Vehicle control method and device, vehicle, chip and medium

Abstract

This application proposes a vehicle control method, device, vehicle, chip, and medium, relating to the field of smart cockpits. The method includes: acquiring CSI signals obtained from at least one WiFi sensing node sensing the cockpit within the vehicle; extracting signal features associated with at least one occupant risk event from the CSI signals based on a feature type associated with that event; detecting occupant risk events based on the signal features to obtain event detection results; wherein the event detection results are used to determine the occupant risk state in the occupant risk event; and controlling the vehicle to execute a vehicle control strategy adapted to the event detection results. This allows the vehicle's control actions / intervention actions to accurately match the occupant's real-time state, avoiding user experience disruption caused by ineffective control, mitigating the privacy leakage risks of image acquisition schemes in related technologies, and improving occupant acceptance and willingness to use in-vehicle sensing functions.

Description

Technical Field

[0001] This application relates to the field of intelligent cockpit technology, and more particularly to a vehicle control method, device, vehicle, chip, and medium. Background Technology

[0002] In related technologies, visual cameras installed inside the cabin continuously collect images of the vehicle interior and perform image recognition to determine occupant status and detect risk events. However, this method requires long-term recording and storage of occupant images, which easily leads to the risk of privacy leaks. Most occupants have a strong psychological aversion to this indiscriminate visual data collection, resulting in poor practical experience and acceptance. Summary of the Invention

[0003] This application proposes a vehicle control method, apparatus, vehicle, chip, and medium to at least partially solve one of the technical problems in the related art.

[0004] One embodiment of this application proposes a vehicle control method, comprising: acquiring a CSI signal obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin; extracting signal features associated with at least one feature type related to occupant risk events from the CSI signal; detecting the occupant risk event based on the signal features to obtain an event detection result; wherein the event detection result is used to determine the occupant risk state of the occupant in the occupant risk event; and controlling the vehicle to execute a vehicle control strategy adapted to the event detection result.

[0005] Another embodiment of this application proposes a vehicle control device, comprising: an acquisition module for acquiring Channel State Information (CSI) signals obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin; an extraction module for extracting signal features associated with at least one occupant risk event type from the CSI signals; a detection module for detecting the occupant risk event based on the signal features and obtaining an event detection result; wherein the event detection result is used to determine the occupant risk state of the occupant in the occupant risk event; and a control module for controlling the vehicle to execute a vehicle control strategy adapted to the event detection result.

[0006] Another aspect of this application provides a vehicle, including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to implement the vehicle control method as described in the preceding aspect.

[0007] In another aspect of this application, a chip is provided that includes an interface circuit and a processing circuit coupled to each other. The interface circuit is used to input or output signals, and the processing circuit is configured to perform the vehicle control method as described in the preceding aspect.

[0008] In another aspect of this application, a non-transitory computer-readable storage medium is provided, on which computer program instructions are stored, which, when executed by a processor, implement the vehicle control method as described in the foregoing aspect.

[0009] Another aspect of this application provides a computer program product having a computer program stored thereon, which, when executed by a processor, implements the vehicle control method as described in the foregoing aspect.

[0010] The vehicle control method, device, vehicle, chip, and medium proposed in this application extract signal features adapted to the feature types associated with each occupant risk event from the CSI signals sensed in the cabin by WiFi sensing nodes, and detect each occupant risk event accordingly. This improves the accuracy and reliability of event detection results, and allows the vehicle to execute an adapted vehicle control strategy based on accurate event detection results. This enables the vehicle's control / intervention actions to accurately match the occupant's real-time state, avoiding the experience disturbance caused by ineffective control. As a result, it avoids the privacy leakage risk of image acquisition schemes in related technologies and improves the occupant's acceptance and willingness to use the vehicle sensing function.

[0011] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0012] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A schematic flowchart of a vehicle control method provided for an exemplary embodiment of this application; Figure 2 A schematic flowchart of another vehicle control method provided for an exemplary embodiment of this application; Figure 3 A schematic flowchart of yet another vehicle control method provided for an exemplary embodiment of this application; Figure 4 A schematic flowchart of another vehicle control method provided for an exemplary embodiment of this application; Figure 5 A schematic diagram illustrating the principle of a personalized thermal comfort control scheme provided for an exemplary embodiment of this application; Figure 6 A schematic flowchart of another vehicle control method provided for an exemplary embodiment of this application; Figure 7 A schematic flowchart of another vehicle control method provided for an exemplary embodiment of this application; Figure 8 A schematic flowchart of yet another vehicle control method provided for an exemplary embodiment of this application; Figure 9 A schematic flowchart of another vehicle control method provided for an exemplary embodiment of this application; Figure 10 A schematic flowchart of another vehicle control method provided for an exemplary embodiment of this application; Figure 11 A schematic diagram illustrating the principle of a driver's sudden health abnormality detection and emergency safety handling scheme provided for an exemplary embodiment of this application; Figure 12 A schematic diagram of the structure of a vehicle control device provided for an exemplary embodiment of this application; Figure 13 This is a block diagram illustrating a vehicle according to an exemplary embodiment; Figure 14 This is a schematic diagram of the structure of a chip proposed as an exemplary embodiment of this application. Detailed Implementation

[0013] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0014] It should be noted that the acquisition, storage, use, and processing of data in this application comply with relevant laws and regulations and do not violate public order and good morals.

[0015] It should also be noted that all data processed in this application is data that has been explicitly authorized by the user or relevant parties, and has been de-identified or anonymized before collection and use, and does not contain any personally identifiable information or user privacy content; all data is used only for vehicle control purposes, ensuring that data security and user privacy rights are fully protected while achieving technical effects.

[0016] The vehicle control method, apparatus, vehicle, chip, and medium according to embodiments of this application are described below with reference to the accompanying drawings.

[0017] Figure 1This is a schematic flowchart of a vehicle control method provided for an exemplary embodiment of this application.

[0018] It should be noted that the vehicle control method of this application embodiment can be applied to a vehicle control device. In some possible embodiments, the vehicle control device can be configured in a vehicle or a chip so that the vehicle or chip can perform vehicle control functions. Additionally, in some possible embodiments, the vehicle control device can also be software within the vehicle.

[0019] The vehicle can be a hybrid vehicle, a non-hybrid vehicle, an electric vehicle, a fuel cell vehicle, or other types of vehicles, and this application embodiment does not limit this.

[0020] For ease of explanation, the following description will use a vehicle as the executor of this vehicle control method.

[0021] like Figure 1 As shown, the vehicle control method may include the following steps S101 to S104: Step S101: Obtain Channel State Information (CSI) signals from at least one WiFi sensing node in the vehicle cabin that senses the cabin.

[0022] The vehicle cabin is equipped with at least one WiFi sensing node. It should be noted that this application does not impose specific restrictions on the deployment location or number of WiFi sensing nodes. For example, one WiFi sensing node can be deployed at either the front or rear of the cabin, with each WiFi sensing node integrating WiFi access point (AP) functionality; alternatively, one WiFi sensing node (or WiFi sensing transmitter Tx) can be deployed in front of the driver's seat headrest or at the steering column, and one WiFi sensing node (or WiFi sensing receiver Rx) can be deployed directly behind the driver (e.g., at the B-pillar); alternatively, one WiFi sensing node (or WiFi sensing transmitter Tx) can be deployed at the front of the cabin (e.g., at the dashboard), and one WiFi sensing node (or WiFi sensing receiver Tx) can be deployed at the rear of the cabin (e.g., at the rear window).

[0023] Each WiFi sensing node is configured with an antenna array. For example, each WiFi sensing node can be configured with a 4×4 Multiple-Input Multiple-Output (MIMO) antenna array, where the spacing between adjacent antennas can be designed to be half the wavelength of the corresponding operating frequency band; alternatively, each WiFi sensing node can be configured with a 2×2 MIMO antenna array. It should be understood that the number of antennas in the above-described antenna array is merely illustrative, and this application is not limited thereto. The number of antennas in each WiFi sensing node can be configured based on actual application requirements.

[0024] The CSI signal sensed by each WiFi sensing node can be used to instruct different receiving antennas in the antenna array of that WiFi sensing node to collect complex channel response data (or complex CSI values) for different subcarriers. For example, taking the CSI signal as a matrix, the element in the i-th row and j-th column of the matrix is ​​used to instruct the i-th receiving antenna in the antenna array of the WiFi sensing node to collect complex channel response data for the j-th subcarrier.

[0025] Step S102: Based on at least one occupant risk event associated feature type, extract signal features associated with the feature type from the CSI signal.

[0026] Occupant risk events refer to predictable negative events that may affect the safety, health, or riding experience of occupants during vehicle operation or parking, including but not limited to: comfort-related risk events, health-related risk events, and motion sickness events. The characteristic types associated with different occupant risk events may be different or the same; this application's embodiments do not impose any limitations on this.

[0027] For example, the feature type associated with comfort-related risk events can be micro-motion features, the feature type associated with health-related risk events can be multi-dimensional physiological features, and the feature type associated with motion sickness events can also be multi-dimensional physiological features. Among them, micro-motions include, but are not limited to: cold-related movements (such as hunching shoulders, rubbing hands, crossing legs, and folding arms), heat-related movements (such as fanning, stretching, loosening collar, and wiping sweat), and neutral state movements (such as normal sitting posture and minor movement).

[0028] In this embodiment of the application, signal features associated with each occupant risk event can be extracted from the CSI signal based on the feature type associated with that event.

[0029] Step S103: Detect occupant risk events based on signal characteristics and obtain event detection results; wherein, the event detection results are used to determine the occupant risk status of the occupants in the occupant risk event.

[0030] Each type of occupant risk event has multiple corresponding occupant risk states. For example, the multiple occupant risk states corresponding to comfort-related risk events can also be called thermal comfort states, including but not limited to: cold, slightly cold, moderate, slightly hot, and hot; the multiple occupant risk states corresponding to health-related risk events can also be called abnormal health states or abnormal health types, including but not limited to: cardiac arrest, myocardial infarction, stroke, epileptic seizure, severe hypoglycemia, and normal state; the multiple occupant risk states corresponding to motion sickness events can also be called motion sickness states, including but not limited to: motion sickness precursors, impending vomiting, and normal state.

[0031] In this embodiment of the application, occupant risk events can be detected based on the signal features extracted above, and event detection results can be obtained. The event detection results are used to determine the occupant risk status of the occupants in the occupant risk event.

[0032] As one possible implementation, the event detection results can be used to indicate the likelihood (or probability) of various occupant risk states under a corresponding occupant risk event within a target time period; wherein, the target time period is associated with the occupant risk event, including the current time period or a future time period.

[0033] As an example, the event detection results may include the probability that an occupant in the cabin is in each occupant risk state under an occupant risk event. In the case that the occupant risk event includes comfort-related risk events, the event detection results may be used to indicate the probability that each occupant in the cabin will experience multiple thermal comfort states under comfort-related risk events during the current time period.

[0034] It is understandable that by outputting the probability distribution of different thermal comfort states for each occupant in the cabin, rather than just outputting a single thermal comfort state, the differentiated comfort needs of different occupants can be accurately perceived. This allows for subsequent implementation of zoned temperature control based on the differentiated comfort needs of different occupants, thereby improving the occupant's riding comfort experience.

[0035] As another example, the event detection results can include the probability that an occupant in the cabin is in each occupant risk state under an occupant risk event. In the case that the occupant risk event includes a health-related risk event, the event detection results can be used to indicate the probability that an occupant will experience multiple types of health abnormalities under a health-related risk event in the future.

[0036] It should be noted that, unlike related technologies that only passively detect health-related risk events (or sudden health abnormalities) after they occur, this application can proactively predict the probability of passengers experiencing various types of health abnormalities in the future, upgrading health protection from "post-event handling" to "pre-event warning," thus avoiding safety accidents caused by passengers' sudden health problems.

[0037] As another example, the event detection results may include the probability that an occupant in the cabin is in each occupant risk state under an occupant risk event. In the case that the occupant risk event includes motion sickness, the event detection results are used to indicate the probability that an occupant will experience multiple motion sickness states under a motion sickness event during the current time period.

[0038] Understandably, by outputting the probability of passengers experiencing different motion sickness states during the current time period, the potential trend of motion sickness can be identified in advance, without having to wait until passengers have already shown obvious motion sickness reactions before passively intervening. This can suppress the occurrence and development of motion sickness symptoms in advance and effectively improve the riding experience of people prone to motion sickness.

[0039] It should be noted that the above examples are only illustrative of events where the event detection results include one of the following: comfort-related risk events, health-related risk events, and motion sickness events. However, this application is not limited to this. In actual applications, the event detection results may include any two of the following: comfort-related risk events, health-related risk events, and motion sickness events. Alternatively, the event detection results may include all three of these events simultaneously. This application does not impose any specific limitations on this.

[0040] Step S104: Control the vehicle to execute a vehicle control strategy adapted to the event detection results.

[0041] Vehicle control strategies include, but are not limited to: temperature control strategies, warning strategies (such as health warning strategies and motion sickness warning strategies), etc.

[0042] In this embodiment, a vehicle control strategy that is compatible with the current event detection result can be determined from a variety of vehicle control strategies, and the adapted vehicle control strategy can be controlled to be executed, so as to achieve dynamic matching between the cabin environment, vehicle control strategy and real-time status of the occupants, allowing vehicle control to be upgraded from passively responding to commands to actively adapting to the needs of the occupants, thereby optimizing the riding experience of the occupants while ensuring driving safety.

[0043] The vehicle control method of this application embodiment extracts signal features adapted to the feature types associated with each occupant risk event from the CSI signals sensed in the cabin by WiFi sensing nodes, and detects each occupant risk event accordingly. This improves the accuracy and reliability of event detection results, and controls the vehicle to execute an adapted vehicle control strategy based on accurate event detection results. This allows the vehicle's control actions / intervention actions to accurately match the real-time state of the occupants, avoiding the experience disturbance caused by ineffective control. Thus, it avoids the privacy leakage risk of image acquisition schemes in related technologies and improves the occupants' acceptance and willingness to use the vehicle sensing function.

[0044] As one possible implementation method, Figure 2 This is a schematic flowchart illustrating another vehicle control method provided as an exemplary embodiment of this application. It should be noted that this vehicle control method can be executed alone, or it can be executed together with any embodiment or possible implementation thereof in this application, or it can be executed together with any technical solution in related technologies. This application does not impose any limitations on this.

[0045] like Figure 2 As shown, the vehicle control method may include the following steps S201 to S207: Step S201: Obtain the CSI signal obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin.

[0046] It should be noted that the explanation of step S201 can be found in the relevant description in any embodiment of this application, and will not be repeated here.

[0047] Step S202, in response to occupant risk events, including comfort-related risk events, extracts local CSI signals corresponding to multiple seating areas within the cabin from the CSI signals; wherein, the CSI signals include complex channel response data of multiple subcarriers collected by multiple receiving antennas in the antenna array of the WiFi sensing node.

[0048] The cabin includes multiple seating areas (also known as seating perception areas). For example, in a 4-seater vehicle, these areas include the driver's seat area, the front passenger seat area, the left rear seat area, and the right rear seat area.

[0049] The CSI signal includes complex channel response data of multiple subcarriers collected by multiple receiving antennas in the antenna array of the WiFi sensing node. For example, if the CSI signal is in matrix form, the element in the i-th row and j-th column of the matrix indicates that the i-th receiving antenna in the antenna array of the WiFi sensing node is collecting complex channel response data of the j-th subcarrier.

[0050] In the embodiments of this application, when occupant risk events include comfort-related risk events, the local CSI signal corresponding to each seat area can be extracted from the CSI signal.

[0051] One possible implementation involves dividing the cabin into multiple perceptible seating areas; using a Time Difference of Arrival (TDOA) algorithm, the occupant's position is calculated based on the time difference of the CSI signals sensed by each WiFi sensing node; based on the estimated occupant position, each CSI signal is marked as a corresponding area; the CSI signals are grouped according to the marked areas to form multiple independent signal streams (referred to as CSI local signals in this application). That is, in this application, a location-based spatial filtering algorithm can be used to distinguish the reflection paths of different seating areas using the time difference of arrival of the CSI signals, thereby achieving the separation of CSI local signals from multiple seating areas.

[0052] Step S203: Based on the target frequency band associated with various micro-movements, extract the time-frequency characteristics of the CSI local signal of each seat area in the cockpit.

[0053] The explanation of the micro-movements can be found in the relevant descriptions in the above embodiments, and will not be repeated here. It should be noted that the frequency characteristics of the micro-movements are concentrated in the range of 0.5Hz-10Hz, which is separate from the frequency bands of physiological signals such as heart rate / respiration. Therefore, in this application, the target frequency band is, for example, the 0.5Hz-10Hz band, or the target frequency band may be included within the 0.5Hz-10Hz band.

[0054] In this embodiment, the time-frequency characteristics of the local CSI signal in each seat area within the cabin can be extracted based on the target frequency band. For example, a Short-Time Fourier Transform (STFT) can be used to convert the local CSI signal of each seat area into a time-frequency spectrum. Based on the target frequency band, the time-frequency spectrum is band-pass filtered to obtain the corresponding time-frequency characteristics, thereby accurately removing frequency interference from irrelevant physiological signals such as heart rate and respiration.

[0055] For example, STFT parameters, such as a window size of 128 points and a step size of 32 points, using a Hamming window, can avoid spectrum leakage while ensuring that the time-frequency resolution matches the target frequency band, further enhancing the distinction between micro-motion features and irrelevant physiological signals such as heart rate and respiration, and improving the accuracy of time-frequency features.

[0056] Step S204: Based on the difference in Angle of Arrival (AOA) between the complex channel response data collected by different receiving antennas in the CSI local signal of each seat area, determine the spatial characteristics of the CSI local signal of each seat area.

[0057] For example, AOA information of the CSI local signal of each seat area can be extracted based on the spatial diversity characteristics of the antenna array. Considering that the micro-movements of different occupants will cause the CSI local signal of the seat area to produce different perturbation patterns in the spatial dimension, the spatial features in the form of a matrix with the dimension of [number of antennas × number of subcarriers] can be constructed by relying on the AOA differences between the complex channel response data collected by different receiving antennas in the AOA information, so as to fully preserve the spatial domain signal change rules of each seat area.

[0058] Step S205: The time-frequency features and spatial features corresponding to each seat area are fused to obtain the micro-motion features of each seat area.

[0059] In this embodiment of the application, a feature fusion algorithm can be used to fuse the time-frequency features and spatial features corresponding to each seat area in a targeted manner to obtain the micro-motion features of each seat area.

[0060] Step S206: Detect comfort-related risk events based on the micro-motion characteristics of each seating area, and obtain the event detection results.

[0061] For example, a thermal comfort prediction model in the field of artificial intelligence can be used to detect comfort-related risk events based on the micro-motion characteristics of each seating area and output the event detection results; wherein, the event detection results can be used to indicate the probability of occupants in each seating area of ​​the cabin experiencing multiple thermal comfort states under comfort-related risk events in the current time period.

[0062] Step S207: Control the vehicle to execute a temperature control strategy adapted to the event detection results.

[0063] As one possible approach, a temperature control strategy that matches the event detection result can be determined from a variety of preset temperature control strategies, and the vehicle can be controlled to execute the adapted temperature control strategy.

[0064] As an example, for any target seating area in the cabin, the probability of an occupant in that target seating area experiencing multiple thermal comfort states can be determined based on the event detection results. The target thermal comfort state with the highest probability can then be determined from among the multiple thermal comfort states. Based on the target thermal comfort state corresponding to that target seating area, a temperature control strategy that is compatible with (or associated with) that target thermal comfort state can be determined from among multiple temperature control strategies. Based on the temperature control strategy, the operation of the on-board components associated with that target seating area can be controlled to accurately match the differentiated thermal needs of different occupants.

[0065] For example, taking the thermal comfort states as an example, which include five states: cold, slightly cold, moderate, slightly warm, and warm, the temperature control strategy includes: The first type: Temperature control strategy related to cold: Increase the temperature setting of the vehicle air conditioner by 2°C; turn on the seat heating function if the vehicle seat is equipped with seat heating components; increase the air outlet temperature of the vehicle air conditioner; The second approach: Temperature control strategy related to slightly colder temperatures: Increase the vehicle's air conditioning temperature setting by 1°C; reduce the airflow of the vehicle's air conditioning. The third type: Temperature control strategy related to moderation: Maintain the current temperature setting and airflow of the vehicle's air conditioning; The fourth type: Temperature control strategy related to slightly heat: Set the temperature of the vehicle air conditioner to -1°C and increase the air volume of the vehicle air conditioner; The fifth strategy: Heat-related temperature control: Set the vehicle air conditioner temperature to -2°C; if the vehicle seat is equipped with a seat ventilation component, activate the seat ventilation function of the seat ventilation component to maximize cooling.

[0066] The vehicle control method of this application extracts independent CSI local signals for each seat area from the CSI signal, and extracts time-frequency features associated with multiple micro-movements from the CSI local signals of each seat area based on the target frequency bands associated with multiple micro-movements, so as to filter out frequency band interference that is irrelevant to physiological characteristics; determines spatial features based on the AOA difference of complex channel response data of different receiving antennas in the CSI local signals of each seat area, and fuses spatial features and time-frequency features to generate micro-movement features specific to each seat area, which can realize parallel feature extraction of multiple seat areas and improve the recognition and extraction accuracy of micro-movement features of each seat area.

[0067] As one possible implementation method, Figure 3This is a schematic flowchart illustrating another vehicle control method provided as an exemplary embodiment of this application. It should be noted that this vehicle control method can be executed alone, or it can be executed together with any embodiment of this application or any possible implementation thereof, or it can be executed together with any technical solution in related technologies. This application does not impose any limitations on this.

[0068] like Figure 3 As shown, the vehicle control method may include the following steps S301 to S307: Step S301: Obtain the CSI signal obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin.

[0069] Step S302: In response to occupant risk events, including comfort-related risk events, extract micro-motion features corresponding to multiple seating areas in the cabin from the CSI signal.

[0070] It should be noted that the explanations of steps S301 to S302 can be found in the relevant descriptions in any embodiment of this application, and will not be repeated here.

[0071] Step S303: Using the coding network in the thermal comfort prediction model, extract the common features among the micro-motion features of multiple seating areas.

[0072] Among them, the thermal comfort prediction model can be a trained artificial intelligence model that has learned the correspondence between input data (micro-motion characteristics of multiple seating areas) and output data (such as whether there are occupants in each seating area, and the thermal comfort state of occupants in the target seating area during comfort-related risk events); where the target seating area refers to the seating area where there are occupants.

[0073] The thermal comfort prediction model can adopt an encoder-decoder structure, which includes an encoding network (also known as a shared encoder, used to extract common features of CSI local signals in different seating areas) and multiple decoding networks (also known as position decoders, used to independently predict the thermal comfort state of occupants for each seating area).

[0074] In this embodiment of the application, the coding network in the thermal comfort prediction model can be used to extract the common features among the micro-motion features of multiple seating areas.

[0075] Step S304: Using the occupant detection network in the thermal comfort prediction model, the presence of occupants is detected based on common features to obtain occupant detection results; wherein, the occupant detection results are used to indicate whether there are occupants in each seating area.

[0076] The thermal comfort prediction model also includes an occupant detection network, which detects the presence of occupants in each seating area to determine whether there are occupants in each seating area.

[0077] In this embodiment of the application, an occupant detection network in the thermal comfort prediction model can be used to detect the presence of occupants based on common features, so as to determine whether there are occupants in each seating area in the cabin.

[0078] Step S305: Activate the target decoding network corresponding to the target seat area among the multiple decoding networks in the thermal comfort prediction model; wherein, the target seat area is the seat area where the occupant detection result indicates the presence of an occupant.

[0079] The thermal comfort prediction model includes multiple decoding networks, each corresponding to a seating area, used to predict the thermal comfort status of occupants in the corresponding seating area.

[0080] In this embodiment of the application, a target seat area with occupants can be determined from multiple seat areas based on the occupant detection results, and the target decoding network corresponding to the target seat area can be activated among multiple decoding networks in the thermal comfort prediction model.

[0081] Step S306: The common features are decoded using a target decoding network to output a probability distribution; wherein the probability distribution is used to indicate the probability of an occupant in the target seating area experiencing multiple thermal comfort states during the current time period.

[0082] In this embodiment of the application, an activated target decoding network can be used to decode common features to output a probability distribution, wherein the probability distribution is used to indicate the probability that an occupant in the target seating area will experience multiple thermal comfort states during the current time period.

[0083] In any embodiment of this application, a supervised training mechanism is used to train the thermal comfort prediction model as an example. The training method of the thermal comfort prediction model is as follows: acquiring training data, wherein the training data includes sample micro-motion features, thermal comfort annotation information, occupant annotation information, and micro-motion annotation information associated with at least one sample seat area; processing the sample micro-motion features using the thermal comfort prediction model to obtain output results, wherein the output results include thermal comfort prediction information, occupant prediction information, and micro-motion prediction information for the sample seat area; and training the thermal comfort prediction model based on the differences between thermal comfort annotation information and thermal comfort prediction information, the differences between occupant annotation information and occupant prediction information, and the differences between micro-motion annotation information and micro-motion prediction information.

[0084] Among them, thermal comfort labeling information can be used to indicate the actual thermal comfort status of occupants in the sample seating area; thermal comfort prediction information can be used to indicate the predicted thermal comfort status of the sample as an occupant in the area.

[0085] The occupant labeling results can be used to indicate the actual occupant presence status in the sample seat area; the occupant prediction results can be used to indicate whether there are occupants in the sample seat area. The method of obtaining these results can be found in the relevant description of step S304, and will not be repeated here.

[0086] Among them, micro-motion annotation information can be used to indicate the type of micro-motion actually performed by the occupant; micro-motion prediction information can be used to indicate the predicted type of micro-motion performed by the occupant. For example, the thermal comfort prediction model also includes a micro-motion recognition network, which can be used to predict the micro-motion prediction information of the sample seat area based on the micro-motion characteristics of the sample.

[0087] In this application, the value of the loss function can be determined based on the differences between thermal comfort annotation information and thermal comfort prediction information (hereinafter referred to as the first difference), the differences between occupant annotation information and occupant prediction information (hereinafter referred to as the second difference), and the differences between micro-motion annotation information and micro-motion prediction information (hereinafter referred to as the third difference). This value is hereinafter referred to as the loss value. Based on the loss value, the model parameters in the thermal comfort prediction model can be adjusted to minimize the loss value. The loss value is positively correlated with the aforementioned first, second, and third differences. For example, the loss value L can be calculated using the following formula. total : L total = α·L comfort + β·L action + γ·L presence ; Where α, β, and γ are preset weights, with α > β > γ. For example, α = 1.0, β = 0.3, and γ = 0.2; L comfort For the loss of comfort determined based on the first difference; L action For the loss of motion determined based on the third difference; L presence The existence loss is determined based on the second difference.

[0088] It should be noted that the above example only uses minimizing the loss value as the termination condition for model training, but this application is not limited to this. For example, the termination condition may also include: the training time reaches a set time, the training rounds reach a set number of rounds, etc. The embodiments of this application do not impose specific limitations on this.

[0089] In summary, by adopting a supervised training approach, the output of the thermal comfort prediction model is simultaneously constrained and corrected based on three types of labeled information: thermal comfort state, occupant presence state, and micro-motion type. This allows the feature extraction and prediction processes of the three tasks to share parameters and mutually enhance each other. This avoids feature redundancy and computational waste caused by independent training of a single task, and also strengthens the prediction accuracy of the main task (thermal comfort prediction) through the supervision signals of the two auxiliary tasks, micro-motion recognition and occupant presence detection, thereby improving the inference accuracy of the thermal comfort prediction model.

[0090] Step S307: Control the vehicle to execute a temperature control strategy that is adapted to the probability distribution in the occupant detection results.

[0091] As one possible implementation, when the event detection result indicates the probability of an occupant in the target seating area experiencing multiple thermal comfort states during the current time period, the target thermal comfort state with the highest probability can be determined from the multiple thermal comfort states based on the probability of the multiple thermal comfort states in the event detection result. Based on the target thermal comfort state corresponding to the target seating area, a temperature control strategy can be determined, thereby enabling the operation control of the vehicle components associated with the target seating area based on the temperature control strategy. The vehicle components include at least one of the following: vehicle air conditioning, seat heating components, and seat ventilation components.

[0092] Using the example from step S207, assuming the target thermal comfort state is cold, at least one of the following can be performed: increase the temperature setting of the vehicle air conditioner associated with the target seating area by 2°C; if a seat heating element is provided in the target seating area, activate the seat heating function; increase the air outlet temperature of the vehicle air conditioner. As another example, assuming the target thermal comfort state is hot, at least one of the following can be performed: decrease the temperature setting of the vehicle air conditioner associated with the target seating area by 2°C; if a seat ventilation element is provided in the target seating area, activate the seat ventilation function of that element to maximize cooling.

[0093] Understandably, based on the probability of an occupant in the target seating area experiencing multiple thermal comfort states during the current time period, the actual target thermal comfort state of the occupant is determined, and a corresponding temperature control strategy is matched to drive the operation of the vehicle components (such as vehicle air conditioning, seat heating components, and seat ventilation components) associated with the target seating area. Thus, personalized temperature control at the single-seat level can be achieved without relying on manual operation by the occupant, accurately matching the differentiated thermal needs of different occupants.

[0094] The vehicle control method of this application extracts common features among the micro-motion features of multiple seat areas through an encoding network, and detects the presence of occupants through an occupant detection network. It only activates the target decoding network corresponding to the target seat area where there are occupants to complete the detection of comfort-related risk events. This avoids activating the decoding network corresponding to empty seat areas to perform invalid inference calculations. While ensuring the detection accuracy of comfort-related risk events, it reduces the inference overhead of the model and improves the response speed of the vehicle.

[0095] As one possible implementation method, Figure 4 This is a schematic flowchart illustrating another vehicle control method provided as an exemplary embodiment of the present application. It should be noted that this vehicle control method can be executed alone, or it can be executed together with any embodiment of the present application or any possible implementation thereof, or it can be executed together with any technical solution in related technologies. The embodiments of the present application do not impose any limitations on this.

[0096] like Figure 4 As shown, the vehicle control method may include the following steps S401 to S408: Step S401: Obtain the CSI signal obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin.

[0097] Step S402: In response to occupant risk events, including comfort-related risk events, extract micro-motion features corresponding to multiple seating areas within the cabin from the CSI signal.

[0098] Step S403: Detect comfort-related risk events based on the micro-motion characteristics of multiple seating areas to obtain event detection results; wherein, the event detection results are used to indicate the probability of an occupant in a target seating area within the cabin experiencing multiple thermal comfort states during the current time period, and the target seating area refers to the seating area where an occupant is located.

[0099] It should be noted that the explanations of steps S401 to S403 can be found in the relevant descriptions in any embodiment of this application, and will not be repeated here.

[0100] Step S404: Determine the target thermal comfort state from the multiple thermal comfort states based on the probabilities of the multiple thermal comfort states.

[0101] In this embodiment, for any target seating area, the target thermal comfort state of the passenger in the target seating area can be determined from among multiple thermal comfort states based on the probability of the passenger in the target seating area experiencing multiple thermal comfort states during the current time period. For example, the thermal comfort state with the highest probability can be taken as the target thermal comfort state.

[0102] Step S405: Based on the vehicle's remaining driving range and trip planning information, predict the remaining state of charge (SOC) of the vehicle's battery when the vehicle arrives at its destination.

[0103] The remaining driving range refers to the maximum distance that the vehicle can continue to drive, calculated based on the remaining battery power and real-time energy consumption level in its current state.

[0104] Among them, trip planning information refers to the path data generated by the navigation system from the current location to the destination, including path-related information that will affect the actual energy consumption of the vehicle, such as the total mileage of the path, road condition type, gradient distribution, and the proportion of congested road sections.

[0105] For example, a trained deep learning model can be used to predict the remaining SOC of the vehicle battery when the vehicle arrives at its destination, based on the vehicle's remaining driving range and trip planning information; wherein the deep learning model has learned the correspondence between the input data (remaining driving range and trip planning information) and the output data (remaining SOC).

[0106] Step S406: Based on the actual SOC and remaining SOC of the vehicle battery, determine the target energy mode that the vehicle needs from multiple preset energy modes.

[0107] The energy mode is a pre-set energy mode, and the vehicle control strategy (such as temperature control strategy) is different under different energy modes. For example, the energy modes include, but are not limited to: the first energy mode (also known as the ample mode), which is used to prioritize the comfort of the occupants; the second energy mode (also known as the moderate mode), which is used to balance the comfort of the occupants and the energy consumption of the vehicle; and the third energy mode (also known as the stress mode), which is used to prioritize the vehicle's range and moderately sacrifice the comfort of the occupants.

[0108] As one possible implementation, if the vehicle's actual SOC is greater than a set first SOC, and the vehicle's remaining SOC upon reaching its destination is greater than a set second SOC, the target energy mode can be determined as the first energy mode. Here, the first SOC is greater than the second SOC.

[0109] As another possible implementation, if the vehicle's actual SOC is greater than a set third SOC and less than or equal to the first SOC, and the vehicle's remaining SOC upon reaching its destination is greater than the second SOC, then the target energy mode can be determined as the second energy mode. Specifically, the first SOC is greater than the third SOC, and the third SOC is greater than the second SOC.

[0110] As another possible implementation, if the vehicle's actual SOC is less than or equal to the third SOC, and the vehicle's remaining SOC when it reaches its destination is greater than the second SOC, the target energy mode can be determined as the third energy mode.

[0111] As another possible implementation, regardless of the actual SOC value of the vehicle, if the remaining SOC when the vehicle reaches its destination is less than the second SOC, the target energy mode can be determined as the third energy mode.

[0112] Step S407: Generate a temperature control strategy based on the target energy mode and the target thermal comfort state of the target seating area.

[0113] As one possible implementation, when the target energy mode is the first energy mode and there is only one target seating area, the required temperature of the target seating area can be determined based on its target thermal comfort state. Then, from the vehicle components associated with the target seating area, target vehicle components adapted to the target thermal comfort state are identified. Based on this required temperature, a temperature control strategy for controlling the target vehicle components is generated. Different target thermal comfort states have different required temperatures, and the correspondence between the required temperature and the target thermal comfort state can be preset.

[0114] The target vehicle components to be controlled may differ depending on the target thermal comfort state. For example, when the target thermal comfort state is "cold", the target vehicle components include the vehicle air conditioner and seat heating components; when the target thermal comfort state is "slightly cold", the target vehicle components include the vehicle air conditioner; when the target thermal comfort state is "moderate", the target vehicle components include the vehicle air conditioner; when the target thermal comfort state is "slightly hot", the target vehicle components include the vehicle air conditioner; and when the target thermal comfort state is "hot", the target vehicle components include the vehicle air conditioner and seat ventilation components.

[0115] As another possible implementation, when the target energy mode is the first energy mode and there are multiple target seating areas, the required temperature of each target seating area can be determined based on the target thermal comfort state of each target seating area. Then, the target vehicle components that are compatible with the target thermal comfort state of each target seating area can be determined from the vehicle components associated with each target seating area. After that, the required temperatures of multiple target seating areas can be weighted according to the weighted weights corresponding to the multiple target seating areas to obtain the target required temperature. Thus, a temperature control strategy for controlling the target vehicle components can be generated based on the target required temperature.

[0116] For example, the target required temperature T can be calculated using the following formula.final : T final = Σ(w i × T i ) / Σ(w i ); Among them, T i This refers to the required temperature for the i-th target seating area, w i It refers to the weighted weight of the i-th target seat region.

[0117] In some embodiments, the event recognition result may further include: confidence scores of multiple target seating areas, wherein each confidence score is used to indicate the uncertainty of the target thermal comfort state of the occupant in the corresponding target seating area. For example, each target decoding network may further include a confidence output network for outputting the confidence score of the target seating area corresponding to the target decoding network. In this application, the weighting weights corresponding to multiple target seating areas may also be adjusted according to the confidence scores of multiple target seating areas.

[0118] For example, when the confidence level of a target seat area indicates relatively high uncertainty, the weighting of that target seat area can be reduced to decrease its interference with the temperature control strategy; conversely, when the confidence level of a target seat area indicates relatively low uncertainty, the weighting of that target seat area can be increased. This avoids misleading the temperature control strategy output with low-confidence recognition results and ensures that the final temperature control strategy better matches the actual thermal comfort needs of different occupants, improving the accuracy and rationality of temperature control in multi-occupant scenarios.

[0119] In summary, when the battery has sufficient charge, the appropriate target vehicle components and required temperatures are determined based on the thermal comfort status of each target seating area, and a temperature control strategy is generated accordingly for single-occupant scenarios. In multi-occupant scenarios, weighted averages are set based on the degree of impact of occupants in multiple target seating areas on driving safety, and the required temperatures of multiple target seating areas are weighted and fused to obtain the final target required temperature. A temperature control strategy is then generated based on the target required temperature. Thus, under the premise of sufficient charge, the thermal comfort needs of different occupants can be maximized, and the comfort of occupants with a higher impact on driving safety can be prioritized, avoiding interference with the driver's normal driving state due to unreasonable priority settings of the temperature control strategy.

[0120] As another possible implementation, if the target energy mode is not the first energy mode, the temperature adjustment direction can be determined according to the target thermal comfort state, and the target vehicle component adapted to the target thermal comfort state can be determined from the vehicle components associated with the target seating area. If the target energy mode is the second energy mode, a temperature control strategy for controlling the target vehicle component can be generated according to the temperature adjustment direction and the set first temperature adjustment range. If the target energy mode is the third energy mode, a temperature control strategy for controlling the target vehicle component can be generated according to the temperature adjustment direction and the set second temperature adjustment range.

[0121] The direction of temperature adjustment is determined based on the target thermal comfort state. For example, when the target thermal comfort state is "cold" or "slightly cold", the direction of temperature adjustment is positive "+", while when the target thermal comfort state is "hot" or "slightly hot", the direction of temperature adjustment is negative "-".

[0122] The second temperature adjustment range is smaller than the first temperature adjustment range. For example, the first temperature adjustment range is within ±1.5°C, and the second temperature adjustment range is within ±1.0°C.

[0123] As an example, when the target energy mode is the second energy mode and the temperature adjustment direction is positive, the temperature control strategy can be used to indicate at least one of the following: the temperature adjustment amount of the vehicle air conditioner, turning off the vehicle air conditioner corresponding to the non-target seating area, and activating the seat heating function under the target seating area.

[0124] The temperature adjustment amount of the vehicle air conditioner is determined based on the target thermal comfort state and the first temperature adjustment range. For example, the temperature adjustment amount corresponding to "cold" is greater than the temperature adjustment amount corresponding to "slightly cold". Similarly, the temperature adjustment amount corresponding to "hot" is greater than the temperature adjustment amount corresponding to "slightly hot".

[0125] As another example, when the target energy mode is the second energy mode and the temperature adjustment direction is negative, the temperature control strategy can be used to indicate at least one of the following: the temperature adjustment amount of the vehicle air conditioner, turning off the vehicle air conditioner corresponding to the non-target seating area, and activating the seat ventilation function under the target seating area. The temperature adjustment amount of the vehicle air conditioner is determined based on the target thermal comfort state and the second temperature adjustment range.

[0126] As another example, when the target energy mode is the third energy mode, the temperature control strategy is used to indicate at least one of the following: the desired temperature in the driver's seat area within the target seating area, turning off the vehicle's air conditioning in the non-target seating areas, and triggering an energy-saving prompt; wherein the energy-saving prompt is used to indicate that the vehicle is currently in energy-saving mode. The desired temperature in the driver's seat area is determined based on the target thermal comfort state, the temperature adjustment direction, and the second temperature adjustment range.

[0127] Understandably, in the energy-constrained third energy mode, by prioritizing the temperature control needs of the driver's seat area, turning off the in-vehicle air conditioning in unoccupied areas, and triggering energy-saving prompts, the driver's thermal comfort can be ensured to remain within a comfortable range while minimizing unnecessary temperature control power consumption. This avoids driving inconvenience due to physical discomfort and prevents a sudden drop in driving experience caused by excessive temperature control in extreme range-constrained scenarios.

[0128] In summary, by setting a relatively wide first temperature adjustment range in the second energy mode and a relatively narrow second temperature adjustment range in the third energy mode, a stepped temperature control strategy can be implemented to balance the basic thermal comfort needs of passengers with the safety of remaining range during periods of relatively tight range, thereby achieving dynamic adaptation of temperature control experience and energy consumption priority under different energy reserve scenarios.

[0129] Step S408: Based on the temperature control strategy, perform operation control on the vehicle components associated with the target seating area.

[0130] For example, a target vehicle component adapted to the target thermal comfort state can be determined from vehicle components associated with the target seating area, and operation control can be performed on the target vehicle component according to a temperature control strategy; wherein the vehicle component includes at least one of the following: vehicle air conditioner, seat heating component, and seat ventilation component.

[0131] As an application scenario, the vehicle control method provided in this application can be applied to at least the following scenarios: Scenario 1: Family travel scenario: In scenarios where parents travel with children and the elderly, the temperature requirements of passengers of different ages vary significantly. For example, the elderly are afraid of the cold, while children have a fast metabolism and sweat easily. The vehicle can sense the thermal comfort status of each passenger through WiFi sensing nodes, and can realize personalized temperature control for each seating area.

[0132] Scenario 2: Long-distance driving scenario: When drivers maintain a seated posture for a long time, their blood circulation slows down, which may cause their hands and feet to feel cold; while passengers in the front passenger area may feel too hot due to direct sunlight. The vehicle can sense the thermal comfort status of each passenger through WiFi sensing nodes and maintain the comfort of passengers in each seating area.

[0133] Scenario 3: Vehicle Range Priority Scenario: When the vehicle's actual SOC is relatively low, the vehicle can intelligently optimize the energy consumption of the vehicle's air conditioning while ensuring the basic comfort of the occupants, thus balancing the comfort of the occupants with the vehicle's driving range.

[0134] The vehicle control method of this application combines the vehicle's current actual SOC and the predicted remaining SOC when the vehicle reaches its destination to select a suitable target energy mode, and combines the thermal comfort state of the target seating area to generate a temperature control strategy. This can achieve a dynamic balance between the comfort and energy consumption priority of temperature control in different range scenarios, avoiding insufficient vehicle range due to excessive power consumption caused by temperature control, and can maximize the adaptation to the vehicle's current energy reserve state while ensuring the thermal comfort of the occupants, thus taking into account both the comfort of the occupants and the safety of the range.

[0135] In any embodiment of this application, where occupant risk events include comfort-related risk events, this application provides a method such as... Figure 5 The personalized thermal comfort control scheme for occupants in a smart cockpit based on WiFi CSI signal sensing, as shown, mainly includes the following steps: Step 1: Deploy WiFi sensing nodes inside the vehicle and continuously collect CSI signals from each seating area in the cabin.

[0136] The hardware deployment includes: deploying one WiFi sensing node at the front and one at the rear of the cockpit; each WiFi sensing node is equipped with a 4×4 MIMO antenna array, with the antenna spacing designed to be half wavelength (approximately 6cm@2.4GHz); the WiFi sensing nodes can operate on both 2.4GHz and 5GHz dual-band frequencies; the WiFi sensing nodes have built-in edge computing modules and support local CSI preprocessing.

[0137] The data acquisition method includes: setting the sampling frequency to 200Hz (to meet the requirements for capturing micro-movements); each CSI signal contains complex CSI values ​​of all subcarriers (234 subcarriers in an 80MHz bandwidth); and simultaneously acquiring the CSI signals of the two WiFi sensing nodes before and after the acquisition for subsequent triangulation.

[0138] Taking a four-seater vehicle as an example, the occupant position sensing method includes: using the TDOA algorithm to calculate the occupant position based on the time difference of the CSI signals of the front and rear WiFi sensing nodes; dividing the cabin into four sensing areas: the driver's seat area, the front passenger seat area, the left rear seat area, and the right rear seat area; and marking the CSI signals according to the seat area to achieve multi-person signal separation and obtain the CSI local signals corresponding to multiple seat areas.

[0139] Step 2: Extract micro-motion features related to thermal comfort from the CSI local signal corresponding to each seating area.

[0140] Among them, the micro-movement types related to thermal comfort are defined as follows: cold-related movements: shrugging shoulders, rubbing hands, crossing legs, hugging arms; heat-related movements: fanning, stretching, loosening collar, wiping sweat; neutral state movements: normal sitting posture, slight activity.

[0141] The extraction methods for micro-motion features in each seating area include: 1. Time-frequency feature extraction: Using STFT to convert the CSI local signal of each seating area into a time-frequency spectrum; the frequency features of micro-motions are concentrated in the range of 0.5Hz-10Hz, separated from the frequency bands of physiological signals such as heart rate / respiration. Based on this frequency band, bandpass filtering is performed on the time-frequency spectrum to obtain the time-frequency features; 2. Spatial feature extraction: Utilizing the spatial diversity characteristics of the MIMO antenna array, the angle of arrival information is extracted; different micro-motions will cause different changes in the spatial dimension of the CSI local signal; based on the angle of arrival information, the spatial features of the CSI local signal are constructed; the spatial features are in matrix form, with the dimension being [number of antennas × number of subcarriers]; 3. Feature fusion: The time-frequency features and spatial features are fused to construct a three-dimensional feature tensor [T × F × S]; where T is the time dimension, F is the frequency dimension, and S is the spatial dimension.

[0142] Step 3: Use a multi-task learning thermal comfort prediction model to simultaneously predict the thermal comfort status of multiple occupants from the above three-dimensional feature tensor.

[0143] The model architecture design adopts an encoder-decoder structure, including an encoding network and multiple decoding networks.

[0144] The encoding network consists of: an input layer that receives micro-motion features of dimension [B, T, F, S, 4], where B is the batch size and 4 is the number of seat regions; a 3D convolutional layer 1 with a kernel size of (3,3,3), 64 output channels, and a Rectified Linear Unit (ReLU) activation function; a 3D convolutional layer 2 with a kernel size of (3,3,3), 128 output channels, and a ReLU activation function; a residual connection that adjusts the dimension of the output of 3D convolutional layer 1 by passing it through a 1×1×1 convolution and then adds it to the output of 3D convolutional layer 2; and global average pooling that compresses the features into a global feature set of [B, 128, 4] (referred to as common features in this application).

[0145] The component consists of multiple parallel decoding networks (each with the same structure): Taking the driver's seat area as an example, the decoding networks associated with the driver's seat area include: Attention module: using a self-attention mechanism, the Query is the micro-motion features of the driver's seat area, and the Key / Value is the global feature; Fully connected layer 1: the feature dimension mapping method is 128→64, the activation function is ReLU, and the regularization configuration is Dropout(0.3); Fully connected layer 2: the feature dimension mapping method is 64→32, and the activation function is ReLU; Thermal comfort output network: the feature dimension mapping method is 32→5, the activation function is Softmax, and it outputs the probability distribution of the following 5 thermal comfort states: cold (1), slightly cold (2), moderate (3), slightly hot (4), hot (5); Confidence output network: the feature dimension mapping method is 32→1, the activation function is Sigmoid, and it outputs the predicted confidence, which is used for subsequent weighted fusion.

[0146] The multi-task learning objectives are as follows: Main task: thermal comfort state classification (5 categories); Auxiliary task 1: micro-motion type recognition (10 categories); Auxiliary task 2: occupant presence detection (2 categories); The total loss function is designed as: L total = α·L comfort + β·L action + γ·L presence .

[0147] In summary, the encoding network can learn common patterns of CSI local signals in different seating areas (such as basic human movement patterns), reducing the number of model parameters; the decoding network can capture individual differences in different seating areas, enabling personalized predictions; multi-task learning improves the generalization ability of the main task by simultaneously training two auxiliary tasks: micro-motion recognition and occupant presence detection; and the confidence output network can identify prediction uncertainties and reduce the prediction weight of that seating area when the quality of the CSI local signal is poor (such as signal occlusion).

[0148] Step 4: Based on the predicted thermal comfort status, generate a personalized temperature control strategy to regulate at least one of the following: vehicle air conditioning, seat ventilation components, and seat heating components. This mainly includes the following sub-steps: Step 4.1: Thermal Comfort State - Temperature Mapping Model. Establish the mapping relationship between the thermal comfort state of occupants in each seating area and the air conditioning operating parameters: Thermal Comfort State 1 (Cold): Temperature setpoint for this seating area +2°C, turn on seat heating function (if applicable), increase air outlet temperature; Thermal Comfort State 2 (Slightly Cold): Temperature setpoint for this seating area +1°C, decrease airflow; Thermal Comfort State 3 (Moderate): Maintain current setting; Thermal Comfort State 4 (Slightly Hot): Temperature setpoint for this seating area -1°C, increase airflow; Thermal Comfort State 5 (Hot): Temperature setpoint for this seating area -2°C, turn on seat ventilation function (if applicable) to maximize cooling.

[0149] Step 4.2: Multi-occupant Coordination Strategy: When the comfort needs of multiple occupants conflict, a weighted coordination strategy can be adopted: the driver's weight is the highest (e.g., weight 1.0) to ensure driving safety; the front-seat passengers' weight is the second highest (e.g., weight 0.8); and the rear-seat passengers' weight is the lowest (e.g., weight 0.6). Ultimately, the air conditioning operating parameters are the weighted average of the required temperatures for each seating area.

[0150] Step 4.3: Smoothing Strategy: To avoid discomfort caused by frequent adjustments, a hysteresis mechanism is introduced: A threshold for changes in thermal comfort state is set: When the occupant's target thermal comfort state changes, the duration of the change is recorded. If the duration exceeds a first set duration (e.g., 30 seconds), temperature adjustment is triggered. A limit can be set for the magnitude of a single adjustment: a single temperature adjustment cannot exceed 0.5°C; and an adjustment interval can be set: at least a second set duration (e.g., 60 seconds) must be between two temperature adjustments.

[0151] Step 5: Vehicle Energy Co-optimization: Optimize air conditioning energy consumption to extend driving range while meeting passenger comfort requirements. This mainly includes the following sub-steps: Step 5.1: Energy Status Perception: Real-time acquisition of the vehicle's actual SOC, remaining range, and trip planning information; using a first SOC of 60%, a second SOC of 20%, and a third SOC of 30% as an example, energy modes are divided according to SOC levels: Ample Mode (actual SOC > 60%): Prioritize passenger comfort; Moderate Mode (30% < actual SOC ≤ 60%): Balance comfort and energy consumption; Stressful Mode (actual SOC ≤ 30%): Prioritize range, with some compromise on comfort.

[0152] Step 5.2: Energy consumption optimization strategy: (a) Ample mode: Adjust according to the comfort requirements in Step 4, without limiting the air conditioning power; (b) Moderate mode: Limit the adjustment range of the air conditioning temperature to within ±1.5°C; prioritize the use of seat heating / seat ventilation functions (energy efficiency ratio higher than the vehicle air conditioning); turn off the air conditioning in unoccupied areas; (c) Stressful mode: Limit the adjustment range of the air conditioning temperature to within ±1.0°C; ensure the comfort of only the driver's seat area; completely turn off the rear air conditioning when there is no one in the rear; remind the occupants that the current mode is energy saving.

[0153] Step 5.3: Predictive Adjustment: Based on navigation information, predict the energy consumption demand before the vehicle reaches its destination. During long-distance driving, dynamically adjust the comfort-energy balance strategy according to the remaining driving range and the actual SOC. If the remaining SOC is predicted to be less than 20% when reaching the destination, enter the energy-saving mode in advance.

[0154] Understandably, by deeply integrating the thermal comfort perception of different occupants with vehicle energy management, an intelligent balance between comfort and driving range can be achieved.

[0155] In summary, the personalized thermal comfort control solution provided in this application has at least the following advantages: It utilizes WiFi CSI signals for non-contact sensing of occupant thermal comfort status, requiring no operation or device wearing by the occupant; the vehicle can actively sense the occupant's temperature sensation, achieving a smart cockpit experience; it uses only the CSI signals sensed by WiFi sensing nodes, eliminating the risk of privacy leaks; the deep learning-based thermal comfort prediction model infers the occupant's thermal comfort status from micro-motion features, more closely approximating the occupant's actual feelings; and through location sensing and multi-person signal separation technology, it can simultaneously sense the thermal comfort status of occupants in multiple seating areas, achieving personalized control. It features zoned temperature control; deep integration with the vehicle's energy management system to optimize air conditioning energy consumption while ensuring comfort, with real-world testing showing a 15-20% reduction in air conditioning energy consumption and effectively extending driving range; prediction of future energy consumption needs based on trip planning information to adjust comfort-energy consumption strategies in advance, avoiding forced reductions in comfort due to insufficient battery power at the end of the trip; the ability to learn historical thermal preference data of specific occupants (i.e., training data associated with that specific occupant) to build a personalized thermal comfort prediction model, with predictions becoming increasingly accurate over time; and the ability to utilize existing in-vehicle WiFi sensing nodes (for vehicle-to-everything communication), eliminating the need for additional dedicated sensors and reducing hardware costs.

[0156] As one possible implementation method, Figure 6This is a schematic flowchart illustrating another vehicle control method provided as an exemplary embodiment of the present application. It should be noted that this vehicle control method can be executed alone, or it can be executed together with any embodiment of the present application or any possible implementation thereof, or it can be executed together with any technical solution in the related art. The embodiments of the present application do not impose any limitations on this.

[0157] like Figure 6 As shown, the vehicle control method may include the following steps S601 to S606: Step S601: Obtain the CSI signal obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin.

[0158] It should be noted that the explanation of step S601 can be found in the relevant description in any embodiment of this application, and will not be repeated here.

[0159] Step S602, in response to occupant risk events, including motion sickness, extracts multidimensional physiological characteristics of the occupants in the cabin from the CSI signal. These multidimensional physiological characteristics include, but are not limited to, at least two of the following: heart rate, heart rate variability (HRV) index, respiratory rate, microkinetic energy caused by friction between skin and clothing determined based on high-frequency components in the CSI signal, and head posture characteristics.

[0160] One method for extracting heart rate is, for example, performing a continuous wavelet transform (CWT) on the CSI signal in the heart rate frequency band and extracting the wavelet ridges to obtain the instantaneous heart rate time series HR(t). For instance, the formula for calculating HR(t) is: HR(t) = 60 × f instant (t), where f instant (t) refers to the instantaneous heart rate obtained from the wavelet ridge of the CSI signal at time t, in units of beats per second.

[0161] HRV (Heart Rate Value) refers to the minute fluctuations in the time interval between consecutive heartbeats. HRV includes time-domain and frequency-domain indicators. The time-domain indicators include the standard deviation of all normal to normal intervals (SDNN) and the root mean square of successive differences (RMSSD) between adjacent normal sinus intervals. The frequency-domain indicators include high-frequency (HF) power, low-frequency (LF) power, and the ratio of LF power to HF power.

[0162] The SDNN (also known as the RR interval standard deviation) is used to reflect overall autonomic nervous function. For example, the SDNN is calculated as follows: based on the R-peak detection algorithm, the peak position of the R-wave in the CSI signal of the heart rate band is detected; the time interval (also known as the RR interval) between the i-th and (i+1)-th R-wave peak positions is calculated: RR[i] = t R [i+1] - t R [i]; where t R [i+1] refers to the timestamp corresponding to the (i+1)th R-wave peak position, t R [i] refers to the timestamp corresponding to the i-th R-wave peak position, and RR[i] refers to the time interval between two adjacent R-wave peak positions; an RR interval sequence is generated based on the interval duration between two adjacent R-wave peak positions; and the SDNN is determined based on the standard deviation of the RR interval sequence.

[0163] RMSSD is used to reflect parasympathetic nerve activity. For example, RMSSD can be calculated by taking the root mean square of the difference between adjacent elements in the RR interval sequence.

[0164] Among them, low-frequency (e.g., 0.04-0.15Hz) power is used to reflect the mixed activity of sympathetic and parasympathetic nerves; high-frequency (e.g., 0.15-0.4Hz) power is used to reflect parasympathetic nerve activity; and the ratio of LF power to HF power (denoted as LF / HF) is used to reflect the balance between sympathetic and parasympathetic nerves.

[0165] The method for extracting respiratory rate (BR) is, for example, to perform peak detection on the CSI signal in the respiratory frequency band (e.g., 0.1-0.5Hz); calculate the average respiratory rate within a sliding window (30 seconds), and the calculation formula is: BR(t) = 60 × number of peaks / window duration (unit: breaths / min).

[0166] Among them, micro-kinetic energy E micro The extraction method for (also known as the sweating micro-motion index) is, for example: considering that sweating leads to increased friction between the skin and clothing, in this application, the high-frequency component CSI in the CSI signal can be used. high (e.g., 10-30Hz) energy change detection micro-motion energy, for example, the formula for calculating micro-motion energy is: E micro (t) = Σ|CSI high (t)|².

[0167] One method for extracting head posture features is to extract the occupant's head posture features, such as normal, forward tilt, side tilt, and backward tilt, based on the spatial dimension information of the CSI signal, in order to identify dangerous behaviors of the occupant, such as the occupant looking down for a long time (reading / looking at a mobile phone) or unstable head shaking.

[0168] In any embodiment of this application, before extracting multidimensional physiological features, the CSI signal may be preprocessed to extract the multidimensional physiological features of the occupants in the cabin from the preprocessed CSI signal; wherein, the preprocessing includes, but is not limited to, at least one of the following: The first item is to use a notch filter adapted to the vibration frequency of the vibrating components in the vehicle to filter the CSI signal in order to eliminate vibration noise introduced by the vibrating components; wherein, the center frequency of the notch filter is related to the rotational speed of the vibrating components, including the engine and / or motor.

[0169] In this application, considering the stable frequency of the vibrating component (approximately 20-100Hz), it can be eliminated using a notch filter. The center frequency of the notch filter is dynamically adjusted according to the rotational speed of the vibrating component. For example, the filtering formula for the notch filter is: H notch (t) = H raw (t) - Σ A n × sin(2π×f n ×t + φ n ); Among them, H notch (t) represents the filtered CSI signal (with vibration noise removed); H raw (t) represents the original CSI signal; A n f is the amplitude of the nth vibration component; n φ is the frequency of the nth vibration component (related to the rotational speed of the vibrating component); n It represents the phase of the nth vibration component.

[0170] The second step involves using a Kalman filter to perform fusion filtering on the CSI signal based on the motion data collected by the vehicle's Inertial Measurement Unit (IMU) (hereinafter referred to as IMU data) in order to eliminate vibration noise introduced by road bumps; wherein, the IMU is synchronized with the WiFi sensing node in time.

[0171] For example, an extended Kalman filter (EKF) can be used to fuse CSI signals and IMU data: State vector: X = [CSI signal, CSI vibration component, IMU acceleration]; Observation vector: Z = [raw value of CSI signal, IMU reading]; Prediction steps: X pred = F×X prev + B×u; Update steps: X new = X pred + K×(Z - H×X pred ); Where K is the Kalman gain, which is adaptively adjusted to achieve optimal estimation of the vibration components; X pred X is the predicted state vector; F is the state transition matrix; X is the predicted state vector. prev B is the state vector of the previous time step; B is the control input matrix; u is the control input vector (which can come from IMU data).

[0172] The third step involves using a Butterworth filter adapted to the signal frequency band corresponding to the multidimensional physiological characteristics to filter the CSI signal in order to eliminate out-of-band interference signals; the passband range of the Butterworth filter is determined based on the occupant's motion state.

[0173] For example, a 4th-order Butterworth filter can be used for the heartbeat signal frequency band (0.8-2.0 Hz, corresponding to 48-120 bpm); a 2nd-order Butterworth filter can be used for the respiratory signal frequency band (0.1-0.5 Hz, corresponding to 6-30 bpm). The adaptive adjustment of the Butterworth filter's bandpass range can be achieved, for example, by narrowing the bandpass range when the occupant is detected to be at rest, in order to improve the signal-to-noise ratio.

[0174] Fourthly, motion detection is performed on the CSI signal to label target segments (also known as large motion segments) within the CSI signal. The target segment indicates an occupant's movement amplitude exceeding a set amplitude. The labeling is used to reduce the weight of the target segment during multidimensional physiological feature extraction, and / or, the labeling is used for interpolation and filling of the target segment based on its neighboring segments. In other words, this application uses motion detection algorithms to identify large occupant movements (such as turning the head, reaching out, etc.); large motion segments are labeled to reduce their weight or perform interpolation in subsequent analysis, eliminating occupant motion artifacts.

[0175] The fifth step involves using target correction parameters to perform phase correction processing on the CSI signal in order to eliminate nonlinear errors caused by temperature drift.

[0176] For example, the target correction parameters are determined as follows: The original CSI signal collected by the WiFi sensing node in the most recent correction period is obtained. The original CSI signal is in matrix form, and the element in the i-th row and j-th column of the matrix indicates the complex channel response data of the j-th subcarrier collected by the i-th receiving antenna in the antenna array of the WiFi sensing node. A phase correction model is constructed with the correction parameters as variables. The phase correction model is used to subtract the temperature drift component determined by the correction parameters from the phase components of the original CSI signal. The phase correction model is solved with the objective function of minimizing the first-order difference variance in the time dimension of the CSI phase sequence corresponding to the same subcarrier and the same receiving antenna after correction by the phase correction model to obtain the target correction parameters. The CSI phase sequence includes the phase components of each original CSI signal collected in the most recent correction period.

[0177] For example, the following dual-anchor correction method can be used to perform phase correction on the CSI signal: φ corrected (k) = φ raw (k) - [a1×(k-k0) + a2×(k-k0) 2 + b]; Where, φ corrected (k) is the corrected phase; φ raw (k) is the original phase before correction; k is the subcarrier index; k0 is the reference subcarrier index; a1, a2, b are correction parameters, which are determined by minimizing the variance of the CSI phase of adjacent time slots.

[0178] In summary, various preprocessing strategies can eliminate the impact of multi-source vibration and environmental interference on CSI signals in vehicle scenarios, improve the purity of subsequently extracted multi-dimensional physiological features, and thus improve the detection accuracy of subsequent occupant risk events.

[0179] Step S603: Obtain the reference physiological characteristics of the occupant; wherein, the reference physiological characteristics are determined based on the occupant's historical physiological data.

[0180] The extraction method for reference physiological features is similar to the extraction method for the aforementioned multidimensional physiological features, and will not be elaborated upon here.

[0181] Step S604: Determine the difference data between the reference physiological characteristics and the multidimensional physiological characteristics.

[0182] In the embodiments of this application, the difference data between the occupant's reference physiological characteristics and multidimensional physiological characteristics can be calculated.

[0183] Step S605: Detect motion sickness events based on the difference data to obtain event detection results; wherein, the event detection results are used to indicate the probability of a passenger experiencing various motion sickness states under the current time period.

[0184] For example, a motion sickness prediction model from the field of artificial intelligence can be used to detect motion sickness events based on differential data and output a probability distribution. This probability distribution indicates the probability of a passenger experiencing various motion sickness states during the current time period. The motion sickness prediction model has learned the correspondence between input data (multidimensional physiological characteristics) and output data (passenger's motion sickness state).

[0185] Step S606: Control the vehicle to execute a vehicle control strategy adapted to the event detection results.

[0186] For example, if the event detection result indicates that the occupant is not carsick, and the probability of being in a normal state is much higher than the probability of carsickness symptoms / imminent vomiting, the current vehicle control strategy can be maintained; however, if the event detection result indicates that the occupant is experiencing carsickness symptoms or is about to vomit, the vehicle can be controlled to execute a warning strategy to trigger a carsickness warning prompt.

[0187] The vehicle control method of this application adopts personalized reference physiological characteristics generated based on the occupant's own historical physiological data as a baseline. It completes motion sickness event detection by calculating the difference data between the occupant's multidimensional physiological characteristics and the baseline. It can adapt to the physiological baseline differences of occupants of different ages and physical conditions, avoid the problem of misjudgment and missed judgment of general thresholds for different occupants, and thus improve the individual adaptability and recognition accuracy of motion sickness event detection.

[0188] As one possible implementation method, Figure 7This is a schematic flowchart illustrating another vehicle control method provided as an exemplary embodiment of the present application. It should be noted that this vehicle control method can be executed alone, or it can be executed together with any embodiment of the present application or any possible implementation thereof, or it can be executed together with any technical solution in related technologies. The embodiments of the present application do not impose any limitations on this.

[0189] like Figure 7 As shown, the vehicle control method may include the following steps S701 to S705: Step S701: Obtain the CSI signal obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin.

[0190] Step S702, in response to occupant risk events, including motion sickness, extracts multidimensional physiological characteristics of the occupants in the cabin from the CSI signal. These multidimensional physiological characteristics include, but are not limited to, at least two of the following: heart rate, heart rate variability (HRV) index, respiratory rate, microkinetic energy caused by friction between skin and clothing determined based on high-frequency components in the CSI signal, and head posture characteristics.

[0191] Step S703: Detect motion sickness events based on multidimensional physiological characteristics and obtain event detection results; wherein, the event detection results are used to indicate the probability of a passenger experiencing various motion sickness states under the current time period.

[0192] It should be noted that the explanations of steps S701 to S703 can be found in the relevant descriptions in any embodiment of this application, and will not be repeated here.

[0193] Step S704: Based on the probabilities of various motion sickness states in the event detection results, determine a first target warning strategy from the preset multi-level first warning strategies. The various motion sickness states include, but are not limited to: a first motion sickness state and a second motion sickness state, wherein the degree of motion sickness in the second motion sickness state is higher than that in the first motion sickness state. For example, the second motion sickness state may be, for instance, impending vomiting, and the first motion sickness state may be, for instance, pre-motion sickness symptoms.

[0194] In this embodiment of the application, a first target warning strategy can be determined from multiple levels of first warning strategies based on the probability of various motion sickness states in the event detection results.

[0195] As one possible implementation, if the probability of a first motion sickness state in the event detection results is higher than a first probability threshold, the first target warning strategy can be used to prompt passengers to wear smart glasses and / or trigger a motion sickness risk warning. The first probability threshold can be a pre-set probability threshold; for example, the value of the first probability threshold is 0.4.

[0196] For example, taking the first state of motion sickness as a precursor to motion sickness, the probability of marking the first state of motion sickness is P. warning Assuming the first probability threshold is 0.4, and P warning If the value is greater than 0.4, a Level 1 warning will be issued, and passengers are advised to wear smart glasses.

[0197] As another possible implementation, if the probability of a second motion sickness state in the event detection result is higher than a second probability threshold, a first target warning strategy is used to instruct the activation of a first visual intervention mechanism in the smart glasses to alleviate the passenger's anxiety and nausea. The second probability threshold is also a pre-set probability threshold, and this second probability threshold is higher than the first probability threshold; for example, the value of the second probability threshold is 0.6.

[0198] For example, taking the second motion sickness state as impending vomiting as an example, the probability of marking the second motion sickness state as P sickness Assuming the second probability threshold is 0.6, and P sickness If the value is greater than 0.6, a level 2 warning will be issued, and the first visual intervention mechanism in the smart glasses will be activated.

[0199] In some embodiments, the first visual intervention mechanism includes, but is not limited to, at least one of the following: displaying a fixed reference point to enable the occupant's visual system to establish a stable reference frame; displaying breathing guidance material to guide the occupant to take deep breaths; and adjusting at least one of the color temperature, brightness, and hue of the displayed content. The fixed reference point can remain spatially fixed with slight head movements of the occupant, helping the occupant's visual system to establish a stable reference frame.

[0200] In some embodiments, the fixed reference point is displayed in a manner such as: tracking the occupant's head posture using the IMU of the smart glasses and obtaining the fixed display position of the fixed reference point in virtual space. Then, based on the head posture, a set compensation coefficient, and the fixed display position, the actual display position of the fixed reference point on the smart glasses can be determined. Therefore, in this application, the fixed reference point can be displayed on the smart glasses based on this actual display position to maintain its fixed position in virtual space. For example, the actual display position of the fixed reference point on the smart glasses can be calculated using the following formula: P display = P fixed - ΔHeadPose × K compensation ; Among them, P fixed ΔHeadPose is the fixed display position of a fixed reference point in virtual space, that is, the ideal spatial coordinates that the fixed reference point should be in relative to the occupant's head or vehicle coordinate system; ΔHeadPose is the change in the occupant's head posture; K compensationIt is the calibrated compensation coefficient; P display It is the position of the fixed reference point to be displayed on the smart glasses, which is referred to as the actual display position in this application.

[0201] Therefore, when the occupant's head moves, the display position of the fixed reference point can be adjusted to keep the fixed reference point in a fixed position in the virtual space, thereby providing a stable visual reference and alleviating the occupant's nausea.

[0202] The display forms of the breathing guidance material include, but are not limited to, animation, images, and text. When the display form of the breathing guidance material is an image or animation, this application does not impose specific restrictions on the shape and size of the breathing guidance material.

[0203] In the case where the breathing guidance material is displayed in the form of animation, this application does not impose specific restrictions on the breathing rhythm of the breathing guidance material. For example, the breathing rhythm is 12 times per minute (e.g., 5 seconds of inhalation, 5 seconds of exhalation).

[0204] In some embodiments, the animation is exemplified by the display of breathing guidance material. The display method of the breathing guidance material is, for example, to periodically contract and expand the displayed reference graphic with a set contraction cycle and expansion cycle; wherein, the contraction process is used to guide the occupant to exhale; and the expansion process is used to guide the occupant to inhale.

[0205] This application does not specifically display the shape and size of the reference graphic. For example, the shape of the reference graphic may be a circle, ellipse, cube, etc. The duration of the contraction cycle and the expansion cycle may be equal or unequal; this application does not impose specific limitations on this. For example, the duration of both the contraction cycle and the expansion cycle may be 5 seconds.

[0206] Therefore, by displaying breathing guidance materials in smart glasses, passengers can be guided to take deep breaths, alleviating their anxiety and nausea.

[0207] In some embodiments, if the probability of a second motion sickness state is higher than a second probability threshold, the first target warning strategy is further used to instruct the vehicle to broadcast breathing prompts associated with breathing guidance material.

[0208] Understandably, by displaying breathing guidance materials and simultaneously broadcasting breathing prompts, passengers can actively adjust their breathing rhythm, quickly alleviate the discomfort caused by the misalignment between the vestibular and visual systems, and intervene in the development of motion sickness symptoms in advance without relying on additional hardware, reducing the probability of passengers experiencing severe motion sickness reactions later. This effectively improves the riding comfort of people prone to motion sickness without interrupting the normal driving rhythm.

[0209] Adjusting the color temperature of the smart glasses' display content includes: warming the color temperature of the display content to reduce blue light; adjusting the brightness of the display content includes: reducing the brightness of the display content to reduce stimulation; and adjusting the hue of the display content includes: using soothing hues such as light blue or green.

[0210] In summary, different visual interference mechanisms can be used to alleviate passengers' anxiety and nausea, which can not only improve the flexibility and applicability of the method, but also enhance the passengers' riding experience.

[0211] As another possible implementation, when the probability of a second motion sickness state in the event detection result is higher than a third probability threshold, the first target warning strategy is used to indicate, but is not limited to, at least one of the following: activating the second visual intervention mechanism in the smart glasses to alleviate the occupant's motion sickness; triggering a stop and rest prompt; triggering a window ventilation prompt. The third probability threshold is also a pre-set probability threshold, and the third probability threshold is higher than the second probability threshold; the intervention intensity of the second visual intervention mechanism is higher than that of the first visual intervention mechanism.

[0212] For example, taking the second motion sickness state as impending vomiting as an example, the probability of marking the second motion sickness state as P sickness Assuming the second probability threshold is 0.8, and P sickness If the reading is >0.8, an emergency warning will be issued, the second visual intervention mechanism in the smart glasses will be activated, and voice suggestions will be given to stop the car for a rest and / or open the windows for ventilation.

[0213] In some embodiments, the second visual intervention mechanism includes, but is not limited to, at least one of the following: displaying a virtual horizon line to simulate a horizon; reducing the refresh rate of the displayed content; performing motion compensation on the displayed content; and blurring the visual field edges of the displayed content to guide the occupant to focus on the central area of ​​the displayed content.

[0214] The virtual horizon line, used to simulate the horizon, always remains horizontal and aligned with the direction of gravity to help occupants align their vestibular and visual systems.

[0215] Reducing the refresh rate of the display content on smart glasses can decrease the conflict between the occupant's visual and vestibular systems; motion compensation for the display content on smart glasses can also decrease the conflict between the occupant's visual and vestibular systems.

[0216] Blurring the edges of the displayed content can reduce motion information in the occupant's peripheral vision and guide the occupant to focus on the central area of ​​the displayed content.

[0217] In some embodiments, the first visual intervention mechanism may also be included in the second visual intervention mechanism, that is, the second visual intervention mechanism may include at least one of the following: displaying a fixed reference point to enable the occupant's visual system to establish a stable reference frame; displaying breathing guidance material to guide the occupant to take deep breaths; adjusting at least one of the color temperature, brightness, and hue of the displayed content; displaying a virtual horizon line to simulate a horizon; reducing the refresh rate of the displayed content; performing motion compensation on the displayed content; and blurring the visual field edges of the displayed content to guide the occupant to focus on the central area of ​​the displayed content.

[0218] In summary, by matching corresponding graded early warning strategies with three progressively increasing probability thresholds, only mild risk warnings are given in the early stages of motion sickness to guide passengers to take preventative measures. In the moderate stages of motion sickness, low-intensity visual intervention mechanisms are activated to alleviate passengers' anxiety and nausea. In the severe stages of motion sickness, high-intensity visual intervention mechanisms, rest reminders, and window ventilation reminders are employed. This approach perfectly matches the physiological changes in passengers as their motion sickness gradually worsens. It avoids excessive intervention in the early stages from disrupting the passengers' normal riding experience, while accurately delivering appropriate levels of relief measures at different stages of motion sickness, reducing the probability of passengers experiencing worsening motion sickness.

[0219] Step S705: Control the vehicle to execute the first target warning strategy to trigger a motion sickness warning prompt.

[0220] In some embodiments, if the probability of experiencing a first motion sickness state is higher than a first probability threshold, the passenger may be prompted to wear smart glasses and / or a motion sickness risk warning may be triggered. For example, the passenger may be advised to wear smart glasses via voice, and a motion sickness risk warning may be displayed on the dashboard.

[0221] In some embodiments, when the probability of a second motion sickness state is higher than a second probability threshold, and the occupant is wearing smart glasses, the smart glasses can be controlled to activate a first visual intervention mechanism to alleviate the occupant's anxiety and nausea. Exemplarily, the smart glasses can be controlled to perform at least one of the following: display a fixed reference point, display breathing guidance material, and adjust the color temperature and brightness of the displayed content.

[0222] In some embodiments, when the probability of a second motion sickness state is higher than a third probability threshold, and the occupant is wearing smart glasses, at least one of the following can be performed: controlling the smart glasses to activate a second visual intervention mechanism to alleviate the occupant's motion sickness; triggering a stop and rest prompt; triggering a window ventilation prompt. For example, the smart glasses can be controlled to display a virtual horizon line, reduce the refresh rate of the displayed content, perform motion compensation on the displayed content, and provide voice suggestions to the occupant to stop and rest and to open the windows for ventilation.

[0223] The vehicle control method of this application embodiment can trigger a graded warning prompt adapted to the current motion sickness state of the passenger, so as to give a warning prompt of corresponding intensity in a timely manner when the motion sickness state gradually worsens, help the passenger take intervention operations in advance, and reduce the probability of motion sickness discomfort.

[0224] As one possible implementation method, Figure 8 This is a schematic flowchart illustrating another vehicle control method provided as an exemplary embodiment of this application. It should be noted that this vehicle control method can be executed alone, or it can be executed together with any embodiment of this application or any possible implementation thereof, or it can be executed together with any technical solution in related technologies. This application does not impose any limitations on this.

[0225] like Figure 8 As shown, the vehicle control method may include the following steps S801 to S807: Step S801: Obtain the CSI signal obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin.

[0226] Step S802, in response to occupant risk events, including motion sickness events, extract multidimensional physiological characteristics of occupants in the cabin from the CSI signal.

[0227] Step S803: Detect motion sickness events based on multidimensional physiological characteristics and obtain event detection results; wherein, the event detection results are used to indicate the probability of a passenger experiencing various motion sickness states under a motion sickness event during the current time period.

[0228] Step S804: Based on the probability of various motion sickness states in the event detection results, determine the first target warning strategy from the preset multi-level first warning strategies, and control the vehicle to execute the first target warning strategy to trigger a motion sickness warning prompt.

[0229] It should be noted that the explanations of steps S801 to S804 can be found in the relevant descriptions in any embodiment of this application, and will not be repeated here.

[0230] Step S805: Obtain the latest physiological characteristics of the occupants after the vehicle executes the first warning strategy. These latest physiological characteristics are extracted from the latest CSI signal sensed by the WiFi sensing node. The method of obtaining these characteristics is similar to that of obtaining the multi-dimensional physiological characteristics described above, and will not be elaborated upon here.

[0231] Step S806: Based on the latest physiological characteristics, determine whether the degree of motion sickness of the passengers has decreased.

[0232] For example, if the BR (breast temperature) of the occupant decreases and the HRV (human vegetative state) index recovers, the degree of motion sickness is determined to have decreased; if the HRV index continues to deteriorate, the degree of motion sickness is determined to have not decreased.

[0233] For example, the following formula can be used to calculate the intervention effect index Score, so as to intuitively determine the trend of change in the degree of motion sickness of passengers based on the positive and negative values ​​and magnitude of the intervention effect index, and thus determine whether the degree of motion sickness of passengers has decreased.

[0234] Score = w1×ΔBR + w2×ΔHRV + w3×ΔMicroMove; Wherein, ΔBR refers to the change in respiratory rate, ΔHRV refers to the change in HRV index, ΔMicroMove refers to the change in micromotion energy, where micromotion energy refers to the micromotion energy caused by friction between skin and clothing, determined based on the high-frequency components in the CSI signal; w1, w2, and w3 are preset weighting weights.

[0235] Step S807: If not, determine the second target warning strategy from the multi-level first warning strategy and control the vehicle to execute the second target warning strategy, wherein the warning level of the second target warning strategy is higher than that of the first target warning strategy.

[0236] In this application, the level difference between the early warning level of the second target early warning strategy and the early warning level of the first target prediction strategy is not specifically limited. That is, the level difference can be one level, two levels, three levels, etc.

[0237] In this embodiment of the application, if the degree of motion sickness of the occupants does not decrease after the vehicle executes the first warning strategy, a second target warning strategy can be determined from the multi-level first warning strategies, and the vehicle can be controlled to execute the second target warning strategy to alleviate the degree of motion sickness of the occupants.

[0238] In any embodiment of this application, feedback information and parameter adjustment information of the occupants regarding at least one first warning strategy can also be obtained, and the at least one first warning strategy can be updated based on the feedback information and parameter adjustment information to match the occupants' latest strategy preferences.

[0239] Understandably, by continuously collecting feedback information from passengers on various warning strategies and personalized parameter adjustment information, and dynamically iterating and updating the configuration details of the corresponding warning strategies, such as the intervention brightness of smart glasses and the display size of fixed reference points, it is possible to continuously learn and adapt to the motion sickness intervention preferences of different passengers. This avoids the problem that fixed preset warning strategies cannot adapt to the usage habits of different passengers, so that the adaptability of the graded warning strategies and visual intervention mechanisms continuously improves with the duration of use, further enhancing the motion sickness relief effect and passenger experience.

[0240] In any embodiment of this application, contextual information of the vehicle when executing the first target warning strategy can also be obtained. The contextual information includes the vehicle's road condition information and the occupant's behavior information. In this application, a personalized motion sickness risk profile of the occupant can be generated based on the road condition information and behavior information. The motion sickness risk profile is used to execute the first target warning strategy when the vehicle is traveling in similar contextual information.

[0241] Road condition information refers to road environment data collected during vehicle operation, including but not limited to road curvature, road surface bumpiness, current acceleration / deceleration frequency, attributes of enclosed scenarios such as tunnels / elevated roads, and road slope, which directly affect occupant vestibular perception. Occupant behavior information refers to occupant status data collected by onboard sensors (including WiFi sensing nodes), including but not limited to head posture, torso posture, and micro-movement types.

[0242] In this application, a correspondence can be established between the passenger's motion sickness risk profile and the first target prediction strategy. Thus, when the vehicle subsequently travels to a road condition similar to the historical scenario and a context environment matching the passenger's behavior, the pre-stored first target warning strategy can be directly retrieved to quickly trigger intervention without repeating the real-time inference calculation of the strategy. This significantly reduces the onboard computing power consumption and further compresses the warning response latency, avoiding the problem of missing the opportunity for motion sickness intervention due to the lag in strategy triggering, and achieving a personalized active protection effect against motion sickness.

[0243] As an application scenario, the vehicle control method provided in this application can be applied to at least the following scenarios: Scenario 1: Family travel scenario where children are prone to motion sickness: In scenarios where parents and children travel together, the incidence of motion sickness in children is as high as 40-50%. The vehicle uses WiFi sensing nodes to detect changes in the child's breathing rate, providing an early warning before motion sickness occurs, and suggesting that the child wear smart glasses in the cabin for visual intervention to prevent the child from vomiting and soiling the car interior.

[0244] Scenario 2: Long-distance travel scenario: Due to the long travel time, even passengers who do not usually get motion sickness may experience discomfort. Therefore, the vehicle can continuously monitor the multidimensional physiological characteristics of passengers through WiFi sensing nodes. When motion sickness tendencies are detected, the vehicle can proactively recommend intervention measures (such as first visual intervention mechanism and second visual intervention mechanism).

[0245] Scenario 3: Mountain Road / Bumpy Road Scenarios: Frequent turns and bumps in the vehicle can easily trigger motion sickness. Based on the vehicle's real-time location and road conditions, high-risk road sections can be predicted, and a preventative first-visual intervention mechanism can be activated in advance. Specifically, this application can also determine whether the vehicle is entering / about to enter a high-risk road section based on its actual location and road conditions. If so, passengers will be prompted to wear smart glasses, and the first-visual intervention mechanism within the smart glasses will be activated.

[0246] Scenario 4: Passenger reading / looking at mobile phone scenario: Based on the CSI signal sensed by the WiFi sensing node, it can be determined that the passenger is reading / looking at their mobile phone, which indicates that the passenger is engaging in a high-risk behavior for motion sickness. At this time, it can be proactively suggested that the passenger wear smart glasses or enable Augmented Reality (AR) mode to display fixed reference points.

[0247] The vehicle control method of this application, when the early warning intervention fails to alleviate motion sickness, switches to a second target early warning strategy with a higher early warning level. This allows for adaptive upgrading of the intervention intensity without manual operation by the passenger, avoiding the problem that a fixed-level early warning strategy cannot adapt to the dynamic changes in the passenger's motion sickness state. When the passenger's motion sickness symptoms continue to worsen, a higher intensity relief intervention can be output in a timely manner, improving the success rate of motion sickness intervention.

[0248] In any embodiment of this application, when occupant risk events include motion sickness, this application provides an occupant motion sickness prediction and visual intervention scheme based on WiFi CSI signal sensing and smart glasses collaboration, mainly including the following steps: Step 1': Deploy WiFi sensing nodes in the cabin and continuously collect CSI signals from each seating area to extract multidimensional physiological characteristics of the occupants from the CSI signals.

[0249] The hardware deployment method includes: deploying one WiFi sensing node (Tx) at the front of the cockpit (such as the instrument panel); and deploying one WiFi sensing node (Rx) at the rear of the cockpit (such as the rear window); both Tx and Rx are equipped with 2×2 MIMO antennas; the collected CSI signals cover the seating area of ​​all occupants.

[0250] The data acquisition methods include: sampling frequency: 200Hz; each CSI signal contains complex CSI values ​​of multiple subcarriers; continuous acquisition to form time series data for each seating area.

[0251] The occupant position separation method includes: using CSI spatial characteristics (different signal attenuation modes in different positions) to separate the CSI signals of each occupant's seat area; dividing the cabin into 4 perceptible seat areas: driver's seat area, front passenger seat area, left rear seat area, and right rear seat area.

[0252] Step 2': Extraction of physiological features of motion sickness precursors: Extract multidimensional physiological features related to motion sickness from the preprocessed CSI signals of each seat area.

[0253] The mechanism of motion sickness is explained as follows: Motion sickness is caused by a sensory conflict between the visual system (movement seen by the eyes) and the vestibular system (movement felt by the body). Precursors to motion sickness typically include: an abnormally rapid respiratory rate (increasing from 12-16 breaths / min to 20-25 breaths / min); a slight increase in heart rate; a decrease in HRV (human vegetative state); and increased skin electrical activity (a precursor to sweating, manifested as increased microkinetic activity).

[0254] In this application, the following physiological features are mainly extracted: respiratory rate BR(t), heart rate HR(t), SDNN(t) and RMSSD(t) in HRV indices, microkinetic energy, and microkinetic energy E. micro (t), Head Pose(t).

[0255] Step 3': Motion Sickness Prediction Model: Using deep learning technology, this model comprehensively predicts the probability of various types of motion sickness based on multi-dimensional physiological features. It mainly includes the following sub-steps: Step 3.1': Feature Vector Construction: Construct a time series feature vector for each occupant: F(t) = [BR(t), HR(t), SDNN(t), RMSSD(t), E micro [(t), HeadPose(t)]. For example, a sliding window (e.g., 5 minutes) can be used to construct the feature vector.

[0256] Step 3.2': The model architecture of the motion sickness prediction model, including the following network layers: Input layer: Receives a feature matrix with dimensions [B, T, 6], where B is the batch size, representing the number of feature vectors input to the motion sickness prediction model in a single operation; T refers to the time steps, representing the time length of the feature vectors (i.e., the number of sampling points within the sliding window); and 6 refers to the feature dimension, corresponding to the feature vectors [BR(t), HR(t), SDNN(t), RMSSD(t), E...]. micro (t), HeadPose(t)].

[0257] One-dimensional convolutional neural network (1DCNN) layers are used to extract local temporal features. They consist of two sets of concatenated convolutional units: the first set is a one-dimensional convolutional layer Conv1D(32, kernel=5) with 32 convolutional kernels and a kernel size of 5, followed by a ReLU activation function to complete the nonlinear transformation; the second set is a one-dimensional convolutional layer Conv1D(64, kernel=3) with 64 convolutional kernels and a kernel size of 3, also followed by a ReLU activation function, which gradually compresses redundant information and enhances the discriminative power of local temporal features.

[0258] Bidirectional Long Short-Term Memory (Bi-LSTM) layers are used to capture the front-to-back dependencies in long-term features. They consist of two stacked Bi-LSTM (64) layers with a hidden dimension of 64, which can simultaneously mine the long-term variation patterns of motion sickness in time series from both forward and reverse directions.

[0259] The attention layer is used to automatically weight and allocate feature weights, focusing on key time segments where motion sickness changes abruptly, and filtering out redundant time-series information that remains stable and unchanged.

[0260] Fully Connected (FC) layer: It consists of three fully connected units connected in series, namely FC (128) with dimension 128, FC (64) with dimension 64, and FC (3) with dimension 3. Finally, it is connected to the normalized exponent Softmax function to map the output value to the probability value in the range of 0-1.

[0261] Output layer: Outputs the probabilities of the three types of states [P] normal , P warning , P sickness ]; where P normal : Probability of the normal state; P warning : Probability of motion sickness precursors (intervention recommended); P sickness The probability of vomiting (requiring emergency intervention).

[0262] Step 3.3': Threshold setting: P warning >0.4: Issue a Level 1 warning; wearing smart glasses is recommended. sickness >0.6: Issue a Level 2 warning and automatically initiate visual intervention; P sickness >0.8: An emergency warning is issued, and it is recommended to stop and rest.

[0263] Step 3.4': Personalized Baseline Learning: The system records the historical physiological data of each occupant and extracts each member's personal baseline (BR) from the historical physiological data. baseline HR baseline HRV baseline (Referring to the reference physiological characteristics in this application), thereby improving the accuracy of prediction results by comparing the individual baseline rather than an absolute threshold during prediction.

[0264] Step 4': Visual intervention mechanism of smart glasses: Based on the event detection results of motion sickness, multi-channel visual intervention is implemented through smart glasses, and a graded early warning strategy (or intervention strategy) is executed; the intervention channel design is shown in Table 1, and the graded early warning strategy design is shown in Table 2.

[0265] Table 1 Different intervention channels under visual intervention mechanism

[0266] Table 2. Tiered Early Warning Strategy

[0267] The principle behind the fixed reference point is as follows: the IMU (gyroscope + accelerometer) of the smart glasses is used to track the head posture and calculate the fixed display position of the fixed reference point in the virtual space; the display position of the fixed reference point is adjusted in real time according to the head posture to keep it fixed in the virtual space.

[0268] Step 5: Closed-loop feedback and strategy optimization: Adjust the early warning strategy in real time based on the intervention effect and continuously optimize the model.

[0269] The intervention effect assessment includes: continuous monitoring of changes in physiological indicators after intervention; if BR decreases and HRV indicators recover after intervention, the intervention is deemed effective; if HRV indicators continue to deteriorate after intervention, the intervention is deemed ineffective, and the warning level / intervention level is upgraded.

[0270] Among them, real-time policy adjustment: effective intervention: maintain the current level policy; ineffective intervention: upgrade to a higher level policy; excessive intervention (such as occupant voluntarily shutting down): reduce the intensity of intervention.

[0271] Among them, personalized preference learning: records the passenger's response to different strategies; records the parameters actively adjusted by the passenger (such as brightness preference, size of fixed reference point); establishes a personalized intervention model, and automatically applies the passenger's preference settings next time.

[0272] Among them, behavioral pattern learning involves learning under what conditions passengers are prone to motion sickness (such as on mountain roads or while reading); initiating preventive intervention strategies in advance under similar conditions; and establishing a motion sickness risk profile for passengers.

[0273] In summary, the technical solution provided in this application has at least the following beneficial effects: It eliminates the need for motion sickness medication, avoiding side effects such as drowsiness and dry mouth, making it particularly suitable for individuals who cannot take medication; it detects pre-motion sickness physiological signals in passengers via WiFi sensing nodes, issuing warnings 3-5 minutes before vomiting occurs, thus gaining valuable time for intervention; it actively applies a multi-channel visual intervention mechanism through smart glasses, rather than passively waiting for symptoms to subside, resulting in better intervention effects; the multi-channel combination of fixed reference points, virtual horizon lines, anti-shake display, breathing guidance, and color temperature adjustment can alleviate motion sickness symptoms from multiple dimensions; it utilizes the in-cabin WiFi infrastructure to detect passengers' motion sickness status, eliminating the need for additional sensors and providing a burden-free user experience; it uses only WiFi CSI signals to identify motion sickness status, without collecting sensitive information such as images and sounds; it adjusts strategies in real-time based on intervention effects to ensure optimal intervention results; and by combining vehicle location and road condition information, it can predict high-risk road sections (such as mountain roads and curves) and initiate preventative intervention strategies in advance, improving the passenger experience.

[0274] As one possible implementation method, Figure 9 This is a schematic flowchart illustrating another vehicle control method provided as an exemplary embodiment of this application. It should be noted that this vehicle control method can be executed alone, or it can be executed together with any embodiment or possible implementation thereof in this application, or it can be executed together with any technical solution in related technologies. This application does not impose any limitations on this.

[0275] like Figure 9 As shown, the vehicle control method may include the following steps S901 to S906: Step S901: Obtain the CSI signal obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin.

[0276] Step S902: In response to occupant status events, including health-related risk events, extract multidimensional physiological characteristics of the occupants in the cabin from the CSI signal. The multidimensional physiological characteristics associated with health-related risk events and those associated with motion sickness events may be the same or different; this embodiment does not impose any limitations on this.

[0277] In some embodiments, the multidimensional physiological characteristics associated with health-related risk events include, but are not limited to, at least two of the following: heart rate, HRV index, respiratory rate, and target posture characteristics; the target posture characteristics include at least two of the following: head posture characteristics, arm posture characteristics, and trunk posture characteristics.

[0278] It should be noted that the explanations of heart rate, HRV index, and respiratory rate in the foregoing embodiments also apply to this embodiment, and will not be repeated here.

[0279] Among them, the target posture features are extracted in the following ways: based on the spatial dimension information of the CSI signal, the occupant's head posture features (such as normal, forward tilt, side tilt, backward tilt), arm posture features (such as normal gripping of the steering wheel, detachment from the steering wheel), and torso posture features are extracted.

[0280] In any embodiment of this application, the CSI signal can be preprocessed before extracting multidimensional physiological features in order to extract multidimensional physiological features of the occupants in the cabin from the preprocessed CSI signal. The implementation principle can be found in the relevant description in step S602, and will not be repeated here.

[0281] Step S903: Detect health-related risk events based on multidimensional physiological characteristics to obtain event detection results; wherein, the event detection results are used to indicate the probability of an occupant experiencing multiple types of health abnormalities under health-related risk events in the future period.

[0282] In some embodiments, the likelihood of an occupant experiencing a health-related risk event in the future can be predicted based on the changing trends of multidimensional physiological characteristics, thus obtaining an event detection result. For example, a health anomaly detection model from the field of artificial intelligence can be used to detect health-related risk events based on the changing trends of multidimensional physiological characteristics, outputting a probability distribution; wherein, this probability distribution indicates the probability of an occupant experiencing multiple types of health anomalies under a health-related risk event in the future.

[0283] Among them, the health anomaly detection model has learned the correspondence between input data (multidimensional physiological characteristics) and output data (occupant health anomaly types).

[0284] Step S904: Query the multi-level second early warning strategy associated with the target health abnormality type; wherein, the target health abnormality type is any one of multiple health abnormality types, and the second early warning strategy associated with different health abnormality types is different. Each multi-level second early warning strategy associated with a health abnormality type is pre-set and used to trigger a health abnormality early warning notification; the second early warning strategies associated with different health abnormality types can be different.

[0285] Step S905: Based on the probability of the target health abnormality type in the event detection results, determine the third target early warning strategy from the multi-level second early warning strategy.

[0286] As a possible implementation, when the probability of the target health abnormality type is higher than the corresponding fourth probability threshold, and the trend of change of the target physiological feature associated with the target health abnormality type in the multidimensional physiological features meets the first abnormality condition, the third target early warning strategy (hereinafter referred to as strategy 1) can be used to indicate at least one of the following: triggering the first health abnormality prompt; triggering the stop and rest prompt.

[0287] The fourth probability thresholds corresponding to different types of health abnormalities may be different or the same, and this application does not impose any restrictions on this.

[0288] The target physiological characteristics associated with different types of health abnormalities can be different. For example, the target physiological characteristic associated with cardiac arrest can be the HRV index, and the target physiological characteristic associated with epileptic seizures can be psychological characteristics related to abnormal limb movements.

[0289] Each physiological characteristic has a corresponding first abnormal condition, and the first abnormal condition associated with different physiological characteristics can be different. For example, taking cardiac arrest as the target health abnormality type, the first abnormal condition associated with the HRV index is, for example, a deteriorating trend in the HRV index (or a continuous decline in the HRV index).

[0290] Among them, the trend of HRV indicator HRV It can be calculated using the following formula: Trend HRV = (HRV current - HRV baseline ) / HRV baseline ; Among them, HRV current HRV refers to the current HRV index in the multidimensional physiological characteristics of the occupant. baseline It refers to the HRV index of an individual passenger, which can be extracted from the passenger's historical physiological data.

[0291] In this application, if Trend HRV If the HRV value falls below a set threshold (e.g., -0.3, meaning a decrease of more than 30% in the HRV indicator), then the trend of HRV value change is considered to have deteriorated. Alternatively, if the trend... HRV If the HRV index falls below the set threshold and remains below the set threshold for a specified duration, then the trend of the HRV index is considered to be deteriorating.

[0292] For example, taking cardiac arrest as the target health abnormality type, if the probability of cardiac arrest in the event detection results is higher than the fourth probability threshold corresponding to myocardial infarction, and the trend of HRV index changes is worsening (or the HRV index continues to decline), at least one of the following can be performed: gently remind the occupant of the health abnormality via voice; display a yellow warning icon on the dashboard to indicate the occupant's health abnormality; or suggest that the occupant pull over and rest.

[0293] As another possible implementation, when the probability of the target health abnormality type is higher than the corresponding fifth probability threshold, and the trend of the target physiological characteristics meets the second abnormality condition, the third target warning strategy (hereinafter referred to as strategy 2) can be used to indicate at least one of the following: triggering the second health abnormality prompt; setting the vibration intensity to control the seat vibration in the occupant's seating area; reducing the vehicle speed; or turning on the hazard warning lights.

[0294] The fifth probability thresholds for different types of health abnormalities may be different or the same, and this application does not impose any restrictions on this. The fifth probability threshold for the same type of health abnormality is higher than the corresponding fourth probability threshold.

[0295] Each physiological characteristic has a corresponding second abnormal condition. Different physiological characteristics may be associated with different second abnormal conditions. Furthermore, the degree of abnormality of the second abnormal condition associated with the same type of health abnormality is higher than that of the corresponding first abnormal condition. For example, taking cardiac arrest as the target health abnormality type, the second abnormal condition associated with the HRV index might be: a continuously deteriorating trend in the HRV index (or a prolonged and sustained decline in the HRV index), such as Trend. HRV If the HRV index falls below a set threshold and remains below it for a second set duration, the trend of HRV change is determined to be continuously deteriorating. The second set duration is longer than the first set duration.

[0296] The second health abnormality warning has a higher intensity than the first health abnormality warning. For example, the second health abnormality warning may be: a voice emergency warning of occupant health abnormality; or a red flashing warning icon displayed on the dashboard.

[0297] For example, taking cardiac arrest as the target health abnormality type, if the probability of cardiac arrest in the event detection results is higher than the corresponding fifth probability threshold, and the trend of HRV index changes continues to deteriorate (or the HRV index continues to decline for a long time), at least one of the following can be performed: a voice emergency warning for the occupant's health abnormality; displaying a red flashing warning icon on the dashboard; controlling the seat in the occupant's seating area to vibrate violently; reducing the vehicle speed to a set speed threshold (e.g., 60 km / h); and activating the hazard warning lights.

[0298] As another possible implementation, when the probability of the target health abnormality type is higher than the corresponding sixth probability threshold, the third target warning strategy (hereinafter referred to as strategy 3) can be used to indicate at least one of the following: triggering the third health abnormality prompt; activating the Advanced Driving Assistance System (ADAS) to assist driving; turning on the right turn signal and controlling the vehicle to slowly stop at the road boundary; controlling the vehicle to drive to a safe position to stop.

[0299] The sixth probability thresholds for different types of health abnormalities may be different or the same, and this application does not impose any restrictions on this. However, the sixth probability threshold for the same type of health abnormality is higher than the corresponding fifth probability threshold.

[0300] Among them, the intensity of the third health abnormality warning is higher than that of the second health abnormality warning.

[0301] In summary, by using a three-tiered, progressively increasing probability threshold and corresponding physiological abnormality judgment conditions, and matching different intensities of tiered warning strategies, a mild alert is triggered in the mild abnormality stage to guide the occupant to take a break. In the moderate abnormality stage, auxiliary interventions such as seat vibration and speed reduction are activated. In the severe abnormality stage, ADAS is activated to complete a safe pullover. This approach avoids excessive early intervention that could disrupt normal driving, and accurately outputs appropriate protective actions at different health risk levels, ensuring driving safety when occupants experience sudden health abnormalities.

[0302] Step S906: Control the vehicle to execute the third target warning strategy to trigger a health abnormality warning prompt.

[0303] In some embodiments, when the third target warning strategy is strategy 1 as described above, the vehicle can be controlled to perform at least one of the following: triggering a first health abnormality warning; triggering a stop and rest warning.

[0304] In some embodiments, when the third target warning strategy is strategy 2 as described above, the vehicle can be controlled to perform at least one of the following: trigger a second health abnormality warning; set the vibration intensity to control the seat vibration in the occupant's seating area; reduce vehicle speed; or turn on hazard warning lights.

[0305] In some embodiments, when the third target warning strategy is strategy 3 as described above, the vehicle can be controlled to perform at least one of the following: trigger a third health abnormality warning; activate ADAS assisted driving; turn on the right turn signal and control the vehicle to slowly stop at the road boundary; control the vehicle to drive to a safe position to stop.

[0306] The vehicle control method of this application embodiment pre-sets exclusive multi-level second warning strategies for different types of health anomalies, and then selects and executes a corresponding level of third target warning strategy from the exclusive multi-level second warning strategies based on the occurrence probability of the detected target health anomaly type. This avoids the problem that general fixed warning strategies cannot adapt to different health anomaly scenarios, and can dynamically match the corresponding intensity of warning prompts according to the severity of the occupant's health anomaly, thereby improving the timeliness of response to occupant health risks and the accuracy of warning adaptation in vehicle scenarios.

[0307] As one possible implementation method, Figure 10 This is a schematic flowchart illustrating another vehicle control method provided as an exemplary embodiment of this application. It should be noted that this vehicle control method can be executed alone, or it can be executed together with any embodiment or possible implementation thereof in this application, or it can be executed together with any technical solution in related technologies. This application does not impose any limitations on this.

[0308] like Figure 10 As shown, the vehicle control method may include the following steps S1001 to S1008: Step S1001: Obtain the CSI signal obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin.

[0309] Step S1002: In response to occupant status events, including health-related risk events, extract multidimensional physiological characteristics of the occupants in the cabin from the CSI signal.

[0310] Step S1003: Detect health-related risk events based on multidimensional physiological characteristics to obtain event detection results; wherein, the event detection results are used to indicate the probability of an occupant experiencing multiple types of health abnormalities under health-related risk events in the future.

[0311] Step S1004: Query the multi-level second early warning strategy associated with the target health abnormality type, and determine the third target early warning strategy from the multi-level second early warning strategy based on the probability of the target health abnormality type in the event detection results.

[0312] The target health abnormality type can be any one of multiple health abnormality types, and the second early warning strategy associated with different health abnormality types can be different. It should be noted that the explanations of steps S1001 to S1004 can be found in the relevant descriptions in any embodiment of this application, and will not be repeated here.

[0313] In any embodiment of this application, environmental information about the vehicle's driving environment can also be obtained, and the third target warning strategy can be adjusted based on this environmental information. The environmental information includes, but is not limited to, at least one of the following: the location information of the road segment the vehicle is traveling on, the degree of congestion, weather information, ambient brightness, etc.

[0314] As an example, in scenarios where vehicles are traveling on highways, the updated third-objective warning strategy can also be used to instruct: prioritize the use of the emergency lane, reduce speed to the first speed (e.g., 40 km / h), and then pull over; in scenarios where vehicles are traveling on urban roads, the updated third-objective warning strategy can also be used to instruct: find the nearest compliant roadside parking space, reduce speed to the second speed (e.g., 20 km / h), and then pull over; in scenarios where vehicles are traveling in tunnels or on bridges, the updated third-objective warning strategy can also be used to instruct: extend the warning duration, wait for the vehicle to exit the current road segment and reach a safe area before taking further action; When a vehicle is traveling in extreme weather conditions, the updated third-target warning strategy can also be used to instruct: increase the anomaly detection threshold (i.e., the fourth probability threshold, fifth probability threshold, and sixth probability threshold in the above embodiments) to reduce the probability of false triggering in complex environments; when a vehicle is traveling in a congested road segment, the updated third-target warning strategy can also be used to instruct: not to perform the pull-over operation, that is, the updated third-target warning strategy does not include strategy 3 mentioned above; when a vehicle is traveling on a night road segment, the updated third-target warning strategy can also be used to instruct: activate the high beam warning action, and remind surrounding vehicles to pay attention and avoid the vehicle by flashing the high and low beams alternately.

[0315] In summary, the dynamic adjustment mechanism of the strategy can avoid the adaptability defects of fixed early warning strategies in different driving scenarios, and improve the environmental adaptability of the vehicle's handling actions and driving safety in the event of sudden health abnormalities.

[0316] Step S1005: Control the vehicle to execute the third target warning strategy to trigger a health abnormality warning prompt.

[0317] It should be noted that the explanation of step S1005 can be found in the relevant description in any embodiment of this application, and will not be repeated here.

[0318] Step S1006: Determine whether the occupants have made a preset effective response after the vehicle executes the third target warning strategy. If yes, proceed to step S1007; otherwise, proceed to step S1008.

[0319] Valid responses include, but are not limited to: occupants issuing voice commands to control the vehicle, and occupants being physically active (such as turning their heads or moving their arms). For example, taking an occupant including the driver, valid responses include, but are not limited to: the driver issuing voice commands to control the vehicle, the driver being physically active, and the driver operating the steering wheel.

[0320] It should be noted that steps S1007 and S1008 are two parallel possible implementations, and either one can be selected for execution.

[0321] Step S1007: Stop executing the third objective prediction strategy.

[0322] In this embodiment, if the occupant makes a preset effective response after the vehicle executes the third target warning strategy, the execution of the third target warning strategy can be stopped to reduce unnecessary interference to the occupant, avoid interrupting the occupant's normal operating rhythm with continuously triggered prompts, vibrations, and other actions, and at the same time reduce the ineffective operating power consumption of vehicle components.

[0323] Step S1008: Determine the fourth target warning strategy from the multi-level second warning strategy, and control the vehicle to execute the fourth target warning strategy, wherein the warning level of the fourth target warning strategy is higher than that of the third target warning strategy.

[0324] In this application, the level difference between the early warning level of the fourth target early warning strategy and the early warning level of the third target prediction strategy is not specifically limited. That is, the level difference can be one level, two levels, three levels, etc.

[0325] In this embodiment of the application, if the occupants do not make a preset effective response after the vehicle executes the third target warning strategy, a fourth target warning strategy can be determined from the multi-level second warning strategies, and the vehicle can be controlled to execute the fourth target warning strategy, wherein the warning level of the fourth target warning strategy is higher than that of the third target warning strategy.

[0326] In any embodiment of this application, when the occupants include the driver, the driver fails to respond effectively, including but not limited to at least one of the following: The first step: Within the target determination timeframe after the third target warning strategy is triggered, acquire the driver's updated physiological characteristics, and based on these updated physiological characteristics, determine that the driver's body is in an inactive state. The target determination timeframe is associated with the third target warning strategy; different warning strategies may have different associated target determination timeframes. The updated physiological characteristics are extracted from the latest CSI signal sensed by the WiFi sensing node.

[0327] The second point is that no effective steering wheel control actions by the driver were detected within the target judgment time.

[0328] Thirdly, no voice commands issued by the driver were recognized within the target judgment time.

[0329] In summary, the driver unresponsiveness determination logic through multi-dimensional state cross-validation can accurately identify dangerous scenarios where drivers lose response after a sudden health abnormality, avoiding misjudgment and omission caused by single-dimensional detection, improving the accuracy and timeliness of driver unresponsiveness identification, and providing a reliable basis for subsequent emergency protective actions such as safe parking.

[0330] In any embodiment of this application, when the vehicle stops and the driver has not yet provided a valid response, the vehicle can be controlled to execute the fifth target warning strategy (hereinafter referred to as strategy 4), which is the highest warning level in the multi-level second warning strategy. The fifth target warning strategy is used to instruct, but is not limited to, at least one of the following: controlling the vehicle's electronic parking system to lock the parking brake; controlling the vehicle's horn to continuously emit a warning sound; initiating an emergency rescue call based on the vehicle's location information; pushing a vehicle warning notification to a preset emergency contact; controlling the vehicle's door locks to unlock and activating hazard warning lights (such as hazard lights); controlling the vehicle's windows to open for ventilation; and cyclically broadcasting the vehicle's location information.

[0331] Understandably, if no effective response from the driver is detected after the vehicle has completed a temporary stop, the highest level, the fifth-target warning strategy, is triggered, which executes at least one of the following: locking the parking brake, continuously sounding the horn, initiating an emergency rescue call, sending a warning message to a pre-set emergency contact, unlocking the doors and turning on the hazard lights, opening the windows for ventilation, and continuously broadcasting the vehicle's location information. This can achieve multi-dimensional emergency protection in the first instance, which not only avoids the vehicle being in an unprotected dangerous state after the driver suddenly becomes incapacitated, but also quickly transmits accurate distress location information to surrounding vehicles, pedestrians, and rescuers, improving the efficiency of rescue response after the driver suddenly becomes unwell.

[0332] As an application scenario, the vehicle control method provided in this application can be applied to at least the following scenarios: Scenario 1: Long-distance highway driving scenario: During long-distance highway driving, drivers may experience cardiovascular health abnormalities due to maintaining a fixed posture for extended periods and high levels of mental stress. The vehicle continuously monitors the driver's multidimensional physiological characteristics and issues an early warning reminder to rest when abnormal heart rate or deteriorating HRV indicators are detected. If the driver suffers sudden cardiac arrest and loses consciousness, emergency procedures are implemented, such as Strategy 4.

[0333] Scenario 2: Urban Traffic Congestion: In urban traffic congestion, drivers experience greater emotional fluctuations and are more prone to elevated blood pressure. The vehicle monitors changes in various indicators of the driver's multidimensional physiological characteristics in real time. When hypertension-related physiological characteristics are detected, the vehicle proactively adjusts the cabin environment (such as activating seat massage and playing soothing music) to prevent health-related risk events.

[0334] The vehicle control method of this application embodiment detects whether the occupant makes a preset effective response after executing the third target warning strategy. If so, the current third target warning strategy is terminated to avoid unnecessary disturbance to the occupant caused by the continuous running prompts and intervention actions. If not, it switches to the fourth target warning strategy with a higher warning level. This can realize the timely output of higher-level protective measures when the occupant's health abnormality continues to worsen, thereby improving the reliability of closed-loop handling of sudden health risks of occupants in vehicle scenarios and driving safety.

[0335] In any embodiment of this application, where occupant risk events include health-related risk events, this application provides a method such as... Figure 11 The illustrated intelligent cockpit-based driver health anomaly detection and emergency safety response solution, based on WiFi CSI signal sensing, mainly includes the following steps: Step 1'': Deploy WiFi sensing nodes in the cockpit and continuously collect CSI signals from the driver's seating area.

[0336] The hardware deployment method includes: deploying one WiFi sensing node (Tx) in front of the driver's seat headrest or at the steering wheel column; deploying one WiFi sensing node (Rx) directly behind the driver (such as at the B-pillar); the two ends of Tx and Rx form a WiFi signal path that penetrates the driver's torso to maximize the sensitivity of heartbeat / breathing signal detection; both Tx and Rx are equipped with a 2×2 MIMO antenna array.

[0337] The data acquisition method includes: a sampling frequency of 500Hz (to meet the sampling requirements of the heartbeat signal); each CSI signal contains complex CSI values ​​of 234 subcarriers under an 80MHz bandwidth; continuous acquisition to form time-series data. The CSI signal data format is as follows: the CSI signal at each time point is represented as a complex matrix: H(t) ∈ C^(2×2×234), where 2×2 is the antenna dimension of the MIMO antenna array, and 234 is the number of subcarriers.

[0338] Step 2'': Preprocess the CSI signal under dynamic vehicle conditions to eliminate interference from vehicle vibrations and extract clean signals relevant to the driver's physiology. This mainly includes the following sub-steps: Step 2.1'': Vibration and noise characteristics analysis: Vibration sources and characteristics during vehicle operation: Engine / motor vibration: stable frequency (approximately 20-100Hz), which can be eliminated by notch filters; Road bump vibration: highly random (0.5-20Hz), positively correlated with vehicle speed; Centrifugal force during cornering: causes the driver's body to tilt, resulting in low-frequency CSI offset; Inertial force during rapid acceleration and deceleration: causes the driver to move forward and backward, resulting in transient CSI changes.

[0339] Step 2.2'': IMU-CSI Synchronous Acquisition and Calibration: Use hardware triggering to achieve precise time synchronization between IMU and CSI (error <1ms); collect 10 minutes of baseline data while the vehicle is stationary to establish a static mapping relationship between CSI and IMU; collect calibration data under different vehicle speeds and road conditions to establish a dynamic compensation model.

[0340] Step 2.3'': Multi-level vibration cancellation algorithm: Level 1: Engine / Motor Vibration Elimination: Adaptive notch filter is used, and the center frequency is dynamically adjusted with engine speed.

[0341] Level 2: Road bump elimination: Use an extended Kalman filter to fuse CSI signals and IMU data.

[0342] Level 3: Driver motion artifact elimination: Use motion detection algorithms to identify large driver movements (turning head, reaching out, etc.); mark large movement segments and reduce their weight or perform interpolation processing in subsequent analysis.

[0343] Step 2.4'': Phase Correction: Related techniques only use linear fitting, which cannot eliminate nonlinear errors caused by temperature drift. This application uses the following dual-anchor correction method to perform phase correction on the CSI signal. For example, considering the effect of temperature rise after vehicle startup, the correction parameters are recalculated every 5 minutes.

[0344] Step 2.5'': Adaptive bandpass filtering: Heartbeat signal frequency band: 0.8-2.0 Hz (corresponding to 48-120 bpm), using a 4th-order Butterworth filter; Breathing signal frequency band: 0.1-0.5 Hz (corresponding to 6-30 bpm), using a 2nd-order Butterworth filter; Adaptive adjustment: When the driver is detected to be in a resting state, the bandpass range is narrowed to improve the signal-to-noise ratio.

[0345] Step 3'': Multidimensional Physiological Feature Extraction: Extract the driver's multidimensional physiological features from the preprocessed CSI signal. This mainly includes the following sub-steps: Step 3.1'': Heart rate extraction, in bpm.

[0346] Step 3.2'': HRV index extraction, where the time-domain HRV index includes SDNN and RMSSD; the frequency-domain HRV index includes low-frequency power, high-frequency power, and the ratio of low-frequency power to high-frequency power.

[0347] Step 3.3'': Respiratory rate extraction: Perform peak detection on the CSI signal of the respiratory frequency band; calculate the time interval between adjacent respiratory peaks; respiratory rate BR(t) = 60 / mean (breathing interval), unit breaths / min; where mean (breathing interval) refers to the statistical value obtained by averaging the time intervals of all discrete adjacent respiratory peaks.

[0348] Step 3.4'': Posture feature extraction: Utilize CSI spatial dimension information (MIMO antenna array) to extract the driver's torso posture features, detect head posture features (such as normal, forward tilt, side tilt, backward tilt), and arm posture features (such as normal grip on the steering wheel, detachment from the steering wheel), and output the posture feature vector Pose(t) ∈ {normal, abnormal}.

[0349] Step 4'': Multimodal fusion detection of sudden health abnormality types: Based on the extracted multidimensional physiological features, deep learning technology is used to detect the types of sudden health abnormalities. This mainly includes the following sub-steps: Step 4.1'': Definition of Abnormal Types: Define the following types of sudden health abnormalities: Type 1: Cardiac arrest; Type 2: Myocardial infarction; Type 3: Stroke; Type 4: Seizure; Type 5: Severe hypoglycemia; Type 0: Normal state.

[0350] Step 4.2'': Feature Engineering: Construct a multi-dimensional temporal feature vector: F(t) = [HR(t), SDNN(t), RMSSD(t), LF(t), HF(t), LF / HF(t), BR(t), Pose(t)]; For example, a sliding window can be used to construct the feature matrix with a window size of 5 minutes and a sliding step size of 30 seconds.

[0351] Step 4.3'': Model Architecture of the Health Anomaly Detection Model: Input layer: Receives a feature matrix with dimensions [B, T, 8], where B is the batch size, representing the number of feature vectors detected in a single health anomaly detection; T is the time steps, representing the time length of the feature vector (i.e., the number of sampling points within the sliding window); and 8 is the feature dimension. The 1D CNN layer is used to extract local temporal features. It performs two convolutional operations in sequence: the first layer is a one-dimensional convolution Conv1D(64, kernel=5) with 64 convolutional kernels and a kernel size of 5, which is concatenated with the ReLU activation function and the BatchNorm normalization function; the second layer is a one-dimensional convolution Conv1D(128, kernel=3) with 128 convolutional kernels and a kernel size of 3, which is also concatenated with the ReLU activation function and the BatchNorm function, to complete the layer-by-layer extraction and normalization calibration of local fine-grained temporal features.

[0352] The bidirectional LSTM layer is used to capture long-range temporal dependencies. It consists of two stacked bidirectional long short-term memory networks Bi-LSTM (128) with a hidden layer dimension of 128. It traverses the temporal features along both the forward and reverse time directions to fully capture the long-range temporal dependencies in the temporal features, avoiding the omission of reverse correlation features in unidirectional temporal processing.

[0353] The self-attention layer is used to automatically weight and allocate feature weights, focusing on key time segments before and after the occurrence of health abnormalities, and strengthening the expression weight of core features related to health abnormalities.

[0354] Fully connected classification layer: a fully connected layer FC (256) with 256 neurons, a fully connected layer FC (128) with 128 neurons, and a fully connected layer FC (6) with 6 neurons are connected in sequence. Finally, the Softmax function is connected to map the high-dimensional features into the probability distribution of 6 types of health abnormalities.

[0355] Output: Probability distribution of 6 types of health abnormalities P = [P0, P1, P2, P3, P4, P5]; where P0 is the probability of normal state, P1 is the probability of cardiac arrest, P2 is the probability of myocardial infarction, P3 is the probability of stroke, P4 is the probability of epileptic seizure, and P5 is the probability of severe hypoglycemia.

[0356] Step 4.4'': Early warning mechanism: In related technologies, driver monitoring systems mostly use threshold judgment, which can only detect abnormalities after they occur. However, this application introduces the changing trend of the HRV index to achieve early warning. The following explanation uses cardiac arrest as an example of a health abnormality type: (a) HRV deterioration trend identification: Before cardiac arrest, the autonomic nervous system will first show functional disorders, manifested as: continuous decline of SDNN, increased fluctuation of RMSSD, and abnormal LF / HF ratio. This application achieves early identification by analyzing the sliding window trend of HRV indicators; for example, the trend calculation formula is as follows: Trend HRV = (HRV current - HRV baseline) / HRV baseline A warning is issued when Trend_HRV < -0.3 (decreases by more than 30%) and persists for more than 5 minutes.

[0357] Among them, HRV baseline The HRV index, which serves as the driver's individual baseline, can be extracted from the driver's historical physiological data.

[0358] (b) Multi-timescale analysis: Short-term window (5 minutes): detects acute changes; Medium-term window (30 minutes): detects slow deterioration trends; Long-term window (2 hours): establishes individual baselines and detects deviations.

[0359] (c) Warning window: Deterioration of HRV indicators can occur 5-15 minutes before cardiac arrest. This application is designed to issue a warning 5-10 minutes before the event occurs, allowing the driver sufficient reaction time (to pull over, seek medical attention, etc.).

[0360] (d) Warning confidence level classification: Level 1 warning: P1>0.3 or the trend of HRV index deteriorates (rest is recommended); Level 2 warning: P1>0.5 and the trend of HRV index continues to deteriorate (medical treatment is recommended); Level 3 warning: P1>0.7 (emergency situation, immediate action is required).

[0361] Step 4.5'': Multimodal fusion decision: Fusion judgment of abnormal physiological parameters and abnormal posture; when abnormal physiological parameters and abnormal posture (such as sudden forward head tilt) occur simultaneously, increase the abnormality confidence; final abnormality confidence Conf final = α × Conf physio + β × Conf pose Where α = 0.8, β = 0.2, and Conf physio The confidence level for abnormal physiological parameters can be calculated from the probability distribution P in the model output, such as Conf. physio = 1 - P0, or take max(P1, P2, P3, P4, P5); Conf pose The confidence level for abnormal posture.

[0362] Step 5'': Implementation of the tiered early warning strategy (also known as the emergency response strategy): Based on the severity of the anomaly detection results, a progressively escalating tiered emergency response is implemented to achieve a closed-loop process from "passive alert" to "assisted driving." This mainly includes the following sub-steps: Step 5.1'': Using Table 3, define the warning levels (progressive escalation mechanism) for emergency response scenarios: Table 3. Warning Levels in Emergency Response Scenarios

[0363] Among them, P abnormal It refers to the total probability of different types of health abnormalities, i.e., P. abnormal = 1 - P0 (the probability that P0 is the normal state), or P abnormal = max(P1, P2, P3, P4, P5), in the subsequent upgrade logic, when P abnormal When the threshold is reached, the action for the current level is executed.

[0364] Step 5.2'': Progressive escalation mechanism: In related technologies, only an alarm is issued, and there is no follow-up if the driver does not respond; however, in this application, a closed-loop mechanism of "detection-alert-verification-escalation" is established. The escalation logic is as follows: if the anomaly probability P > the threshold, then the current level action is executed, and a response detection timer (T) is started. response ), in T response When the value is >0, the system checks whether the driver has made an effective response (e.g., voice response, manual operation, posture change). If so, the emergency response is stopped, normal driving resumes, and the false triggering event is recorded for subsequent model optimization. T is then set to... response = T response - Δt. If T response If the value is 0 and the driver does not respond, then the system will be upgraded to the next level.

[0365] Step 5.3'': Driver Response Detection Mechanism: Sensing whether the driver has regained consciousness via WiFi CSI signal: Detecting whether the driver's body is moving (e.g., head turning, arm movement); detecting whether the driver attempts to operate the steering wheel; detecting whether the driver issues voice commands; if any of the above responses are detected, it is determined that the driver has regained consciousness.

[0366] Step 5.4'': Adjust the adaptive policy safety constraints of the strategy based on the environmental information of the vehicle's driving; wherein, the environmental information includes at least one of the following: the location information of the vehicle's driving segment, the degree of congestion, the weather information, and the ambient brightness.

[0367] Step 5.5'': Human-Computer Interaction Design (Graded Differentiation): L1 warning: Gentle voice prompt "You may need to rest. There is a service area ahead. We recommend that you stop and rest." L2 warning: Emergency voice prompt "Health abnormality detected! Please pull over immediately!" + violent seat vibration; L3 Assisted Driving: Reassuring voice prompt: "The system is activating assisted driving functions. Please relax and don't be nervous." L4 parking: The system continuously broadcasts "Vehicle has been safely parked, emergency services have been called, and location has been sent"; As a result, the tiered emergency response strategy has been upgraded from "just a reminder" to "assisting driving," addressing safety issues after the driver loses consciousness and balancing safety with the risk of accidental triggering.

[0368] In summary, the technical solution provided in this application includes at least the following beneficial effects: Utilizing existing in-vehicle WiFi infrastructure, the system operates automatically without the driver needing to wear any devices; it uses only WiFi CSI signals, without collecting sensitive information such as images or sounds, eliminating the risk of privacy leaks within the vehicle; by continuously monitoring the changing trends of HRV indicators, it can issue warnings 5-10 minutes before serious health abnormalities such as cardiac arrest occur, giving the driver valuable time to rest or seek medical attention; it simultaneously monitors multiple indicators such as heart rate, HRV, respiratory rate, and posture, which is more accurate and comprehensive than monitoring a single indicator; when the driver loses consciousness, the vehicle can perform actions such as pulling over, activating hazard lights, and initiating an emergency rescue call, gaining golden time for life-saving; it selects the optimal emergency response strategy based on the severity of the health abnormality, balancing safety and user experience, and avoiding overreaction; the vibration cancellation algorithm designed for dynamic vehicle scenarios ensures accurate extraction of physiological characteristics even on bumpy roads and during acceleration and deceleration; it can be implemented using existing in-vehicle WiFi modules, eliminating the need for additional expensive dedicated sensors, resulting in lower hardware costs; it can learn from the physiological baseline data of specific drivers to establish personalized health models, with increasingly accurate detection over longer usage time.

[0369] To achieve the above embodiments, this application also proposes a vehicle control device. Figure 12 This is a schematic diagram of the structure of a vehicle control device provided for an exemplary embodiment of this application.

[0370] like Figure 12 As shown, the vehicle control device 1200 may include: an acquisition module 1210, an extraction module 1220, a detection module 1230, and a control module 1240.

[0371] The acquisition module 1210 is used to acquire Channel State Information (CSI) signals obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin; the extraction module 1220 is used to extract signal features associated with at least one occupant risk event from the CSI signal; the detection module 1230 is used to detect occupant risk events based on the signal features and obtain event detection results; wherein, the event detection results are used to determine the occupant risk state in the occupant risk event; and the control module 1240 is used to control the vehicle to execute a vehicle control strategy adapted to the event detection results.

[0372] In one implementation of this application, the event detection result includes the probability that an occupant in the cabin is in each occupant risk state under an occupant risk event; the occupant risk event includes comfort-related risk events, and the event detection result is used to indicate the probability that each occupant in the cabin will experience multiple thermal comfort states under a comfort-related risk event in the current time period; and / or, the occupant state event includes health-related risk events, and the event detection result is used to indicate the probability that an occupant will experience multiple types of health abnormalities under a health-related risk event in the future time period; and / or, the occupant state event includes motion sickness events, and the event detection result is used to indicate the probability that an occupant will experience multiple motion sickness states under a motion sickness event in the current time period.

[0373] In one implementation of this application, in response to occupant risk events including comfort-related risk events, and signal features including micro-motion features corresponding to multiple seating areas within the cabin, an extraction module 1220 is configured to: extract local CSI signals corresponding to multiple seating areas within the cabin from the CSI signal; wherein the CSI signal includes complex channel response data of multiple subcarriers collected by multiple receiving antennas in the antenna array of the WiFi sensing node; extract time-frequency features of the local CSI signal of each seating area based on the target frequency band associated with multiple micro-motions; determine the spatial features of the local CSI signal of each seating area based on the difference in the angle of arrival (AOA) between the complex channel response data collected by different receiving antennas in the local CSI signal of each seating area; and fuse the time-frequency features and spatial features corresponding to each seating area to obtain the micro-motion features of each seating area.

[0374] In one implementation of this application, in response to occupant risk events including comfort-related risk events, and signal features including micro-motion features corresponding to multiple seating areas within the cabin, a detection module 1230 is configured to: extract common features among the micro-motion features of multiple seating areas using an encoding network in a thermal comfort prediction model; detect the presence of occupants based on the common features using an occupant detection network in the thermal comfort prediction model to obtain occupant detection results; wherein the occupant detection results are used to indicate whether an occupant exists in each seating area; activate the target decoding network corresponding to the target seating area among multiple decoding networks in the thermal comfort prediction model; wherein the target seating area is the seating area where the occupant detection results indicate the presence of an occupant, and each decoding network corresponds to one seating area; decode the common features using the target decoding network to output a probability distribution; wherein the probability distribution is used to indicate the probability that an occupant in the target seating area will experience multiple thermal comfort states during the current time period.

[0375] In one implementation of this application, the thermal comfort prediction model is trained using a training module. The training module is configured to: acquire training data; wherein the training data includes sample micro-motion features, thermal comfort annotation information, occupant annotation information, and micro-motion annotation information associated with at least one sample seat area; the thermal comfort annotation information indicates the thermal comfort state of an occupant in the sample seat area, the occupant annotation results indicate whether an occupant is present in the sample seat area, and the micro-motion annotation information indicates the type of micro-motion performed by the occupant; process the sample micro-motion features using the thermal comfort prediction model to obtain output results; wherein the output results include thermal comfort prediction information, occupant prediction information, and micro-motion prediction information for the sample seat area; and train the thermal comfort prediction model based on the differences between the thermal comfort annotation information and the thermal comfort prediction information, the differences between the occupant annotation information and the occupant prediction information, and the differences between the micro-motion annotation information and the micro-motion prediction information.

[0376] In one implementation of this application, the occupant risk event includes comfort-related risk events. The event detection result is used to indicate the probability of an occupant in a target seating area within the cabin experiencing multiple thermal comfort states during the current time period. The control module 1240 is used to: determine a target thermal comfort state from multiple thermal comfort states based on the probability of these states; determine a temperature control strategy based on the target thermal comfort state corresponding to the target seating area; and perform operational control on the vehicle components associated with the target seating area according to the temperature control strategy. The vehicle components include at least one of the following: a vehicle air conditioner, a seat heating component, and a seat ventilation component.

[0377] In one implementation of this application, the control module 1240 is configured to: predict the remaining state of charge (SOC) of the vehicle battery when the vehicle arrives at its destination based on the vehicle's remaining driving range and trip planning information; determine the target energy mode that the vehicle needs from a plurality of preset energy modes based on the actual SOC and remaining SOC of the vehicle battery; and generate a temperature control strategy based on the target energy mode and the target thermal comfort state of the target seating area.

[0378] In one implementation of this application, the target energy mode includes: a first energy mode, a second energy mode, or a third energy mode. In the first energy mode, the vehicle's actual SOC is greater than a set first SOC, and the vehicle's remaining SOC upon arrival at the destination is greater than a set second SOC. In the second energy mode, the vehicle's actual SOC is greater than a set third SOC and less than or equal to the first SOC, and the vehicle's remaining SOC upon arrival at the destination is greater than the second SOC. The third SOC is greater than the second SOC. In the third energy mode, the vehicle's actual SOC is less than or equal to the third SOC, and the vehicle's remaining SOC upon arrival at the destination is greater than the second SOC. Alternatively, in the third energy mode, the vehicle's remaining SOC upon arrival at the destination is less than the second SOC.

[0379] In one implementation of this application, in response to the target energy mode being a first energy mode, the control module 1240 is configured to: determine the required temperature of the target seating area based on the target thermal comfort state of the target seating area, and determine the target vehicle components adapted to the target thermal comfort state from the vehicle components associated with the target seating area; in response to the number of target seating areas being one, generate a temperature control strategy for controlling the target vehicle components based on the required temperature; in response to the number of target seating areas being multiple, weight the required temperatures of the multiple target seating areas according to the weighted weights corresponding to the multiple target seating areas to obtain the target required temperature, and generate a temperature control strategy for controlling the target vehicle components based on the target required temperature; wherein, the weighted weights are used to indicate the degree of influence of the occupants in the corresponding target seating areas on the safe operation of the vehicle.

[0380] In one implementation of this application, the event recognition result further includes: confidence levels of multiple target seating areas, wherein the confidence level is used to indicate the uncertainty of the target thermal comfort state of the occupant in the corresponding target seating area; the vehicle control device 1200 may further include: a first adjustment module, used to adjust the weighted weights corresponding to the multiple target seating areas according to the confidence levels of the multiple target seating areas.

[0381] In one implementation of this application, the control module 1240 is configured to: determine the temperature adjustment direction based on the target thermal comfort state, and determine the target vehicle component adapted to the target thermal comfort state from the vehicle components associated with the target seating area; in response to the target energy mode being a second energy mode, generate a temperature control strategy for controlling the target vehicle component based on the temperature adjustment direction and a set first temperature adjustment range; and in response to the target energy mode being a third energy mode, generate a temperature control strategy for controlling the target vehicle component based on the temperature adjustment direction and a set second temperature adjustment range; wherein the second temperature adjustment range is smaller than the first temperature adjustment range.

[0382] In one implementation of this application embodiment, the temperature control strategy satisfies any one of the following: In response to the target energy mode being the second energy mode and the temperature adjustment direction being positive, the temperature control strategy indicates at least one of the following: the temperature adjustment amount of the vehicle air conditioner, turning off the vehicle air conditioner corresponding to the non-target seat area, and activating the seat heating function under the target seat area; In response to the target energy mode being the second energy mode and the temperature adjustment direction being negative, the temperature control strategy indicates at least one of the following: the temperature adjustment amount of the vehicle air conditioner, turning off the vehicle air conditioner corresponding to the non-target seat area, and activating the seat ventilation function under the target seat area; In response to the target energy mode being the third energy mode, the temperature control strategy indicates at least one of the following: the required temperature of the driver's seat area in the target seat area, turning off the vehicle air conditioner corresponding to the non-target seat area, and triggering an energy-saving reminder; wherein, the energy-saving reminder indicates that the vehicle is currently in energy-saving mode.

[0383] In one implementation of this application, the response to occupant status events includes health-related risk events and / or motion sickness events, and the signal features include multidimensional physiological features of the occupants in the cabin. An extraction module 1220 is used to: preprocess the CSI signal and extract the multidimensional physiological features of the occupants in the cabin from the preprocessed CSI signal; wherein the preprocessing includes at least one of the following: using a notch filter adapted to the vibration frequency of the vibrating components in the vehicle to filter the CSI signal to eliminate vibration noise introduced by the vibrating components; wherein the center frequency of the notch filter is related to the rotational speed of the vibrating components, and the vibrating components include an engine and / or a motor; using a Kalman filter to perform fusion filtering on the CSI signal based on motion data collected by the vehicle's inertial measurement unit (IMU) to... The system eliminates vibration noise introduced by road bumps; the IMU is time-synchronized with the WiFi sensing node; a Butterworth filter adapted to the signal frequency band corresponding to the multidimensional physiological characteristics is used to filter the CSI signal to eliminate out-of-band interference signals; the passband range of the Butterworth filter is determined according to the occupant's motion state; motion detection is performed on the CSI signal to mark target segments in the CSI signal; if the occupant's motion amplitude indicated by the target segment is higher than a set amplitude, the marking is used to reduce the weight of the target segment when extracting multidimensional physiological characteristics, and / or, the marking is used to perform interpolation filling processing on the target segment based on adjacent segments; target correction parameters are used to perform phase correction processing on the CSI signal to eliminate nonlinear errors caused by temperature drift.

[0384] In one implementation of this application, the target correction parameter is calculated using a determination module. This determination module is configured to: acquire the original CSI signal collected by the WiFi sensing node within the most recent correction period; wherein the original CSI signal is in matrix form, and the element in the i-th row and j-th column of the matrix indicates the complex channel response data of the j-th subcarrier collected by the i-th receiving antenna in the antenna array of the WiFi sensing node; construct a phase correction model with the correction parameter as a variable, wherein the phase correction model is used to subtract the temperature drift component determined by the correction parameter from the phase component of the original CSI signal; and solve the phase correction model with the objective function of minimizing the first-order difference variance in the time dimension of the CSI phase sequence corresponding to the same subcarrier and the same receiving antenna after correction by the phase correction model to obtain the target correction parameter; wherein the CSI phase sequence includes the phase components of each original CSI signal collected in the most recent correction period.

[0385] In one implementation of this application, in response to an occupant risk event including motion sickness, the signal features include multidimensional physiological features, including at least two of the following: heart rate, heart rate variability (HRV) index, respiratory rate, microkinetic energy caused by friction between skin and clothing determined based on high-frequency components in the CSI signal, and head posture features; the detection module 1230 is used to: acquire reference physiological features of the occupant; wherein the reference physiological features are determined based on the occupant's historical physiological data; determine the difference data between the reference physiological features and the multidimensional physiological features; and detect motion sickness events based on the difference data to obtain event detection results.

[0386] In one implementation of this application, in response to an occupant risk event including a motion sickness event, the control module 1240 is configured to: determine a first target warning strategy from a preset multi-level first warning strategy based on the probability of multiple motion sickness states in the event detection results; and control the vehicle to execute the first target warning strategy to trigger a motion sickness warning prompt.

[0387] In one implementation of this application, multiple motion sickness states include a first motion sickness state and a second motion sickness state, wherein the degree of motion sickness in the second motion sickness state is higher than that in the first motion sickness state. A first target warning strategy satisfies any one of the following: in response to the probability of the first motion sickness state being higher than a first probability threshold, the first target warning strategy is used to prompt the passenger to wear smart glasses and / or trigger a motion sickness risk warning; in response to the probability of the second motion sickness state being higher than a second probability threshold, the first target warning strategy is used to instruct the activation of a first visual intervention mechanism in the smart glasses to alleviate the passenger's anxiety and nausea; in response to the probability of the second motion sickness state being higher than a third probability threshold, the first target warning strategy is used to instruct at least one of the following: activating a second visual intervention mechanism in the smart glasses to alleviate the degree of motion sickness in the passenger; triggering a stop and rest prompt; triggering a window ventilation prompt; wherein the third probability threshold is higher than the second probability threshold, and the second probability threshold is higher than the first probability threshold, and the intervention intensity of the second visual intervention mechanism is higher than that of the first visual intervention mechanism.

[0388] In one implementation of this application, the first visual intervention mechanism includes at least one of the following: displaying a fixed reference point to establish a stable reference frame for the occupant's visual system; displaying breathing guidance material to guide the occupant to take deep breaths; adjusting at least one of the color temperature, brightness, and hue of the displayed content; and / or, the second visual intervention mechanism includes at least one of the following: displaying a virtual horizon line to simulate a horizon; reducing the refresh rate of the displayed content; performing motion compensation on the displayed content; blurring the visual field edges of the displayed content to guide the occupant to focus on the central area of ​​the displayed content.

[0389] In one implementation of this application, displaying a fixed reference point includes: tracking the occupant's head posture using the inertial measurement unit (IMU) of the smart glasses; obtaining the fixed display position of the fixed reference point in virtual space; determining the actual display position of the fixed reference point on the smart glasses based on the head posture, a set compensation coefficient, and the fixed display position; and displaying the fixed reference point on the smart glasses based on the actual display position to keep the fixed reference point fixed in virtual space.

[0390] In one implementation of this application, the breathing guidance material is displayed as an animation. Displaying the breathing guidance material includes: periodically contracting and expanding the displayed reference graphic with a set contraction cycle and expansion cycle; wherein, the contraction process is used to guide the occupant to exhale; and the expansion process is used to guide the occupant to inhale.

[0391] In one implementation of this application, in response to the probability of a second motion sickness state being higher than a second probability threshold, the first target warning strategy is further used to instruct the vehicle to broadcast breathing prompt information associated with breathing guidance material.

[0392] In one implementation of this application, the control module 1240 is further configured to: acquire the latest physiological characteristics of the occupants after the vehicle executes the first warning strategy; determine whether the degree of motion sickness of the occupants has decreased based on the latest physiological characteristics; if not, determine a second target warning strategy from the multi-level first warning strategies, wherein the warning level of the second target warning strategy is higher than that of the first target warning strategy; and control the vehicle to execute the second target warning strategy.

[0393] In one implementation of this application, the vehicle control device 1200 may further include: an update module, configured to acquire occupant feedback information and parameter adjustment information on at least one first warning strategy; and update at least one first warning strategy based on the feedback information and parameter adjustment information.

[0394] In one implementation of this application, the vehicle control device 1200 may further include: a generation module, configured to acquire context information of the vehicle when executing a first target warning strategy; wherein the context information includes road condition information of the vehicle and behavior information of the occupants; and generate a motion sickness risk profile of the occupants based on the road condition information and behavior information; wherein the motion sickness risk profile is used to execute the first target warning strategy when the vehicle is traveling in similar context information.

[0395] In one implementation of this application, in response to an occupant risk event, including a health-related risk event, the signal features include multidimensional physiological features, which include at least two of heart rate, heart rate variability (HRV), respiratory rate, and target posture features; the target posture features include at least two of head posture features, arm posture features, and trunk posture features; the detection module 1230 is used to: predict whether an occupant will experience a health-related risk event in the future based on the changing trend of the multidimensional physiological features, and obtain an event detection result.

[0396] In one implementation of this application, in response to an occupant risk event including a health-related risk event, the control module 1240 is configured to: query a multi-level second warning strategy associated with a target health abnormality type; wherein the target health abnormality type is any one of multiple health abnormality types, and the second warning strategies associated with different health abnormality types are different; determine a third target warning strategy from the multi-level second warning strategies based on the probability of the target health abnormality type in the event detection results; and control the vehicle to execute the third target warning strategy to trigger a health abnormality warning prompt.

[0397] In one implementation of this application, the third target warning strategy satisfies any one of the following: In response to the probability of a target health abnormality type being higher than the corresponding fourth probability threshold, and the changing trend of the target physiological characteristic associated with the target health abnormality type in the multidimensional physiological characteristics satisfying the first abnormality condition, the third target warning strategy is used to indicate at least one of the following: triggering a first health abnormality alert; triggering a stop and rest alert; In response to the probability of a target health abnormality type being higher than the corresponding fifth probability threshold, and the changing trend of the target physiological characteristic satisfying the second abnormality condition, the third target warning strategy is used to indicate at least one of the following: triggering a second health abnormality alert; To set the vibration intensity and control the occupant's seat. The system triggers a third target warning strategy in response to a target health anomaly type having a probability higher than the corresponding sixth probability threshold. This strategy indicates at least one of the following: triggering a third health anomaly warning; activating an Advanced Driver Assistance System (ADAS); activating the right turn signal and controlling the vehicle to slowly stop at the road boundary; controlling the vehicle to a safe location to stop. The sixth probability threshold is higher than the fifth probability threshold, and the fifth probability threshold is higher than the fourth probability threshold. The degree of abnormality of the second anomaly condition is higher than that of the first anomaly condition. The warning intensity of the third health anomaly warning is higher than that of the second health anomaly warning, and the warning intensity of the second health anomaly warning is higher than that of the first health anomaly warning.

[0398] In one implementation of this application, the control module 1240 is further configured to: determine whether the occupants make a preset effective response after the vehicle executes the third target warning strategy; if yes, stop executing the third target prediction strategy; if no, determine a fourth target warning strategy from the multi-level second warning strategies, and control the vehicle to execute the fourth target warning strategy, wherein the warning level of the fourth target warning strategy is higher than that of the third target warning strategy.

[0399] In one implementation of this application embodiment, the control module 1240 is further configured to: in response to the vehicle stopping and the driver still not outputting a valid response, control the vehicle to execute the fifth target warning strategy, which is the highest warning level in the multi-level second warning strategy; wherein the fifth target warning strategy is configured to instruct at least one of the following: control the vehicle's electronic parking system to complete the parking brake locking; control the vehicle's horn device to continuously emit a warning sound; initiate an emergency rescue call based on the vehicle's location information; push a vehicle warning notification to a preset emergency contact; control the vehicle's door locks to perform an unlocking action and activate the hazard warning lights; control the vehicle's windows to perform a window ventilation operation; and cyclically broadcast the vehicle's location information.

[0400] In one implementation of this application, the occupant includes a driver. The driver's failure to respond effectively includes at least one of the following: within the target determination time after the third target warning strategy is triggered, the driver's updated physiological characteristics are acquired, and based on the updated physiological characteristics, it is determined that the driver's body is in an inactive state; wherein, the target determination time is associated with the third target warning strategy; within the target determination time, no effective control action of the driver on the steering wheel is detected; within the target determination time, no voice command issued by the driver is recognized.

[0401] In one implementation of this application, the vehicle control device 1200 may further include: a second adjustment module, used to acquire environmental information of the vehicle driving; wherein the environmental information includes at least one of the following: the location information of the road segment the vehicle is driving, the degree of congestion, weather information, and ambient brightness; and adjust the third target warning strategy according to the environmental information.

[0402] It should be noted that the explanation of the vehicle control method embodiment performed on the vehicle described above also applies to the vehicle control device of this embodiment, and will not be repeated here.

[0403] In the vehicle control device of this application embodiment, by selectively extracting signal features adapted to the feature types associated with each occupant risk event from the CSI signals sensed in the cabin by the WiFi sensing node, and detecting each occupant risk event accordingly, the accuracy and reliability of event detection results can be improved. Based on accurate event detection results, the vehicle can be controlled to execute an adapted vehicle control strategy, allowing the vehicle's control / intervention actions to precisely match the occupant's real-time state, avoiding experience disruption caused by ineffective control. This avoids the privacy leakage risks of image acquisition schemes in related technologies and increases occupants' acceptance and willingness to use in-vehicle sensing functions.

[0404] To implement the above embodiments, this application also proposes a vehicle, including: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to implement the vehicle control method as described in any of the foregoing embodiments.

[0405] Figure 13 This is a block diagram illustrating a vehicle 1300 according to an exemplary embodiment. For example, vehicle 1300 may be a hybrid vehicle, a non-hybrid vehicle, an electric vehicle, a fuel cell vehicle, or other types of vehicle. Vehicle 1300 may be a driver-assisted vehicle, a semi-driver-assisted vehicle, or a driver-free vehicle.

[0406] Reference Figure 13The vehicle 1300 may include various subsystems, such as an infotainment system 1310, a perception system 1320, a decision control system 1330, a drive system 1340, and a computing platform 1350. The vehicle 1300 may also include more or fewer subsystems, and each subsystem may include multiple components. Furthermore, each subsystem and component of the vehicle 1300 can be interconnected via wired or wireless means.

[0407] In some embodiments, the infotainment system 1310 may include a communication system, an entertainment system, and a navigation system, etc.

[0408] The perception system 1320 may include several sensors for sensing information about the environment surrounding the vehicle 1300. For example, the perception system 1320 may include a global positioning system (which may be a GPS system, a BeiDou system, or another positioning system), an inertial measurement unit (IMU), a lidar, a millimeter-wave radar, an ultrasonic radar, and a camera device.

[0409] The decision control system 1330 may include a computing system, a vehicle controller, a steering system, a throttle, and a braking system.

[0410] The drive system 1340 may include components that provide powered motion to the vehicle 1300. In one embodiment, the drive system 1340 may include an engine, an energy source, a transmission system, and wheels. The engine may be one or a combination of internal combustion engines, electric motors, and compressed air engines. The engine is capable of converting energy provided by the energy source into mechanical energy.

[0411] Some or all of the functions of the vehicle 1300 are controlled by a computing platform 1350. The computing platform 1350 may include at least one processor 1351 and a memory 1352, the processor 1351 being able to execute instructions 1353 stored in the memory 1352.

[0412] Processor 1351 can be any conventional processor, such as a commercially available CPU. Processors may also include graphics processing units (GPUs), field-programmable gate arrays (FPGAs), systems-on-chips (SoCs), application-specific integrated circuits (ASICs), or combinations thereof.

[0413] The memory 1352 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0414] In addition to instruction 1353, memory 1352 can also store data, such as road maps, route information, vehicle position, direction, speed, and other data. The data stored in memory 1352 can be used by computing platform 1350.

[0415] In this embodiment of the application, processor 1351 may execute instruction 1353 to complete all or part of the steps of any of the above method embodiments.

[0416] To implement the above embodiments, this application also proposes a chip, wherein the chip includes an interface circuit and a processing circuit coupled to each other. The interface circuit is used to input or output signals, and the processing circuit is configured to execute the vehicle control method provided in any of the foregoing embodiments.

[0417] Figure 14 This is a schematic diagram of a chip structure proposed as an exemplary embodiment of this application. See also... Figure 14 The diagram shown is a schematic representation of the structure of chip 1400, but it is not limited to this.

[0418] Chip 1400 includes processing circuitry 1401, which is configured to execute any of the above vehicle control methods.

[0419] In some embodiments, chip 1400 further includes one or more interface circuits 1402. Optionally, interface circuit 1402 is connected to memory 1403, and interface circuit 1402 can be used to receive signals from memory 1403 or other devices, and interface circuit 1402 can be used to send signals to memory 1403 or other devices. For example, interface circuit 1402 can read instructions stored in memory 1403 and send the instructions to processing circuit 1401.

[0420] In some embodiments, the interface circuit 1402 performs at least one of the communication steps such as sending and / or receiving in the above method, while the processing circuit 1401 performs other steps.

[0421] In some embodiments, the terms interface circuit, interface, transceiver pin, transceiver, etc., can be used interchangeably.

[0422] In some embodiments, chip 1400 further includes one or more memories 1403 for storing instructions. Optionally, all or part of the memories 1403 may be located outside of chip 1400.

[0423] To implement the above embodiments, this application also proposes a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the vehicle control method as described in any of the foregoing method embodiments.

[0424] To implement the above embodiments, this application also proposes a computer program product having a computer program stored thereon, which, when executed by a processor, implements the vehicle control method as described in any of the foregoing method embodiments.

[0425] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0426] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0427] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0428] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and compact disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0429] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0430] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0431] Furthermore, the functional units in the various embodiments of this application can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0432] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of this application.

Claims

1. A vehicle control method, characterized in that, include: Acquire the Channel State Information (CSI) signal obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin. Based on at least one occupant risk event associated feature type, extract signal features associated with the feature type from the CSI signal; The occupant risk event is detected based on the signal characteristics to obtain an event detection result; wherein, the event detection result is used to determine the occupant risk status of the occupant in the occupant risk event. Control the vehicle to execute a vehicle control strategy adapted to the event detection results.

2. The method according to claim 1, characterized in that, The event detection results include the probability that the occupants in the cabin are in their respective occupant risk states under the occupant risk event; The occupant risk events include comfort-related risk events. The event detection results are used to indicate the probability of each occupant in the cabin experiencing multiple thermal comfort states under the comfort-related risk events during the current time period. And / or, The occupant status events include health-related risk events, and the event detection results are used to indicate the probability of the occupant experiencing multiple types of health abnormalities under the health-related risk events in the future period. And / or, The occupant status events include motion sickness events, and the event detection results are used to indicate the probability of the occupant experiencing various motion sickness states under the motion sickness event during the current time period.

3. The method according to claim 2, characterized in that, In response to the occupant risk events, including comfort-related risk events, the signal characteristics include micro-motion characteristics corresponding to multiple seating areas within the cabin. The extraction of signal features associated with at least one occupant risk event from the CSI signal, based on a feature type, includes: From the CSI signal, extract the CSI local signals corresponding to multiple seat areas in the cockpit; wherein, the CSI signal includes complex channel response data of multiple subcarriers collected by multiple receiving antennas in the antenna array of the WiFi sensing node; Based on the target frequency band associated with multiple micro-movements, the time-frequency characteristics of the CSI local signal for each seat region are extracted; Based on the difference in the angle of arrival (AOA) between the complex channel response data collected by different receiving antennas in the CSI local signal of each seat area, the spatial characteristics of the CSI local signal of each seat area are determined; The time-frequency features and spatial features corresponding to each seat area are fused to obtain the micro-motion features of each seat area.

4. The method according to claim 2, characterized in that, In response to the occupant risk events, including comfort-related risk events, the signal characteristics include micro-motion characteristics corresponding to multiple seating areas within the cabin. The step of detecting the occupant risk event based on the signal characteristics and obtaining the event detection result includes: The common features among the micro-motion features of the multiple seating areas are extracted using the coding network in the thermal comfort prediction model. The occupant detection network in the thermal comfort prediction model is used to detect the presence of occupants based on the common features, and the occupant detection results are obtained; wherein, the occupant detection results are used to indicate whether there are occupants in each of the seat areas; Activate the target decoding network corresponding to the target seat area among the multiple decoding networks in the thermal comfort prediction model; wherein, the target seat area is the seat area where the occupant detection result indicates the presence of an occupant, and each decoding network corresponds to one seat area; The common features are decoded using the target decoding network to output a probability distribution; wherein the probability distribution is used to indicate the probability that an occupant in the target seating area will experience the various thermal comfort states during the current time period.

5. The method according to claim 4, characterized in that, The training methods for the thermal comfort prediction model include: Acquire training data; wherein the training data includes sample micro-motion features, thermal comfort annotation information, occupant annotation information and micro-motion annotation information associated with at least one sample seat area; the thermal comfort annotation information is used to indicate the thermal comfort state of the occupant in the sample seat area, the occupant annotation results are used to indicate whether there is an occupant in the sample seat area, and the micro-motion annotation information is used to indicate the type of micro-motion performed by the occupant; The thermal comfort prediction model is used to process the micro-motion characteristics of the sample to obtain the output results; wherein, the output results include thermal comfort prediction information, occupant prediction information and micro-motion prediction information of the sample seat area; The thermal comfort prediction model is trained based on the differences between the thermal comfort labeling information and the thermal comfort prediction information, the differences between the occupant labeling information and the occupant prediction information, and the differences between the micro-motion labeling information and the micro-motion prediction information.

6. The method according to claim 2, characterized in that, The occupant risk events include comfort-related risk events. The event detection results are used to indicate the probability that an occupant in a target seating area within the cabin will experience any of the various thermal comfort states during the current time period. The control of the vehicle to execute a vehicle control strategy adapted to the event detection result includes: Based on the probabilities of the various thermal comfort states, a target thermal comfort state is determined from the various thermal comfort states; Based on the target thermal comfort state corresponding to the target seating area, a temperature control strategy is determined; According to the temperature control strategy, the operation of the vehicle components associated with the target seating area is controlled. The vehicle-mounted components include at least one of the following: vehicle air conditioning, seat heating components, and seat ventilation components.

7. The method according to claim 6, characterized in that, The step of generating a temperature control strategy based on the target thermal comfort state corresponding to the target seating area includes: Based on the vehicle's remaining driving range and trip planning information, predict the vehicle's remaining state of charge (SOC) when the vehicle reaches its destination. Based on the actual SOC and the remaining SOC of the vehicle battery, the target energy mode that the vehicle must meet is determined from multiple preset energy modes. The temperature control strategy is generated based on the target energy mode and the target thermal comfort state of the target seating area.

8. The method according to claim 7, characterized in that, The target energy mode includes: a first energy mode, a second energy mode, or a third energy mode, wherein, In the first energy mode, the actual SOC of the vehicle is greater than the set first SOC, and the remaining SOC of the vehicle when it reaches its destination is greater than the set second SOC; In the second energy mode, the actual SOC of the vehicle is greater than the set third SOC and less than or equal to the first SOC, and the remaining SOC of the vehicle when it reaches its destination is greater than the second SOC; wherein, the third SOC is greater than the second SOC; In the third energy mode, the vehicle's actual SOC is less than or equal to the third SOC, and the vehicle's remaining SOC upon arrival at the destination is greater than the second SOC; or, in the third energy mode, the vehicle's remaining SOC upon arrival at the destination is less than the second SOC.

9. The method according to claim 8, characterized in that, In response to the target energy mode being the first energy mode, the step of generating the temperature control strategy based on the target energy mode and the target thermal comfort state of the target seating area includes: Based on the target thermal comfort state of the target seating area, the required temperature of the target seating area is determined, and from the vehicle components associated with the target seating area, the target vehicle components adapted to the target thermal comfort state are determined. In response to the fact that the number of target seating areas is one, a temperature control strategy for controlling the target vehicle component is generated based on the required temperature. In response to the fact that there are multiple target seating areas, the required temperatures of the multiple target seating areas are weighted according to the weighted weights corresponding to the multiple target seating areas to obtain the target required temperature, and the temperature control strategy for controlling the target vehicle components is generated based on the target required temperature. The weighted weight is used to indicate the degree of influence of occupants in the corresponding target seating area on the safe operation of the vehicle.

10. The method according to claim 9, characterized in that, The event identification result also includes: confidence levels for multiple target seating areas, wherein the confidence levels are used to indicate the uncertainty of the target thermal comfort state of the occupant in the corresponding target seating area; The method further includes: The weighted weights corresponding to the multiple target seat regions are adjusted based on the confidence levels of the multiple target seat regions.

11. The method according to claim 8, characterized in that, The step of generating a temperature control strategy based on the target energy mode and the target thermal comfort state of the target seating area includes: Based on the target thermal comfort state, the temperature adjustment direction is determined, and a target vehicle component that is adapted to the target thermal comfort state is determined from the vehicle components associated with the target seating area. In response to the target energy mode being the second energy mode, a temperature control strategy for controlling the target vehicle component is generated based on the temperature adjustment direction and the set first temperature adjustment range. In response to the target energy mode being the third energy mode, a temperature control strategy for controlling the target vehicle component is generated based on the temperature adjustment direction and the set second temperature adjustment range. The second temperature adjustment range is smaller than the first temperature adjustment range.

12. The method according to claim 10, characterized in that, The temperature control strategy satisfies any one of the following: In response to the target energy mode being the second energy mode and the temperature adjustment direction being positive, the temperature control strategy is used to indicate at least one of the following: the temperature adjustment amount of the vehicle air conditioner, turning off the vehicle air conditioner that is not corresponding to the target seat area, and activating the seat heating function under the target seat area; In response to the target energy mode being the second energy mode and the temperature adjustment direction being negative, the temperature control strategy is used to indicate at least one of the following: the temperature adjustment amount of the vehicle air conditioner, turning off the vehicle air conditioner that is not corresponding to the target seat area, and activating the seat ventilation function under the target seat area; In response to the target energy mode being the third energy mode, the temperature control strategy is used to indicate at least one of the following: the required temperature of the driver's seat area in the target seating area, turning off the vehicle air conditioning corresponding to the non-target seating area, and triggering an energy-saving reminder; wherein the energy-saving reminder is used to indicate that the vehicle is currently in energy-saving mode.

13. The method according to claim 2, characterized in that, In response to the occupant status events, including health-related risk events and / or motion sickness events, the signal characteristics include the multidimensional physiological characteristics of the occupants in the cabin; The extraction of signal features associated with at least one occupant risk event from the CSI signal, based on a feature type, includes: The CSI signal is preprocessed, and the multidimensional physiological characteristics of the occupants in the cabin are extracted from the preprocessed CSI signal; wherein the preprocessing includes at least one of the following: A notch filter adapted to the vibration frequency of the vibrating components in the vehicle is used to filter the CSI signal to eliminate vibration noise introduced by the vibrating components; wherein, the center frequency of the notch filter is related to the rotational speed of the vibrating components, and the vibrating components include an engine and / or a motor; A Kalman filter is used to perform fusion filtering on the CSI signal based on the motion data collected by the vehicle's inertial measurement unit (IMU) to eliminate vibration noise introduced by road bumps; wherein, the IMU is time-synchronized with the WiFi sensing node; The CSI signal is filtered using a Butterworth filter that is adapted to the signal frequency band corresponding to the multidimensional physiological characteristics to eliminate out-of-band interference signals; wherein, the bandpass range of the Butterworth filter is determined according to the motion state of the occupant. Motion detection is performed on the CSI signal to mark target segments in the CSI signal; wherein the target segment indicates that the occupant's movement amplitude is higher than a set amplitude, the marking is used to reduce the weight of the target segment when extracting the multidimensional physiological features, and / or the marking is used to perform interpolation filling processing on the target segment based on the adjacent segments of the target segment; The CSI signal is phase-corrected using target correction parameters to eliminate nonlinear errors caused by temperature drift.

14. The method according to claim 13, characterized in that, The method for determining the target correction parameters includes: Acquire the raw CSI signal collected by the WiFi sensing node in the most recent calibration period; wherein the raw CSI signal is in matrix form, and the element in the i-th row and j-th column of the matrix is ​​used to indicate the complex channel response data of the j-th subcarrier collected by the i-th receiving antenna in the antenna array of the WiFi sensing node; A phase correction model is constructed with correction parameters as variables, wherein the phase correction model is used to subtract the temperature drift component determined by the correction parameters from the phase component of the original CSI signal; The objective function is to minimize the first-order difference variance in the time dimension of the CSI phase sequence corresponding to the same subcarrier and the same receiving antenna after correction by the phase correction model, and then solve the phase correction model to obtain the target correction parameters. The CSI phase sequence includes the phase components of each original CSI signal acquired in the most recent correction cycle.

15. The method according to claim 2, characterized in that, In response to the occupant risk event, including motion sickness, the signal features include multidimensional physiological features, including at least two of the following: heart rate, heart rate variability (HRV) index, respiratory rate, microkinetic energy caused by friction between skin and clothing determined based on the high-frequency component in the CSI signal, and head posture features. The step of detecting the occupant risk event based on the signal characteristics and obtaining the event detection result includes: Obtain the reference physiological characteristics of the occupant; wherein the reference physiological characteristics are determined based on the occupant's historical physiological data; Determine the difference data between the reference physiological characteristics and the multidimensional physiological characteristics; The motion sickness event is detected based on the difference data, and the event detection result is obtained.

16. The method according to claim 2, characterized in that, In response to the occupant risk event, including motion sickness, controlling the vehicle to execute a vehicle control strategy adapted to the event detection result includes: Based on the probability of various motion sickness states in the event detection results, a first target warning strategy is determined from the preset multi-level first warning strategies; Control the vehicle to execute the first target warning strategy to trigger a motion sickness warning.

17. The method according to claim 16, characterized in that, The multiple motion sickness states include a first motion sickness state and a second motion sickness state, wherein the degree of motion sickness in the second motion sickness state is higher than that in the first motion sickness state, and the first target warning strategy satisfies any one of the following: In response to the probability of the first motion sickness state being higher than a first probability threshold, the first target warning strategy is used to prompt the passenger to wear smart glasses and / or trigger a motion sickness risk warning. In response to the probability of the second motion sickness state being higher than the second probability threshold, the first target warning strategy is used to instruct the activation of the first visual intervention mechanism in the smart glasses to alleviate the passenger's anxiety and nausea. In response to the probability of the second motion sickness state being higher than a third probability threshold, the first target warning strategy is used to indicate at least one of the following: activating the second visual intervention mechanism in the smart glasses to alleviate the degree of motion sickness of the occupant; triggering a stop and rest prompt; triggering a window ventilation prompt; Wherein, the third probability threshold is higher than the second probability threshold, and the second probability threshold is higher than the first probability threshold, and the intervention intensity of the second visual intervention mechanism is higher than that of the first visual intervention mechanism.

18. The method according to claim 17, characterized in that, The first visual intervention mechanism includes at least one of the following: displaying a fixed reference point to establish a stable reference frame for the occupant's visual system; displaying breathing guidance material to guide the occupant to take deep breaths; and adjusting at least one of the color temperature, brightness, and hue of the displayed content. And / or, The second visual intervention mechanism includes at least one of the following: displaying a virtual horizon line to simulate a horizon; reducing the refresh rate of the displayed content; performing motion compensation on the displayed content; and blurring the visual field edges of the displayed content to guide the occupant to focus on the central area of ​​the displayed content.

19. The method according to claim 18, characterized in that, The display of the fixed reference point includes: The occupant's head posture is tracked by the inertial measurement unit (IMU) of the smart glasses; Obtain the fixed display position of the fixed reference point in the virtual space; Based on the head posture, the set compensation coefficient, and the fixed display position, the actual display position of the fixed reference point on the smart glasses is determined; Based on the actual display position, the fixed reference point is displayed on the smart glasses to keep the fixed reference point fixed in the virtual space.

20. The method according to claim 18, characterized in that, The breathing guidance material is displayed in the form of animation, and the display of the breathing guidance material includes: The reference graphic being displayed is periodically contracted and expanded using a set contraction and expansion cycle. The contraction process is used to guide the occupant to exhale; the expansion process is used to guide the occupant to inhale.

21. The method according to claim 20, characterized in that, In response to the probability of the second motion sickness state being higher than the second probability threshold, the first target warning strategy is further used to instruct the vehicle to broadcast breathing prompt information associated with the breathing guidance material.

22. The method according to any one of claims 16-21, characterized in that, The method further includes: Obtain the latest physiological characteristics of the occupants after the vehicle executes the first warning strategy; Based on the latest physiological characteristics, determine whether the degree of motion sickness of the occupant has decreased; If not, then a second target early warning strategy is determined from the first early warning strategy of multiple levels, wherein the early warning level of the second target early warning strategy is higher than that of the first target early warning strategy; Control the vehicle to execute the second target warning strategy.

23. The method according to any one of claims 16-21, characterized in that, The method further includes: Obtain feedback information and parameter adjustment information from the occupants regarding at least one of the first warning strategies; Based on the feedback information and the parameter adjustment information, at least one of the first early warning strategies is updated.

24. The method according to any one of claims 16-21, characterized in that, The method further includes: Obtain the context information of the vehicle when the first target warning strategy is executed; wherein, the context information includes the road condition information of the vehicle and the behavior information of the occupants; Based on the road condition information and the behavioral information, a motion sickness risk profile of the passenger is generated; The motion sickness risk profile is used to execute the first target warning strategy when the vehicle is traveling in similar context information.

25. The method according to claim 2, characterized in that, In response to the occupant risk event including a health risk event, the signal features include multidimensional physiological features, which include at least two of heart rate, heart rate variability (HRV) index, respiratory rate, and target posture features; the target posture features include at least two of head posture features, arm posture features, and trunk posture features. The step of detecting the occupant risk event based on the signal characteristics and obtaining the event detection result includes: Based on the changing trends of the multidimensional physiological characteristics, it is predicted whether the occupant will experience a health-related risk event in the future, and the event detection results are obtained.

26. The method according to claim 2, characterized in that, In response to the aforementioned occupant risk events, including health-related risk events, The control of the vehicle to execute a vehicle control strategy adapted to the event detection result includes: Query the multi-level second early warning strategy associated with the target health abnormality type; wherein, the target health abnormality type is any one of the multiple health abnormality types, and the second early warning strategy associated with different health abnormality types is different; Based on the probability of the target health abnormality type in the event detection results, a third target early warning strategy is determined from the multi-level second early warning strategy; Control the vehicle to execute the third target warning strategy to trigger a health abnormality warning.

27. The method according to claim 26, characterized in that, The third target early warning strategy satisfies any one of the following: In response to the probability of the target health abnormality type being higher than the corresponding fourth probability threshold, and the changing trend of the target physiological feature associated with the target health abnormality type among the multidimensional physiological features satisfying the first abnormality condition, the third target early warning strategy is used to indicate at least one of the following: triggering the first health abnormality prompt; Trigger a parking rest prompt; In response to the probability of the target health abnormality type being higher than the corresponding fifth probability threshold, and the trend of the target physiological characteristics meeting the second abnormality condition, the third target early warning strategy is used to indicate at least one of the following: triggering a second health abnormality prompt; The vibration intensity is set to control the seat vibration in the occupant's seating area; the vehicle speed is reduced; and the hazard warning lights are activated. In response to the probability of the target health abnormality type being higher than the corresponding sixth probability threshold, the third target warning strategy is used to indicate at least one of the following: triggering a third health abnormality prompt; activating the Advanced Driver Assistance System (ADAS); turning on the right turn signal and controlling the vehicle to slowly stop at the road boundary; controlling the vehicle to drive to a safe position to stop; Wherein, the sixth probability threshold is higher than the fifth probability threshold, and the fifth probability threshold is higher than the fourth probability threshold; the degree of abnormality of the second abnormal condition is higher than that of the first abnormal condition; the intensity of the third health abnormality prompt is higher than that of the second health abnormality prompt, and the intensity of the second health abnormality prompt is higher than that of the first health abnormality prompt.

28. The method according to claim 26, characterized in that, The method further includes: Determine whether the occupants make a preset effective response after the vehicle executes the third target warning strategy; If so, then stop executing the third target prediction strategy; If not, a fourth target warning strategy is determined from the multi-level second warning strategy, and the vehicle is controlled to execute the fourth target warning strategy, wherein the warning level of the fourth target warning strategy is higher than that of the third target warning strategy.

29. The method according to claim 28, characterized in that, The method further includes: In response to the vehicle stopping and the driver still not outputting the valid response, the vehicle is controlled to execute the fifth target warning strategy, which is the highest warning level in the multi-level second warning strategy; The fifth target warning strategy is used to instruct at least one of the following: control the vehicle's electronic parking system to complete the parking brake lock; control the vehicle's horn device to continuously emit a warning sound; initiate an emergency rescue call based on the vehicle's location information; push a vehicle warning notification to a preset emergency contact; control the vehicle's door locks to perform an unlocking action and activate the hazard warning lights; control the vehicle's windows to perform a window ventilation operation; and cyclically broadcast the vehicle's location information.

30. The method according to claim 28 or 29, characterized in that, The occupants include the driver, and the driver's failure to respond effectively includes at least one of the following: Within the target determination time after the third target warning strategy is triggered, the driver's updated physiological characteristics are acquired, and based on the updated physiological characteristics, it is determined that the driver's body is in an inactive state; wherein, the target determination time is associated with the third target warning strategy; During the target determination period, no effective steering wheel control action by the driver was detected. No voice commands issued by the driver were recognized within the target determination period.

31. The method according to any one of claims 26-29, characterized in that, The method further includes: Obtain environmental information about the vehicle's driving environment; wherein, the environmental information includes at least one of the following: the location information of the road segment the vehicle is traveling on, the degree of congestion, weather information, and ambient brightness; Based on the environmental information, adjust the early warning strategy for the third target.

32. A vehicle control device, characterized in that, include: The acquisition module is used to acquire the Channel State Information (CSI) signal obtained by at least one WiFi sensing node in the vehicle cabin sensing the cabin. An extraction module is used to extract signal features associated with at least one occupant risk event from the CSI signal based on the feature type associated with the event. The detection module is used to detect the occupant risk event based on the signal characteristics and obtain an event detection result; wherein the event detection result is used to determine the occupant risk status of the occupant in the occupant risk event. The control module is used to control the vehicle to execute a vehicle control strategy adapted to the event detection results.

33. The apparatus according to claim 32, characterized in that, The event detection results include the probability that the occupants in the cabin are in their respective occupant risk states under the occupant risk event; The occupant risk events include comfort-related risk events. The event detection results are used to indicate the probability of each occupant in the cabin experiencing multiple thermal comfort states under the comfort-related risk events during the current time period. And / or, The occupant status events include health-related risk events, and the event detection results are used to indicate the probability of the occupant experiencing multiple types of health abnormalities under the health-related risk events in the future period. And / or, The occupant status events include motion sickness events, and the event detection results are used to indicate the probability of the occupant experiencing various motion sickness states under the motion sickness event during the current time period.

34. The apparatus according to claim 33, characterized in that, In response to the occupant risk events, including comfort-related risk events, and given that the signal characteristics include micro-motion characteristics corresponding to multiple seating areas within the cabin, the detection module is used for: The common features among the micro-motion features of the multiple seating areas are extracted using the coding network in the thermal comfort prediction model. The occupant detection network in the thermal comfort prediction model is used to detect the presence of occupants based on the common features, and the occupant detection results are obtained; wherein, the occupant detection results are used to indicate whether there are occupants in each of the seat areas; Activate the target decoding network corresponding to the target seat area among the multiple decoding networks in the thermal comfort prediction model; wherein, the target seat area is the seat area where the occupant detection result indicates the presence of an occupant, and each decoding network corresponds to one seat area; The common features are decoded using the target decoding network to output a probability distribution; wherein the probability distribution is used to indicate the probability that an occupant in the target seating area will experience the various thermal comfort states during the current time period.

35. The apparatus according to claim 33, characterized in that, In response to the occupant risk event, including motion sickness, the signal features include multidimensional physiological features, including at least two of the following: heart rate, heart rate variability (HRV), respiratory rate, microkinetic energy caused by friction between skin and clothing determined based on the high-frequency component of the CSI signal, and head posture features; the detection module is used for: Obtain the reference physiological characteristics of the occupant; wherein the reference physiological characteristics are determined based on the occupant's historical physiological data; Determine the difference data between the reference physiological characteristics and the multidimensional physiological characteristics; The motion sickness event is detected based on the difference data, and the event detection result is obtained.

36. The apparatus according to claim 33, characterized in that, In response to the occupant risk event including a health-related risk event, the signal features include multidimensional physiological features, which include at least two of heart rate, heart rate variability (HRV), respiratory rate, and target posture features; the target posture features include at least one of head posture features, arm posture features, and trunk posture features; the detection module is used for: Based on the changing trends of the multidimensional physiological characteristics, it is predicted whether the occupant will experience a health-related risk event in the future, and the event detection results are obtained.

37. A vehicle, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured as follows: The steps of implementing the method as described in any one of claims 1 to 31.

38. A non-transitory computer-readable storage medium having computer program instructions stored thereon, characterized in that, When executed by a processor, the program instructions implement the steps of the method described in any one of claims 1 to 31.

39. A chip, characterized in that, The chip includes an interface circuit and a processing circuit that are coupled to each other. The interface circuit is used to input or output signals, and the processing circuit is used to implement the method of any one of claims 1 to 31.

40. A computer program product, characterized in that, Includes a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 31.

Images