplary implementation, the automatic control parameters of the air conditioner can be adjusted periodically. The control parameters are fine-tuned in the next control cycle (e.g., cyclically every 100ms). Furthermore, feedback data is stored in a storage module for offline model optimization and fault diagnosis.

[0104] An exemplary embodiment of this disclosure also provides a vehicle cabin environment control device, the structure of which is as follows: Figure 6 As shown, it includes: The information collection module 601 is used to acquire physiological and environmental information of drivers and passengers; The feature fusion module 602 is used to perform feature fusion on the physiological state information and the environmental information to obtain the thermal comfort coefficient. The instruction generation module 603 is used to generate air conditioning control instructions based on the thermal comfort coefficient, the air conditioning control instructions including the temperature control parameters of the car air conditioning.

[0105] The aforementioned device can be integrated into the vehicle's central domain controller or a standalone air conditioning controller, with the modules communicating via a CAN bus or in-vehicle Ethernet. The device also includes a storage module (not shown in the figure) for storing neural network model parameters, historical data, user preference settings, etc., supporting online model updates and data backtracking analysis.

[0106] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0107] An exemplary embodiment of this disclosure also provides a method for controlling the environment inside a vehicle cabin, the specific process of which is as follows: I. Multi-source data acquisition and preprocessing This embodiment first acquires the physiological state information and environmental information of the driver and passengers through an onboard sensor array, camera, microphone, and external communication interface. The physiological state information includes individual characteristic data, numerical physiological data, image feature data, quantitative auxiliary data, and audio data; the environmental information includes in-vehicle and out-of-vehicle environmental data and third-party environmental data.

[0108] The specific data collection process is as follows: (1) Individual characteristic data: The images of drivers and passengers before getting into the vehicle are obtained through the vehicle-mounted AVM camera, and the images after getting into the vehicle are obtained through the OMS camera. The images are input into a pre-trained multimodal large model (such as a multimodal fusion model based on the Transformer architecture) and an object detection algorithm (such as YOLOv8-nano) to extract the age, gender and ethnicity of the drivers and passengers, and to estimate the clothing index (quantitative value) based on this.

[0109] (2) Numerical physiological data: The surface temperature of key parts (N≥3) of the driver and passengers, such as the forehead, neck, and hands, was collected by infrared sensors and normalized to the 0-1 range to form a 1×N dimensional vector; at the same time, the temperature gradient of the surrounding environment was collected. The OMS video stream was analyzed by the vehicle health detection module, and non-contact physiological signal extraction technology was used to obtain respiratory rate, heart rate, blood pressure, blood oxygen, and heart rate variability, and normalized according to preset rules.

[0110] (3) Image feature data: The sweating state (hierarchical quantization), clothing coverage (area ratio) and facial expression (such as thermal expression encoding) are extracted from the OMS image through a multimodal large model and output as feature vectors.

[0111] (4) Audio data: Audio is collected through the in-vehicle microphone, and the frequency of coughing and sneezing is identified by a lightweight audio classification model and normalized.

[0112] (5) Quantitative auxiliary data: Generate 1×8-dimensional vectors by uniquely encoding gender, age and body type; assign health status values ​​as healthy=0, suspected cold=0.5, cold=1; obtain the satisfaction rating of drivers and passengers with the current temperature through a large language model (such as an in-vehicle voice assistant) and normalize it to the 0~1 range.

[0113] (6) Environmental data: Temperature and humidity are collected by the vehicle's interior and exterior temperature and humidity sensors, and corresponding values ​​are collected by the radiation temperature sensor and solar radiation intensity sensor. The data are normalized to the 0-1 range to form a 1×4-dimensional vector. The weather, temperature and air quality index of the vehicle's location are obtained through the weather API.

[0114] After preprocessing, all collected data is concatenated into a 1×141-dimensional original data vector.

[0115] II. Calculation of Thermal Comfort Coefficient using Multi-Source Data Fusion The original data vectors mentioned above are input into a multi-source data fusion neural network, which consists of a four-layer structure: (1) Input layer: Perform secondary standardization on the original data vector to eliminate the influence of dimensions and output the standardized feature vector.

[0116] (2) Feature extraction layer: A 64-unit Bi-GRU is used to extract the temporal dependencies and deep associations between features, and output a 128-dimensional first feature vector. This layer can capture complex combination features such as "increased body surface temperature, increased sweating and decreased temperature satisfaction" or "cold state, slightly higher body surface temperature and lower temperature satisfaction".

[0117] (3) Attention Fusion Layer: The channel attention mechanism (SE-Net) is introduced to compress and activate the first feature vector to generate 128 channel weight coefficients, which are multiplied element by element with the first feature vector to obtain the weighted second feature vector (dimension 1×128), highlighting the features that have the greatest impact on thermal comfort.

[0118] (4) Output layer: A three-layer multilayer perceptron (MLP) is used, with the layers being 128→64 (ReLU), 64→32 (ReLU), and 32→1 (Sigmoid). The second feature vector is mapped to a value in the range of 0 to 1, and then mapped to a thermal comfort coefficient of 0 to 10. At the same time, the network also outputs the original comfort scores for four regions (face, back, buttocks, and feet) for subsequent regional adjustment.

[0119] III. Air Conditioning Adaptive Control Based on the thermal comfort coefficient and comfort values ​​of each zone output above, the following control logic is executed: (1) Using the center value of the optimal human comfort interval Based on this, calculate the comfort deviation for each part: , Where k = face, back, hip, foot. This represents the thermal comfort coefficient for the corresponding part.

[0120] (2) The overall comfort deviation is obtained by weighting and fusing the results based on the physiological weight coefficients: .

[0121] (3) Combined with the real-time ambient temperature inside the vehicle Calculate the target set temperature of the air conditioner: .

[0122] (4) Allocate the air volume at the air outlet according to the weight of each part: Face air outlet airflow set The airflow of the air outlet is set to... The airflow of the seat ventilation vents is set to... The airflow of the foot-blowing vent is set to... .

[0123] (5) Adjust the wind direction dynamically according to the positive or negative of the comfort deviation of each part, so that the airflow is directed towards the hotter parts and avoids the colder parts.

[0124] (6) The system updates the thermal comfort coefficient with a cycle of 100ms and adjusts the air conditioning parameters with a cycle of 1 second. It supports switching between manual mode and intelligent mode.

[0125] This disclosure also provides an on-board controller comprising four main modules: a data acquisition module, an algorithm processing module, an air conditioning control module, and a storage module. The modules communicate via a CAN bus. The data acquisition module is responsible for acquiring and pre-processing data from various sensors; the algorithm processing module runs a multi-source data fusion neural network and control logic; the air conditioning control module executes commands and feeds back operating parameters; and the storage module stores model parameters, historical data, and logs, supporting model iteration and fault diagnosis.

[0126] Through real-vehicle testing in two typical scenarios—summer and winter—this embodiment was compared with traditional air conditioning control schemes and ordinary adaptive schemes. The results were validated based on three indicators: thermal comfort satisfaction (questionnaire survey), control lag time (the time difference between detecting a comfort deviation and executing an adjustment), and energy consumption (power consumption per unit mileage). The test results show that this embodiment outperforms the comparative schemes in all indicators, significantly improving driving comfort and reducing energy consumption.

[0127] An exemplary embodiment of this disclosure also provides a computer apparatus, including: processor; Memory used to store processor-executable instructions; The processor is configured to execute the vehicle cabin environment control method of the embodiments of this disclosure.

[0128] An exemplary embodiment of this disclosure also provides a non-volatile computer-readable storage medium that, when instructions in the storage medium are executed by a computer's processor, enables the computer to perform the vehicle cabin environment control method provided in the embodiments of this disclosure.

[0129] This disclosure provides a method, apparatus, computer device, and storage medium for controlling the in-vehicle environment. It acquires physiological and environmental information of occupants, performs feature fusion on the physiological and environmental information to obtain a thermal comfort coefficient, and then generates air conditioning control commands based on the thermal comfort coefficient. These air conditioning control commands include temperature control parameters for the vehicle's air conditioning system. This method enables real-time perception of the dynamic thermal state of occupants and environmental changes, allowing the air conditioning control commands to accurately match the real-time needs of the occupants. It solves the problems of limited scene adaptation and poor adjustment effects, significantly improving the intelligence level and driving comfort of the vehicle's air conditioning system.

[0130] By integrating multi-source data fusion algorithms, a comprehensive perception system covering factors affecting the thermal comfort of drivers and passengers is constructed. This system integrates individual characteristic data (age, gender, race, clothing index), numerical physiological data (infrared body surface temperature, respiratory rate, heart rate, blood pressure, blood oxygen, heart rate variability), image feature data (sweating status, clothing coverage, facial expression), environmental data (temperature and humidity, radiation temperature, solar radiation intensity), audio data (coughing / sneezing frequency), and quantitative auxiliary data (individual characteristic coding, health status coding, temperature satisfaction).

[0131] In the feature fusion stage, a multi-source data fusion neural network comprising a Bi-GRU feature extraction layer, an SE-Net attention fusion layer, and an MLP output layer is employed. This network extracts deeply correlated features and automatically highlights core influencing factors such as body surface temperature and sweating status through an attention mechanism, while suppressing irrelevant features, resulting in a more accurate and reliable thermal comfort coefficient. Precise temperature control is implemented for different zones, dynamically adjusting temperature, airflow, and airflow direction based on comfort deviations in various body parts, significantly improving thermal comfort adaptability for different population groups.

[0132] By using infrared sensors to collect body surface temperature in real time, an onboard health monitoring module to extract physiological indicators such as respiratory rate and heart rate in real time, and an OMS camera to identify sweating and facial expressions in real time, the system achieves millisecond-level perception of the thermal state of drivers and passengers. The system updates the thermal comfort coefficient every 100ms and adjusts air conditioning parameters every second, enabling rapid response to changes in the physiological state of drivers and passengers and the environment. Simultaneously, the Bi-GRU feature extraction layer, through processing multi-source time-series data, effectively captures the changing trends in human thermal perception, solving the problems of volatility and randomness in single-sensor data, and achieving precise dynamic matching between air conditioning control parameters and the transient thermal perception needs of drivers and passengers.

[0133] By incorporating individual physiological differences such as race, age, and gender into the air conditioning control model, and using the temperature satisfaction values ​​output by the large language model as feedback signals, the air conditioning control system possesses self-learning and adaptive capabilities. Quantifying individual characteristic encoding (unique thermal encoding, 1×8 dimension) and health status encoding (healthy=0, suspected cold=0.5, cold=1) in the auxiliary data enables the model to distinguish the thermal perception characteristics of different groups, achieving truly personalized temperature control. Furthermore, the system supports switching between manual and intelligent modes. In intelligent mode, it automatically executes the entire process control; in manual mode, it responds to user operation commands, taking into account the usage habits and personalized needs of different users.

[0134] Those skilled in the art will also understand that the various illustrative logical blocks and steps listed in the embodiments of this disclosure can be implemented by electronic hardware, computer software, or a combination of both. Whether such functionality is implemented in hardware or software depends on the specific application and the overall system design requirements. Those skilled in the art can implement the described functionality using various methods for each specific application, but such implementation should not be construed as exceeding the scope of protection of the embodiments of this disclosure.

[0135] Furthermore, the term “exemplary” is used herein to mean serving as an example, instance, or illustration. Any aspect or design described herein as “exemplary” is not necessarily to be construed as advantageous compared to other aspects or designs. Rather, the use of the term “exemplary” is intended to present the concept in a concrete manner. As used herein, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or.” That is, unless otherwise specified or clear from the context, “X applies A or B” is intended to mean any of the natural inclusive arrangements. That is, “X applies A or B” satisfies any of the foregoing instances if X applies A; X applies B; or both X applies A and B. Additionally, unless otherwise specified or clear from the context to refer to the singular form, the articles “a” and “an” as used in this application and the appended claims are generally understood to mean “one or more.”

[0136] Similarly, although this disclosure has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art upon reading and understanding this specification and the accompanying drawings. This disclosure includes all such modifications and variations and is limited only by the scope of the claims. In particular, with respect to the various functions performed by the components described above (e.g., elements, resources, etc.), unless otherwise indicated, the terminology used to describe such components is intended to correspond to any component (functionally equivalent) that performs the specific function of the described component, even if structurally not equivalent to the disclosed structure. Furthermore, although specific features of this disclosure may have been disclosed with respect to only one of several implementations, such features may be combined with one or more other features of other implementations, as may be desired and advantageous to any given or particular application. Moreover, with regard to the terms “comprising,” “owning,” “having,” “having,” or variations thereof as used in the detailed description or claims, such terms are intended to be inclusive in a manner similar to the term “including.”

[0137] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0138] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for controlling the environment inside a vehicle cabin, characterized in that, include: Obtain physiological and environmental information of drivers and passengers; The physiological state information and the environmental information are fused to obtain the thermal comfort coefficient; Based on the thermal comfort coefficient, an air conditioning control command is generated, which includes the temperature control parameters of the vehicle's air conditioning system.

2. The vehicle cabin environment control method according to claim 1, characterized in that, The physiological state information includes at least one or more of the following multi-source data: Individual characteristic data, numerical physiological data, image feature data, quantitative auxiliary data, and audio data. The environmental information includes at least one or more of the following data: Vehicle interior and exterior environmental data, third-party environmental data, The steps for obtaining the physiological state information and environmental information of the driver and passengers include: The physiological state information is extracted using in-vehicle sensor arrays and a multimodal large model; The environmental information is extracted through the in-vehicle sensor group and communication module.

3. The vehicle cabin environment control method according to claim 1, characterized in that, The step of fusing features of the physiological state information and the environmental information to obtain the thermal comfort coefficient includes: The physiological state information and the environmental information are preprocessed and quantized to obtain the raw data vector; Feature extraction is performed on the original data vector to obtain a first feature vector; The first feature vector is weighted to obtain the second feature vector; The second feature vector is mapped to the thermal comfort coefficient.

4. The vehicle cabin environment control method according to claim 3, characterized in that, Before the step of extracting features from the original data vector to obtain the first feature vector, the method further includes: The original data vector is standardized to unify the data dimensions.

5. The vehicle cabin environment control method according to claim 3, characterized in that, The step of extracting features from the original data vector to obtain the first feature vector includes: The temporal and correlation features of the original data vector are extracted by a bidirectional gated loop unit, and the first feature vector is output.

6. The vehicle cabin environment control method according to claim 3, characterized in that, The step of weighting the first feature vector to obtain the second feature vector includes: Based on the channel attention mechanism, the first feature vector is weighted to obtain the weighted second feature vector.

7. The vehicle cabin environment control method according to claim 3, characterized in that, The step of mapping the second feature vector to the thermal comfort coefficient includes: A multilayer perceptron is used to map the second feature vector to the thermal comfort coefficient, which is a quantitative value obtained after quantizing the second feature vector and used to characterize the temperature perception state of the driver and passengers.

8. The vehicle cabin environment control method according to claim 1, characterized in that, The step of generating air conditioning control commands based on the thermal comfort coefficient includes: Calculate the overall comfort deviation based on the thermal comfort coefficient; The temperature control parameters of the car air conditioner are dynamically adjusted based on the current environmental information. The air conditioning control command is generated based on the corrected temperature control parameters.

9. The vehicle cabin environment control method according to claim 8, characterized in that, The step of calculating the overall comfort deviation based on the thermal comfort coefficient includes: The overall comfort deviation is calculated using the following expression: ， in, The overall comfort deviation is... For facial comfort deviation, This is the facial weighting coefficient. For back comfort deviation, The back weighting coefficient, For the comfort deviation of the buttocks, This is the weighting factor for the buttocks. For foot comfort deviation, This represents the foot weighting coefficient.

10. The vehicle cabin environment control method according to claim 9, characterized in that, The temperature control parameters include at least the target set temperature of the air conditioner, and the step of dynamically correcting the temperature control parameters of the car air conditioner based on the current environmental information includes: The target set temperature of the air conditioner is calculated according to the following expression: ， in, Set the target temperature for the air conditioner. To collect ambient temperature inside the vehicle in real time, This is the temperature regulation sensitivity coefficient.

11. The vehicle cabin environment control method according to claim 8, characterized in that, The temperature control parameters include the target set temperature of the air conditioner and / or the airflow setting information of each air outlet. The step of generating the air conditioner control command based on the corrected temperature control parameters includes: Obtain the airflow setting information for each air outlet; The air conditioning control command is generated based on the target set temperature and the air volume setting information.

12. The vehicle cabin environment control method according to claim 1, characterized in that, The method further includes: The system sends air conditioning control commands to the air conditioning control module and obtains the feedback operating parameters.

13. A vehicle cabin environment control device, characterized in that, include: The information collection module is used to acquire physiological and environmental information of drivers and passengers; The feature fusion module is used to perform feature fusion on the physiological state information and the environmental information to obtain the thermal comfort coefficient. The instruction generation module is used to generate air conditioning control instructions based on the thermal comfort coefficient, the air conditioning control instructions including the temperature control parameters of the car air conditioning.

14. A computer device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to perform the vehicle cabin environment control method as described in any one of claims 1 to 12.

15. A non-volatile computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of a computer, the computer is able to perform the vehicle cabin environment control method as described in any one of claims 1 to 12.

Images