Seat adjustment method and vehicle

Abstract

The invention provides a seat adjusting method and a vehicle, and relates to the technical field of vehicles, the method comprises the steps that information of a current user on a target seat is acquired, and the information of the current user comprises a target body type feature vector and identification information of the current user; obtaining an initial seat adjustment parameter based on the target body type feature vector, and obtaining a seat adjustment parameter of the current user based on the identification information; adjusting the initial seat adjustment parameter based on the seat adjustment parameter to obtain a first seat adjustment parameter of the current user; and adjusting the target seat based on the first seat adjustment parameter. According to the method, the intelligence of seat adjustment can be improved, so that the driving experience of a user is improved.

Description

Technical Field

[0001] The present application relates to the field of vehicle technology, and more particularly, to a seat adjustment method and a vehicle in the field of vehicle technology. Background Art

[0002] As users' demands for driving comfort increase, seat adjustment becomes increasingly important. Currently, seat adjustment methods primarily rely on manual adjustment, memory seats, and automated adjustment technologies based on fixed rules. However, these methods have limitations and struggle to intelligently adjust seats to individual needs, which can negatively impact the user's driving experience.

[0003] Therefore, how to improve the intelligence of seat adjustment to enhance the user's driving experience is a technical problem that urgently needs to be solved. Summary of the Invention

[0004] The present application provides a seat adjustment method and a vehicle, which can improve the intelligence of seat adjustment to enhance the user's driving experience.

[0005] In a first aspect, a method for seat adjustment is provided, the method comprising: obtaining information of a current user on a target seat, the information of the current user including a target body shape feature vector and identification information of the current user; obtaining initial seat adjustment parameters based on the target body shape feature vector, and obtaining seat adjustment parameters of the current user based on the identification information; adjusting the initial seat adjustment parameters based on the seat adjustment parameters to obtain first seat adjustment parameters of the current user; and adjusting the target seat based on the first seat adjustment parameters.

[0006] In an embodiment of the present application, the target body shape feature vector of the current user and the identification information of the current user are obtained; based on the target body shape feature vector, initial seat adjustment parameters are obtained, and based on the identification information, seat adjustment parameters of the current user are obtained; based on the seat adjustment parameters, the initial seat adjustment parameters are adjusted to obtain the first seat adjustment parameters of the current user; based on the first seat adjustment parameters, the target seat is adjusted; since the initial seat adjustment parameters in this solution are obtained based on the target body shape feature vector, the seat adjustment parameters are obtained based on the identification information of the current user, and the seat adjustment parameters corresponding to the current user are obtained based on the initial seat adjustment parameters and the seat adjustment parameters. That is, when determining the seat adjustment parameters of the current user, both the body shape characteristics of the current user and the personalized needs of the current user are taken into consideration, and thus personalized seat adjustment parameters for the current user can be provided based on both the body shape characteristics and the personalized needs of the current user, thereby improving the intelligence of the seat adjustment to enhance the user's driving experience.

[0007] In combination with the first aspect, in certain implementations of the first aspect, after the target seat is adjusted based on the first seat adjustment parameters, the method provided in the embodiment of the present application also includes: determining whether a feedback operation is detected, the feedback operation being the adjustment operation of the target seat by the current user; if a feedback operation is detected, optimizing the seat adjustment parameters of the current user based on the adjustment parameters corresponding to the feedback operation; and using the optimized seat adjustment parameters as the seat adjustment parameters of the current user.

[0008] In an embodiment of the present application, if a feedback operation is detected, the seat adjustment parameters for the current user can be optimized. Because the personalized adjustment parameters of different users may vary, in this solution, when a change in a user's personalized adjustment parameters is detected, the user's feedback operation can be used to update the user's personalized adjustment parameters, thereby improving the adaptability of the seat adjustment parameters to the current user and thereby enhancing the intelligence of the seat adjustment.

[0009] In combination with the first aspect and the above-mentioned implementation methods, in certain implementation methods of the first aspect, after adjusting the target seat based on the first seat adjustment parameter, the method provided in the embodiment of the present application also includes: detecting the current state of the current user; obtaining the second seat adjustment parameter of the target seat based on the current state; and adjusting the target seat based on the second seat adjustment parameter.

[0010] In an embodiment of the present application, the second seat adjustment parameter is determined based on the current state of the current user, and then the target seat is adjusted based on the second seat adjustment parameter; since the current state can reflect whether the current user is in a fatigue state, the target seat suitable for the current user's sitting posture can be determined through the current state, thereby reducing the incompatibility of seat adjustment parameters caused by user fatigue, and intelligently adjusting the target seat to improve the user's driving experience.

[0011] In combination with the first aspect and the above-mentioned implementation methods, in certain implementation methods of the first aspect, a second seat adjustment parameter of the target seat is obtained based on the current state, including: determining whether the current state is a fatigue state; if the current state is a fatigue state, obtaining the second seat adjustment parameter based on the current posture of the current user.

[0012] In an embodiment of the present application, when the current state is a fatigue state, a second seat adjustment parameter is obtained based on the current posture of the current user; since the current posture is associated with the user's fatigue state, further adjusting the target seat based on the second adjustment parameter obtained based on the current posture can reduce the current user's fatigue and thereby improve driving safety.

[0013] In combination with the first aspect and the above-mentioned implementation methods, in certain implementation methods of the first aspect, determining whether the current state is a fatigue state includes: if the vehicle's driving time is greater than or equal to a preset time, detecting whether the current user's posture is an abnormal posture; if the current user's posture is an abnormal posture, determining that the current state is a fatigue state; if the current user's posture is a posture other than an abnormal posture, determining that the current state is not a fatigue state.

[0014] In an embodiment of the present application, when the vehicle's driving time is greater than or equal to a preset time and the current user's posture is abnormal, the current state is determined to be a fatigue state; since the fatigue state is associated with the driving time and the user's posture, it is possible to quickly and accurately detect that the user is in a fatigue state at this time, and then adjust the target seat when the user is in a fatigue state, thereby improving the intelligence of the target seat adjustment.

[0015] In combination with the first aspect and the above-mentioned implementation methods, in some implementation methods of the first aspect, obtaining the second seat adjustment parameter based on the current posture of the current user includes: obtaining the second seat adjustment parameter based on the identification information of the current user and the current posture of the current user.

[0016] In an embodiment of the present application, the personalized second seat adjustment parameters corresponding to the current posture of the current user can be accurately obtained through the identification information and current posture of the current user, so that the target seat can be intelligently adjusted based on the second seat adjustment parameters to improve the user's driving experience.

[0017] In combination with the first aspect and the above-mentioned implementation methods, in certain implementation methods of the first aspect, the method provided in the embodiments of the present application also includes: taking the initial seat adjustment parameters as the initial parameters, adjusting the initial parameters with a first adjustment amount to obtain the target parameters, and the first adjustment amount is any adjustment amount in the seat adjustment amount set; when adjusting the target seat through the target parameters, determining the reward value of the first adjustment amount based on the user's feedback information on the target seat and the first adjustment amount; training the target reinforcement learning model based on the reward value of each adjustment amount in the seat adjustment amount set to obtain the trained target reinforcement learning model; wherein, the trained target reinforcement learning model is used to output the seat adjustment parameters of the user, and the user includes the current user.

[0018] In an embodiment of the present application, the target reinforcement learning model is trained to obtain a trained target reinforcement learning model. Since the target reinforcement learning model itself can be trained through the adjustment amounts in the seat adjustment amount set, the seat adjustment amount can be accurately output through the trained target reinforcement learning model when the current user subsequently sits on the target seat, thereby further improving the intelligence of the seat adjustment.

[0019] In combination with the first aspect and the above-mentioned implementation methods, in certain implementation methods of the first aspect, obtaining information of the current user includes: obtaining a target image of the current user and a pressure value of a pressure sensor in a target seat, the target image including a color image and a depth image; inputting the color image, the depth image, and the pressure value into a body shape detection model to obtain a target body shape feature vector; wherein the body shape detection model is used to obtain posture key points of the current user based on the color image and the depth image, and to obtain a target body shape feature vector based on the posture key points and the pressure value.

[0020] In an embodiment of the present application, since the body shape detection model can automatically process the data of the current user to generate the body shape feature vector of the current user, the target body shape feature vector of the current user can be obtained quickly and accurately by inputting the color image, depth image and pressure value of the current user into the body shape detection model, and then the target seat can be intelligently adjusted through the target body shape feature vector.

[0021] In combination with the first aspect and the above-mentioned implementation methods, in some implementation methods of the first aspect, initial seat adjustment parameters are obtained based on the target body shape feature vector, including: inputting the target body shape feature vector into a body shape clustering model to obtain initial seat adjustment parameters; wherein, the body shape clustering model is used to obtain seat adjustment parameters corresponding to the body shape feature vector based on the input body shape feature vector.

[0022] In an embodiment of the present application, since the body shape clustering model can automatically identify body shape feature groups that are similar to the user's body shape feature vector, and then determine the seat adjustment parameters corresponding to the body shape feature group, the initial seat adjustment parameters are obtained by inputting the target body shape feature vector into the body shape clustering model. In this way, the seat adjustment parameters adapted to the current user's body shape can be obtained quickly and accurately, thereby improving the intelligence of the seat adjustment.

[0023] In a second aspect, a seat adjustment device is provided, the device comprising:

[0024] An acquisition module is used to acquire information about the current user on the target seat, where the information about the current user includes a target body shape feature vector of the current user and identification information of the current user;

[0025] a determination module for obtaining initial seat adjustment parameters based on the target body shape feature vector and obtaining seat adjustment parameters for the current user based on the identification information;

[0026] The determination module is further configured to adjust the initial seat adjustment parameter based on the seat adjustment parameter to obtain a first seat adjustment parameter for the current user;

[0027] The adjustment module is configured to adjust the target seat based on the first seat adjustment parameter.

[0028] As a possible implementation method, the device provided in the embodiment of the present application also includes: an optimization module, used to determine whether a feedback operation is detected, the feedback operation is the adjustment operation of the target seat by the current user; if the feedback operation is detected, the seat adjustment parameters of the current user are optimized based on the adjustment parameters corresponding to the feedback operation; and the optimized seat adjustment parameters are used as the seat adjustment parameters of the current user.

[0029] As a possible implementation, the determination module is further configured to detect the current state of the current user; obtain a second seat adjustment parameter of the target seat based on the current state; and adjust the target seat based on the second seat adjustment parameter.

[0030] As a possible implementation manner, the determination module is specifically used to: determine whether the current state is a fatigue state; if the current state is a fatigue state, obtain the second seat adjustment parameter based on the current posture of the current user.

[0031] As a possible implementation method, the determination module is specifically used to: if the vehicle's driving time is greater than or equal to a preset time, detect whether the current user's posture is an abnormal posture; if the current user's posture is an abnormal posture, determine that the current state is a fatigue state; if the current user's posture is a posture other than an abnormal posture, determine that the current state is not a fatigue state.

[0032] As a possible implementation manner, the determination module is specifically configured to obtain the second seat adjustment parameter based on the identification information of the current user and the current posture of the current user.

[0033] As a possible implementation method, the device provided in the embodiment of the present application also includes a training module, which is used to take the initial seat adjustment parameters as the initial parameters, adjust the initial parameters with a first adjustment amount, and obtain the target parameters, where the first adjustment amount is any adjustment amount in the seat adjustment amount set; when the target seat is adjusted by the target parameters, the reward value of the first adjustment amount is determined based on the user's feedback information on the target seat and the first adjustment amount; based on the reward value of each adjustment amount in the seat adjustment amount set, the target reinforcement learning model is trained to obtain a trained target reinforcement learning model; wherein the trained target reinforcement learning model is used to output the seat adjustment parameters of the user, and the user includes the current user.

[0034] As a possible implementation method, the determination module is specifically used to: input the target body shape feature vector into the body shape clustering model to obtain initial seat adjustment parameters; wherein the body shape clustering model is used to obtain the seat adjustment parameters corresponding to the body shape feature vector based on the input body shape feature vector.

[0035] In a third aspect, a vehicle is provided, comprising a memory and a processor. The memory is configured to store executable program code, and the processor is configured to retrieve and execute the executable program code from the memory, so that the vehicle executes the method of the first aspect or any possible implementation of the first aspect.

[0036] In a fourth aspect, a computer program product is provided, comprising: a computer program code, which, when executed on a computer, enables the computer to execute the method in the first aspect or any possible implementation of the first aspect.

[0037] In a fifth aspect, a computer-readable storage medium is provided, which stores a computer program code. When the computer program code runs on a computer, the computer executes the method in the above-mentioned first aspect or any possible implementation of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a schematic diagram of a seat adjustment scenario provided by an embodiment of the present application;

[0039] Figure 2 This is a schematic diagram of the process of a seat adjustment method provided in an embodiment of the present application;

[0040] Figure 3 is a flow chart of another seat adjustment method provided in an embodiment of the present application;

[0041] Figure 4 is a flow chart of another seat adjustment method provided in an embodiment of the present application;

[0042] Figure 5 1 is a structural diagram of a seat adjustment device provided in an embodiment of the present application;

[0043] Figure 6 It is a structural schematic diagram of a vehicle provided in an embodiment of the present application. DETAILED DESCRIPTION

[0044] The following will clearly and thoroughly describe the technical solutions in this application in conjunction with the accompanying drawings. In the description of the embodiments of this application, unless otherwise specified, " / " means or, for example, A / B can mean A or B: "and / or" in the text is only a description of the association relationship of associated objects, indicating that there can be three relationships, for example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more than two.

[0045] In the following, the terms "first" and "second" are used for descriptive purposes only and should not be understood to imply or suggest relative importance or implicitly indicate the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features.

[0046] As users' demands for driving comfort increase, seat adjustment becomes more and more important. Figure 1 As shown, Figure 1 A schematic diagram of a seat adjustment scenario provided in an embodiment of the present application. In scenario 100, when the current user is sitting on the target seat, the seat can be adjusted to improve the user's driving experience. Currently, the user can adjust the backrest angle of the target seat (for example, increase it by 10 degrees (°)), adjust the position of the target seat (for example, adjust it forward by 3 centimeters (cm)), adjust the height of the target seat (for example, raise it by 2 cm), or adjust the backrest lumbar support strength of the target seat (for example, adjust the lumbar support strength to level 2).

[0047] Currently, seat adjustment methods mainly rely on manual adjustment by the user, memory seats, and automated adjustment technologies based on fixed rules. However, these methods have limitations and are difficult to intelligently adjust seats to suit different users, thus reducing the user's driving experience.

[0048] Therefore, how to improve the intelligence of seat adjustment to enhance the user's driving experience is a technical problem that urgently needs to be solved.

[0049] In view of this, an embodiment of the present application provides a method and a vehicle for seat adjustment, which can improve the intelligence of seat adjustment to enhance the user's driving experience.

[0050] In order to illustrate the technical solution of this application, the following is a description of the specific embodiments. Figures 2 to 4 The seat adjustment method provided in the embodiment of the present application is described in detail.

[0051] For example, the executor of the seat adjustment method provided in the embodiment of the present application may be a vehicle, a vehicle-side controller, or a seat adjustment device (for example, a chip) used in a vehicle; wherein the vehicle-side controller may be: a vehicle controller, a cockpit domain controller, or a central domain controller.

[0052] Figure 2 It is a flowchart of a seat adjustment method provided in an embodiment of the present application. Figure 2 It includes S201 to S204. S201 to S204 are described in detail below.

[0053] S201. Obtain information of the current user on the target seat.

[0054] The information of the current user includes the target body shape feature vector of the current user and the identification information of the current user.

[0055] Exemplarily, the target seat can be the main driver's seat in the vehicle, or the co-pilot seat, or of course other seats in the vehicle. The embodiment of the present application does not impose any specific restrictions on this.

[0056] For example, the body shape feature vector can be used to reflect the user's body structure, sitting posture habits, and physiological characteristics. For example, the target body shape feature vector can include the current user's body length ratio, sitting posture angle, and pressure distribution.

[0057] For example, the body length ratio can be used to reflect the geometric relationship between the limbs and the torso; taking the ratio between leg length and height as an example, if the ratio between leg length and height is greater than 0.6, it means that the current user has a long-legged body type; if the ratio between leg length and height is less than 0.5, it means that the current user has a short-legged body type.

[0058] For example, the sitting angle can be used to reflect the quantitative bending state of the spine, neck, and legs, that is, the spine angle, neck angle, and leg angle; taking the spine angle as an example, the normal spine angle is less than 15°.

[0059] For example, pressure distribution can be used to reflect the uniformity of body weight load and contact surface. Pressure distribution can be represented by the coordinates of the pressure center, the pressure ratio between the left and right hip areas, and the waist pressure.

[0060] As an example, obtaining the current user's information includes: obtaining a target image of the current user and a pressure value of a pressure sensor in a target seat; inputting the color image, depth image, and pressure value into a body shape detection model to obtain a target body shape feature vector.

[0061] The target image includes a color image and a depth image. The body shape detection model is used to obtain the posture key points of the current user based on the color image and the depth image, and to obtain the target body shape feature vector based on the posture key points and the pressure value.

[0062] For example, the color image can be a two-dimensional image (e.g., a red, green, and blue (RGB) image). The color image can be used to describe the current user's full body image or a partial sitting posture image. Through the color image, the current user's joints (e.g., shoulders, elbows, knees, etc.), body contours (e.g., shoulder width, sitting height), and whether the sitting posture is offset (e.g., when the body leans to the left, the right shoulder joint moves downward).

[0063] The color image can be obtained by a camera deployed inside the cockpit. The camera can be an RGB camera or an infrared (IR) camera, etc., which is not specifically limited in the present embodiment.

[0064] For example, the depth image can be 3D point cloud data. The depth image can be used to reflect information such as the user's bone length and 3D posture. It can also further address issues such as the inability to obtain information such as the user's joints and body contours due to partial occlusion of the color image (for example, a steering wheel obscuring the torso).

[0065] The depth image can be obtained by a depth sensor. The depth sensor can be a time-of-flight (ToF) sensor, a laser radar (LiDAR), etc., which is not specifically limited in the present embodiment.

[0066] For example, the pressure value of the pressure sensor in the target seat can be used to reflect the distribution of the current user's pressure on the target seat, such as the pressure areas of the buttocks and back.

[0067] Exemplarily, the body shape detection model can be a multimodal neural network model that processes the color image, depth image, and pressure values from the pressure sensors in the target seat, inputting the body shape detection model to output a target body shape feature vector. Optionally, the body shape detection model can also output the coordinates of key points of the current user's target posture.

[0068] In actual scenarios, the body shape detection model can be a posture converter (Transformer-Pose) model. The Transformer-Pose model can include an encoder and a multi-scale decoder. The encoder can globally model the posture of the current user through the self-attention mechanism to capture the correlation information between body parts (for example, the spine and legs) to reduce the interference of image occlusion through the correlation information of body parts (such as inferring the position of the right knee blocked by the steering wheel through the visible left knee). The multi-scale decoder can enhance the adaptability to different body shapes and perspectives, such as small targets at a distance or local occlusion. The multi-scale decoder can be used to fuse features at different levels to improve the detection robustness of different body shapes and posture angles.

[0069] For example, a color image, a depth image, and the pressure value of the pressure sensor in the target seat are input into the Transformer-Pose model. The encoder in the Transformer-Pose model can extract features such as texture and color from the color image to identify the body contour of the current user. At the same time, the encoder can also extract geometric information from the depth image to solve the problem of the current user being occluded in the color image. For example, the self-attention mechanism can be used to identify the posture key points of the right ankle that is occluded in the crossed-leg posture. In addition, the encoder can extract the pressure value of the pressure sensor to determine the weight distribution of the current user on the target seat. The multi-scale decoder can fuse low-level features with high-level features to output the target body shape feature vector.

[0070] For example, the encoder can divide the color image and / or depth image into fixed-size pixel blocks and linearly map them into a vector sequence. It then calculates the association information between different pixel blocks (for example, using association weights as an example, the association between the head and shoulders > the association between the head and feet) to address the problem of false detection of complex postures (such as crossing legs or sitting sideways). The multi-scale decoder in the Transformer-Pose model can then fuse low-level features (such as detailed features: finger joints) with high-level features (such as semantic features: torso posture) to accommodate different body shapes (such as joint occlusion in obese individuals). The Transformer-Pose model extracts the current user's posture key points (e.g., head, left shoulder, right shoulder, left elbow, right elbow, ..., left ankle, right ankle) based on features extracted from the color image and / or depth image. The current user's posture key points are then corrected by mapping the current user's weight distribution on the target chair to the current user's posture key point region, for example, by associating the hip pressure center with the pelvic key point offset. The target body shape feature vector of the current user is extracted through the corrected posture key points of the previous user, such as leg length ratio, spinal curvature angle, and torso width-to-height ratio. For example, through the posture key points of the hip joint, ankle joint, top of the head, and hip joint, the leg length ratio is calculated using the formula leg length / height = (distance from hip joint to ankle joint) / (distance from top of head to hip joint), and then the Transformer-Pose model is used to output the target body shape feature vector.

[0071] It can be understood that since the body shape detection model can automatically process the data of the current user to generate the body shape feature vector of the current user, by inputting the color image, depth image and pressure value of the current user into the body shape detection model, the target body shape feature vector of the current user can be obtained quickly and accurately, and then the target seat can be intelligently adjusted through the target body shape feature vector.

[0072] For example, identification information can be information that uniquely identifies a user, such as a user's identity document (ID). Identification information can be used to associate the user with relevant information, allowing for the acquisition of relevant information. For example, the current user's identification information can be associated with the user's historical seat adjustment records, driving habits, and other information.

[0073] For example, the facial image of the current user may be captured by a camera in the vehicle, and then the identification information of the current user may be obtained based on the facial image of the current user.

[0074] S202. Obtain initial seat adjustment parameters based on the target body shape feature vector of the current user, and obtain seat adjustment parameters of the current user based on the identification information of the current user.

[0075] For example, the seat adjustment parameters may represent parameters such as the seat back angle (e.g., 108°), the distance between the seat and the pedal (e.g., 50 cm), the seat height (e.g., 30 cm), and the seat back lumbar support strength (e.g., level 3). It is understood that the seat back angle is related to the current user's height, the distance between the seat and the pedal and the seat height are related to the current user's leg length, and the seat back lumbar support strength is related to the current user's weight.

[0076] For example, the seat adjustment parameter can be used to represent the adjustment amount of the seat adjustment parameter. Taking the seat back angle as an example, if the current seat adjustment parameter is 108° and the seat adjustment parameter is +2°, the current seat adjustment parameter is adjusted based on the seat adjustment parameter, resulting in an adjusted current seat adjustment parameter of 110°.

[0077] For example, the initial seat adjustment parameters can be used to represent the seat adjustment parameters that are matched to the user corresponding to the target body shape feature vector when sitting in the seat. For example, body shape feature vector A corresponds to seat adjustment parameter 1, body shape feature vector B corresponds to seat adjustment parameter 2, and body shape feature vector C corresponds to seat adjustment parameter 3. If the current user's target body shape feature vector is body shape feature vector B, the initial seat adjustment parameters are determined to be seat adjustment parameter 2.

[0078] As an example, obtaining initial seat adjustment parameters based on the target body shape feature vector includes: inputting the target body shape feature vector into a body shape clustering model to obtain the initial seat adjustment parameters.

[0079] The body shape clustering model is used to obtain the seat adjustment parameters corresponding to the body shape feature vector based on the input body shape feature vector.

[0080] For example, the body shape clustering model can identify groups of body shape features similar to the user's body shape feature vector, and then determine seat adjustment parameters corresponding to these groups of body shape features. The body shape clustering model can perform dimensionality reduction and clustering on the input body shape feature vector to determine seat adjustment parameters that match the body shape feature vector. The seat adjustment parameters that match the body shape feature vector can be stored in a database corresponding to the body shape clustering model.

[0081] In practical applications, the body shape clustering model can be a density-based spatial clustering algorithm (DBSCAN). The DBSCAN model forms clusters by connecting high-density areas and can identify noise points (outliers). In body shape clustering, each data point represents a user's body shape feature vector.

[0082] For example, the body shape clustering model can reduce the dimensionality of the input 128-dimensional body shape feature vector, compressing it to 8-dimensional principal components to accelerate calculations and remove redundancy. The reduced body shape feature vector is then clustered to obtain the user's cluster label. The user's cluster label is then used to determine the seat adjustment parameters corresponding to the cluster label from historical data within the cluster, and these seat adjustment parameters are determined as the initial seat adjustment parameters. For example, if the cluster label is a standard body shape, the seat adjustment parameters corresponding to the standard body shape may include a backrest angle of 105°, a distance of 70 cm between the seat and the pedals, a seat height (e.g., 40 cm), and a seat back lumbar support strength (e.g., level 2).

[0083] The above-mentioned cluster historical data includes multiple cluster labels, and different cluster labels correspond to different seat adjustment parameters. For example, as shown in Table 1, Table 1 is a correspondence table between cluster labels and seat adjustment parameters provided in an embodiment of the present application. It should be noted that Table 1 only provides an exemplary description of the correspondence between cluster labels and seat adjustment parameters, and the specific correspondence can be determined by actual scenarios.

[0084] Table 1 Correspondence between cluster labels and seat adjustment parameters

[0085]

[0086] For example: as shown in Table 1, if the cluster label is cluster label 1 (standard body type), the seat adjustment parameters include: seat back angle: 105°, distance between seat and pedal: 70cm, seat height: 40cm, lumbar support strength: 2 levels; if the cluster label is cluster label 2 (tall and long-legged type), the seat adjustment parameters include: seat back angle: 112°, distance between seat and pedal: 76cm, seat height: 50cm, lumbar support strength: 3 levels; if the cluster label is cluster label 3 (wide and strong type), the seat adjustment parameters include: seat back angle: 100°, distance between seat and pedal: 68cm, seat height: 40cm, lumbar support strength: 4 levels; if the cluster label is cluster label 4 (petite and flexible type), the seat adjustment parameters include: seat back angle: 108°, distance between seat and pedal: 62cm, seat height: 36cm, lumbar support strength: 2 levels.

[0087] It can be understood that since the body shape clustering model can automatically identify body shape feature groups that are similar to the user's body shape feature vector, and then determine the seat adjustment parameters corresponding to the body shape feature group, the initial seat adjustment parameters are obtained by inputting the target body shape feature vector into the body shape clustering model. In this way, the seat adjustment parameters suitable for the current user's body shape can be obtained quickly and accurately, thereby improving the intelligence of the seat adjustment.

[0088] As a possible implementation method, the method provided in an embodiment of the present application also includes: if the target body shape feature vector is input into the body shape clustering model and the seat adjustment parameters corresponding to the target body shape feature vector are not obtained, then detecting the current user's manual adjustment of the target seat; based on the manually adjusted seat adjustment parameters, determining the initial seat adjustment parameters.

[0089] For example, if no cluster label matches the current user's body shape feature vector in the historical data within the cluster, this indicates that the body shape clustering model cannot match the current user's body shape feature vector to any existing cluster, meaning that the body shape feature vector is an outlier. In this case, the body shape feature vector can be used as a new cluster label; the current user can manually adjust the target seat; and the seat adjustment parameters obtained after the manual adjustment are obtained and determined as the seat adjustment parameters corresponding to the new cluster label.

[0090] It needs to be explained that when the initial seat adjustment parameters corresponding to the target body shape feature vector are obtained, the seat adjustment parameters for the current user can be obtained based on the current user's own situation (for example, the current user's sitting habits, historical feedback, etc.), and then the initial seat adjustment parameters can be adjusted based on the seat adjustment parameters to obtain personalized seat adjustment parameters for the current user.

[0091] For example, the current user's seat adjustment parameters can be obtained using their identification information. For example, a historical database may store seat adjustment parameters for multiple users, each with unique identification information. When the current user enters the vehicle, their identity (user identification information) can be confirmed through facial recognition or account login, and then the current user's seat adjustment parameters can be retrieved from the historical database.

[0092] Exemplarily, identification information of the current user may be input into the target reinforcement learning model so that the target reinforcement learning model outputs seat adjustment parameters of the current user.

[0093] S203 . Adjust the initial seat adjustment parameters based on the seat adjustment parameters to obtain the first seat adjustment parameters of the current user.

[0094] For example, the seat adjustment parameter may be an adjustment amount for the initial seat adjustment parameter. The seat adjustment parameter may be a positive value, a negative value, or zero, depending on the actual situation of the current user.

[0095] As an example, take the seat adjustment parameter as the seat back angle: if the initial seat adjustment parameter is 105° and the current user's seat adjustment parameter is +4°, then the current user's first seat adjustment parameter is 109°.

[0096] As another example, take the seat adjustment parameter as the distance between the seat and the pedal: if the initial seat adjustment parameter is 70 cm and the current user's seat adjustment parameter is 0, then the current user's first seat adjustment parameter is 70 cm.

[0097] As another example, take the seat adjustment parameter as seat height: if the initial seat adjustment parameter is 40 cm and the current user's seat adjustment parameter is -3 cm, then the current user's first seat adjustment parameter is 37 cm.

[0098] As another example, take the seat adjustment parameter as lumbar support strength: if the initial seat adjustment parameter is level 2 and the current user's seat adjustment parameter is level 1, then the current user's first seat adjustment parameter is level 3.

[0099] S204 . Adjust the target seat based on the first seat adjustment parameter.

[0100] For example, if the first seat adjustment parameter includes a seat back angle, the back angle of the target seat can be adjusted. If the first seat adjustment parameter includes the distance between the seat and the pedal, the position of the target seat can be adjusted. If the first seat adjustment parameter includes seat height, the height of the target seat can be adjusted. If the first seat adjustment parameter includes lumbar support strength, the lumbar inflation device of the target seat can be adjusted.

[0101] In an embodiment of the present application, the target body shape feature vector of the current user and the identification information of the current user are obtained; based on the target body shape feature vector, initial seat adjustment parameters are obtained, and based on the identification information, seat adjustment parameters of the current user are obtained; based on the seat adjustment parameters, the initial seat adjustment parameters are adjusted to obtain the first seat adjustment parameters of the current user; based on the first seat adjustment parameters, the target seat is adjusted; since the initial seat adjustment parameters in this solution are obtained based on the target body shape feature vector, the seat adjustment parameters are obtained based on the identification information of the current user, and the seat adjustment parameters corresponding to the current user are obtained based on the initial seat adjustment parameters and the seat adjustment parameters. That is, when determining the seat adjustment parameters of the current user, both the body shape characteristics of the current user and the personalized needs of the current user are taken into consideration, and thus personalized seat adjustment parameters for the current user can be provided based on both the body shape characteristics and the personalized needs of the current user, thereby improving the intelligence of the seat adjustment to enhance the user's driving experience.

[0102] In a possible embodiment of the present application, after the above S204, the method provided in the embodiment of the present application also includes: determining whether a feedback operation is detected; if a feedback operation is detected, optimizing the seat adjustment parameters of the current user based on the adjustment parameters corresponding to the feedback operation; and using the optimized seat adjustment parameters as the seat adjustment parameters of the current user.

[0103] The feedback operation is the adjustment operation of the target seat by the current user.

[0104] For example, the feedback operation may refer to a micro-adjustment operation (e.g., adjusting the seat back angle) performed by the current user on the target seat's seat adjustment parameters after the target seat has been adjusted to the first seat adjustment parameters. Alternatively, the feedback operation may also be a persistent sign of discomfort experienced by the current user (e.g., uneven seat cushion pressure distribution).

[0105] For example, if it is detected that the current user has made a fine-tuning operation on the seat adjustment parameters of the target seat, for example, the current user adjusts the seat back angle from 109° (the initial seat adjustment parameter is 105°, and the seat adjustment parameter is 4°) to 115°, it can be determined that the current user has increased the seat back angle by 6° on the basis of 109°. At this time, the seat adjustment parameter of 4° can be optimized based on the increased seat back angle of 6°, and the optimized seat adjustment parameter can be used as the seat adjustment parameter of the current user.

[0106] For example, the aforementioned optimization of the current user's seat adjustment parameters can update the current user's seat adjustment parameters to 10°, so that the next time the current user sits on the target seat, the initial seat adjustment parameters (105°) can be adjusted to the first seat adjustment parameters (115°).

[0107] For another example, the aforementioned optimization of the current user's seat adjustment parameters can also be to update the current user's seat adjustment parameters to the average value between the seat adjustment parameter of 4° and the increased seat back angle of 6°, that is, 5°; in this way, the next time the current user sits in the target seat, the initial seat adjustment parameter (105°) can be adjusted to the first seat adjustment parameter (110°).

[0108] Of course, the seat adjustment parameters of the current user may be optimized in other ways, such as optimizing (i.e., training) the seat adjustment parameters of the current user by combining the historical feedback operations of the current user with the above-mentioned feedback operations of the current user. For details, please refer to the following embodiments, which will not be repeated here.

[0109] In an embodiment of the present application, if a feedback operation is detected, the seat adjustment parameters for the current user can be optimized. Because the personalized adjustment parameters of different users may vary, in this solution, when a change in a user's personalized adjustment parameters is detected, the user's feedback operation can be used to update the user's personalized adjustment parameters, thereby improving the adaptability of the seat adjustment parameters to the current user and thereby enhancing the intelligence of the seat adjustment.

[0110] In a possible embodiment of the present application, the method provided in the embodiment of the present application also includes: taking the initial seat adjustment parameters as the initial parameters, adjusting the initial parameters with a first adjustment amount to obtain the target parameters; when adjusting the target seat with the target parameters, determining the reward value of the first adjustment amount based on the user's feedback information on the target seat and the first adjustment amount; training the target reinforcement learning model based on the reward value of each adjustment amount in the seat adjustment amount set to obtain the trained target reinforcement learning model.

[0111] The first adjustment amount is any adjustment amount in the seat adjustment amount set, and the trained target reinforcement learning model is used to output the seat adjustment parameters of the user, including the current user.

[0112] For example, the seat adjustment amount set may include multiple adjustment amounts, at least two of which may be the same. The adjustment amounts may be positive or negative. For example, taking the seat back angle as an example, the multiple adjustment amounts may include: 2°, 3°, 3°, 3°, 5°, -1°, and -1°. In other words, the first adjustment amount may be any one of the multiple adjustment amounts.

[0113] Exemplarily, the target parameter may be the sum of the initial parameter and the first adjustment amount. For example, if the initial parameter is a seat back angle of 109°, and the multiple adjustment amounts include: 2°, 3°, 3°, 3°, 5°, -1°, and -1°, then the target parameters may include: 111°, 112°, 114°, and 108°.

[0114] For example, the user's feedback on the target seat may include the user's manual fine-tuning operation, whether the user's sitting posture is normal, etc. For example, if the initial parameter is a seat back angle of 109°, the first adjustment amount is 2°, and the target parameter is 111°; if the seat back angle of the target seat is adjusted to the target parameter of 111° and the user adjusts the seat back angle back to 109° within a preset time period, it indicates that the first adjustment amount may not be appropriate.

[0115] Exemplarily, the reward value of the first adjustment amount can be greater than zero (i.e., a positive reward value) or less than zero (i.e., a negative reward value). For example, a positive reward value (e.g., +1 point to +5 points) can mean that when the target seat is adjusted by the target parameters, the user does not make subsequent adjustments. In this case, it is indicated that the first adjustment amount meets the user's riding needs. Therefore, a positive reward is given to the first adjustment amount to encourage the user to continue to repeat the execution action of the first adjustment amount. The larger the positive reward value, the more the reward value meets the user's riding needs. For another example, a negative reward value (e.g., -1 point to -5 points) can mean that when the target seat is adjusted by the target parameters, the user adjusts the target seat more frequently, or has an abnormal posture, etc. In this case, it is indicated that the first adjustment amount does not meet the user's riding needs. Therefore, a negative reward is given to the first adjustment amount to prevent the user from continuing to repeat the execution action of the first adjustment amount. The larger the absolute value of the negative reward value, the less the reward value meets the user's riding needs.

[0116] For example, a target reinforcement learning model can be used to train seat adjustment parameters based on current user feedback, enabling the trained target reinforcement learning model to improve the accuracy and intelligence of personalized seat adjustment parameters for users. In practical applications, the target reinforcement learning model can be a Deep Q-Network (DQN) reinforcement learning model. The DQN reinforcement learning model combines deep learning and reinforcement learning algorithms (Q-Learning).

[0117] For example: the reward value of each adjustment amount in the seat adjustment amount set can be associated with the reward value in the DQN reinforcement learning model; the target parameter corresponding to each adjustment amount can be associated with the target state in the DQN reinforcement learning model; each adjustment amount can be associated with the execution action in the DQN reinforcement learning model; it should be explained that since the DQN model mainly includes three main elements, namely reward value, state, and action, the DQN reinforcement learning model can be trained through the reward value in the DQN reinforcement learning model associated with the reward value of each adjustment amount, the target state associated with the target parameter, and the execution action associated with the adjustment amount, so that the trained DQN reinforcement learning model can determine the seat adjustment amount of the current user.

[0118] In an embodiment of the present application, the target reinforcement learning model is trained to obtain a trained target reinforcement learning model. Since the target reinforcement learning model itself can be trained through the adjustment amounts in the seat adjustment amount set, the seat adjustment amount can be accurately output through the trained target reinforcement learning model when the current user subsequently sits on the target seat, thereby further improving the intelligence of the seat adjustment.

[0119] In a possible embodiment of the present application, after the above S204, the method provided in the embodiment of the present application also includes: detecting the current state of the current user; obtaining a second seat adjustment parameter of the target seat based on the current state; and adjusting the target seat based on the second seat adjustment parameter.

[0120] For example, during driving, the current state of the current user can be detected in real time. Based on the current state of the current user, it can be determined whether the current user is fatigued, and the target seat can be adjusted in a timely manner to improve the current user's driving experience and driving safety. The current state can be fatigued, awake, etc.

[0121] For example, the current state of the current user can be detected by the duration of the current user's abnormal sitting posture, or by the current user's physiological state. The embodiments of the present application do not impose specific restrictions on this.

[0122] For example, the second seat adjustment parameter may be used to further adjust the target seat after the target seat is adjusted based on the first seat adjustment parameter.

[0123] It can be understood that based on the current state of the current user, the second seat adjustment parameter is determined, and then the target seat is adjusted based on the second seat adjustment parameter; since the current state can reflect whether the current user is in a fatigue state, the target seat suitable for the current user's sitting posture can be determined through the current state, thereby reducing the incompatibility of seat adjustment parameters caused by user fatigue, and intelligently adjusting the target seat to improve the user's driving experience.

[0124] As an example, based on the current state, obtaining the second seat adjustment parameter of the target seat includes: determining whether the current state is a fatigue state; if the current state is a fatigue state, obtaining the second seat adjustment parameter based on the current posture of the current user.

[0125] It should be understood that when the current user is in a fatigued state, the current user's sitting posture may be abnormal, such as excessive forward leaning for a long time, abnormal spinal curvature, abnormal difference in left and right ischial pressure, etc.; at this time, the second seat adjustment parameter can be obtained based on the current user's current posture, and then the target seat can be further adjusted to reduce the current user's fatigue and thereby improve driving safety.

[0126] It can be understood that when the current state is a fatigue state, the second seat adjustment parameter is obtained based on the current posture of the current user; since the current posture is related to the user's fatigue state, further adjustment of the target seat based on the second adjustment parameter obtained based on the current posture can reduce the current user's fatigue and thereby improve driving safety.

[0127] As an example, determining whether the current state is a fatigue state includes: if the vehicle's driving time is greater than or equal to a preset time, detecting whether the current user's posture is an abnormal posture; if the current user's posture is an abnormal posture, determining that the current state is a fatigue state; if the current user's posture is a posture other than an abnormal posture, determining that the current state is not a fatigue state.

[0128] For example, an abnormal posture may reflect that the posture of the current user sitting in the target seat is significantly different from the normal sitting posture of the current user. For example, an abnormal posture includes one or more of an abnormal spinal flexion angle, an abnormal thigh-torso angle, a large forward neck tilt, a large left and right ischium pressure difference, and an abnormal lumbar support contact area. Among them, an abnormal spinal flexion angle may refer to a spinal flexion angle less than 12° or greater than 18°; an abnormal thigh-torso angle may refer to a thigh-torso angle less than 100° or greater than 120°; a large forward neck tilt may refer to a forward neck tilt greater than 6cm; a large left and right ischium pressure difference may refer to a left and right ischium pressure difference ratio greater than 25%; and an abnormal lumbar support contact area may refer to a lumbar support contact area less than 100cm2 or greater than 200cm2.

[0129] For example, a body shape detection model (eg, a Transformer-Pose model) may be used to detect whether the current user's posture is an abnormal sitting posture.

[0130] For example, the preset duration may be a value pre-configured by the vehicle or a value manually set, and this embodiment of the present application does not specifically limit this. For example, the preset duration may be 1 hour (h).

[0131] For example, after the vehicle has been traveling for 1 hour, if it is detected that the current user's spinal curvature angle is abnormal and the neck tilts forward a large distance, it can be determined that the current user's posture is abnormal at this time; therefore, the target seat can be adjusted through the second seat adjustment parameter to reduce user fatigue.

[0132] It can be understood that when the vehicle's driving time is greater than or equal to the preset time and the current user's posture is abnormal, the current state is determined to be a fatigue state; since the fatigue state is associated with the driving time and the user's posture, it is possible to quickly and accurately detect that the user is in a fatigue state at this time, and then adjust the target seat when the user is in a fatigue state, thereby improving the intelligence of the target seat adjustment.

[0133] As an example, obtaining the second seat adjustment parameter based on the current posture of the current user includes: obtaining the second seat adjustment parameter based on the identification information of the current user and the current posture of the current user.

[0134] For example, since the abnormal posture includes one or more abnormal points among abnormal spinal curvature angle, abnormal angle between thigh and trunk, large forward neck tilt, large pressure difference between left and right ischial tubes, and abnormal lumbar support contact area, the abnormal posture may include multiple abnormal postures, and different abnormal postures may include different abnormal points. For example, abnormal posture A may include abnormal point 1 (abnormal spinal curvature angle) and abnormal point 2 (abnormal angle between thigh and trunk); abnormal posture B may include abnormal point 1 (abnormal spinal curvature angle) and abnormal point 3 (large pressure difference between left and right ischial tubes).

[0135] For example, the degree of abnormal points in the same abnormal posture may vary for different users. For example, in abnormal posture A of user a, abnormal point 1 corresponds to a spinal curvature angle of 20°, and abnormal point 2 corresponds to a thigh-torso angle of 90°. In abnormal posture A of user b, abnormal point 1 corresponds to a spinal curvature angle of 25°, and abnormal point 2 corresponds to a thigh-torso angle of 125°.

[0136] For example, the target reinforcement learning model can be used to input the current user's identification information and current posture into the target reinforcement learning model, so that the target reinforcement learning model can determine a second seat adjustment parameter that matches the current user's current sitting posture (abnormal sitting posture). For example, if the spinal curvature angle is abnormal, the seat back angle and lumbar support strength are adjusted; if the thigh-torso angle is abnormal, the seat height is increased; if the neck tilt distance is large, the seat height and lumbar support strength are adjusted; if the left and right ischial pressure differences are large, the target seat's air cushion inflation value is adjusted; if the lumbar support contact area is abnormal, the lumbar support strength is adjusted.

[0137] It can be understood that through the current user's identification information and current posture, the personalized second seat adjustment parameters corresponding to the current user's current posture can be accurately obtained, so that the target seat can be intelligently adjusted based on the second seat adjustment parameters to improve the user's driving experience.

[0138] In an embodiment of the present application, the second seat adjustment parameter is determined based on the current state of the current user, and then the target seat is adjusted based on the second seat adjustment parameter. In this way, the target seat can be intelligently adjusted according to the current state of the user, thereby improving the user's driving experience.

[0139] Figure 3 It is a schematic flow chart of another seat adjustment method provided in an embodiment of the present application.

[0140] For example, Figure 3The executor of the seat adjustment method shown can be a vehicle, a vehicle-side controller, or a seat adjustment device (for example, a chip) used in a vehicle; wherein the vehicle-side controller can be: a vehicle controller, a cockpit domain controller or a central domain controller.

[0141] like Figure 3 As shown, the seat adjustment method includes S301 to S309, which are described in detail below.

[0142] S301. Obtain a color image, a depth image, a pressure value of a pressure sensor in the target seat, and identification information of the current user on the target seat.

[0143] For example, the color image can be a two-dimensional image (e.g., a red, green, and blue image). The color image can be used to describe the current user's full body image or a partial sitting posture image. Through the color image, the current user's joints (e.g., shoulders, elbows, knees, etc.), body contours (e.g., shoulder width, sitting height), and whether the sitting posture is offset (e.g., when the body leans to the left, the right shoulder joint moves downward).

[0144] The color image can be obtained by a camera deployed inside the cabin. The camera can be an RGB camera or an infrared camera, etc., and this embodiment of the application does not impose specific restrictions on this.

[0145] For example, the depth image can be 3D point cloud data. The depth image can be used to reflect information such as the user's bone length and 3D posture. It can also further address issues such as the inability to obtain information such as the user's joints and body contours due to partial occlusion of the color image (for example, a steering wheel obscuring the torso).

[0146] The depth image can be obtained by a depth sensor, which can be a time-of-flight sensor, a laser radar, etc., and the present embodiment does not impose any specific restrictions on this.

[0147] For example, the pressure value of the pressure sensor in the target seat can be used to reflect the distribution of the current user's pressure on the target seat, such as the pressure areas of the buttocks and back.

[0148] For example, identification information may be information that uniquely identifies a user, such as a user's personal identification number. Identification information can be used to associate the user with relevant information, allowing for the acquisition of relevant information. For example, the current user's identification information can be associated with information such as the user's historical seat adjustment history and driving habits.

[0149] S302. Input the color image, depth image and pressure value into the body shape detection model to obtain the target body shape feature vector.

[0150] Exemplarily, the body shape detection model can be a multimodal neural network model that processes the color image, depth image, and pressure values from the pressure sensors in the target seat, inputting the body shape detection model to output a target body shape feature vector. Optionally, the body shape detection model can also output the coordinates of key points of the current user's target posture.

[0151] For example, the body shape feature vector can be used to reflect the user's body structure, sitting posture habits, and physiological characteristics. For example, the target body shape feature vector can include the current user's body length ratio, sitting posture angle, and pressure distribution.

[0152] S303. Input the target body shape feature vector into the body shape clustering model to obtain initial seat adjustment parameters.

[0153] Exemplarily, the body shape clustering model is used to obtain seat adjustment parameters corresponding to the body shape feature vector based on the input body shape feature vector.

[0154] For example, the seat adjustment parameters may represent parameters such as the seat back angle (e.g., 108°), the distance between the seat and the pedal (e.g., 50 cm), the seat height (e.g., 30 cm), and the seat back lumbar support strength (e.g., level 3). It is understood that the seat back angle is related to the current user's height, the distance between the seat and the pedal and the seat height are related to the current user's leg length, and the seat back lumbar support strength is related to the current user's weight.

[0155] S304: Input the identification information of the current user into the target reinforcement learning model to obtain seat adjustment parameters.

[0156] For example, the current user's seat adjustment parameters can be obtained using their identification information. For example, a historical database may store seat adjustment parameters for multiple users, each with unique identification information. When the current user enters the vehicle, their identity (user identification information) can be confirmed through facial recognition or account login, and then the current user's seat adjustment parameters can be retrieved from the historical database.

[0157] For example, the target reinforcement learning model can be used to train seat adjustment parameters based on the current user's feedback, so that the trained target reinforcement learning model can improve the accuracy and intelligence of personalized seat adjustment parameters for the user when sitting in the seat. In actual application scenarios, the target reinforcement learning model can be a deep Q-network reinforcement learning model.

[0158] S305 . Obtain a first seat adjustment parameter based on the initial seat adjustment parameter and the seat adjustment parameter.

[0159] Optionally, the specific implementation of the above S305 can be found in Figure 2 The relevant description of S203 is omitted here.

[0160] S306 . Adjust the target seat based on the first seat adjustment parameter.

[0161] Optionally, the specific implementation of the above S306 can be found in Figure 2 The relevant description of S204 is omitted here.

[0162] S307. During the driving of the vehicle, detect whether the current state of the current user is a fatigue state; if so, execute S308.

[0163] In a possible implementation, if not (the current state of the current user is not a fatigue state), S307 is executed.

[0164] For example, during driving, the current state of the current user can be detected in real time. Based on the current state of the current user, it can be determined whether the current user is fatigued, and the target seat can be adjusted in a timely manner to improve the current user's driving experience and driving safety. The current state can be fatigued, awake, etc.

[0165] S308. Input the identification information and current posture into the target reinforcement learning model to obtain the second seat adjustment parameter.

[0166] For example, the target reinforcement learning model can be used to input the current user's identification information and current posture into the target reinforcement learning model, so that the target reinforcement learning model can determine a second seat adjustment parameter that matches the current user's current sitting posture (abnormal sitting posture). For example, if the spinal curvature angle is abnormal, the seat back angle and lumbar support strength are adjusted; if the thigh-torso angle is abnormal, the seat height is increased; if the neck tilt distance is large, the seat height and lumbar support strength are adjusted; if the left and right ischial pressure differences are large, the target seat's air cushion inflation value is adjusted; if the lumbar support contact area is abnormal, the lumbar support strength is adjusted.

[0167] S309 . Adjust the target seat based on the second seat adjustment parameter.

[0168] For detailed description of S301 to S309, please refer to the above Figures 1 to 2 ,as well as Figures 1 to 2 The relevant description will not be repeated here.

[0169] In an embodiment of the present application, a color image, a depth image, a pressure value of a pressure sensor in the target seat, and identification information of the current user on the target seat are obtained; by inputting the color image, depth image, and pressure value into a body shape detection model, a target body shape feature vector can be quickly and accurately obtained; the target body shape feature vector is then input into a body shape clustering model to quickly and accurately obtain initial seat adjustment parameters, thereby improving the accuracy and intelligence of seat adjustment; then, the identification information of the current user is input into a target reinforcement learning model to obtain seat adjustment parameters, and then obtain first seat adjustment parameters; based on the first seat adjustment parameters, the target seat is adjusted; since the initial seat adjustment parameters in this solution are obtained based on the target body shape feature vector and the seat adjustment parameters are obtained based on the identification information of the current user, personalized seat adjustment parameters for the current user can be provided based on the body shape characteristics and personal situation of the current user; and, based on the current posture of the current user, second seat adjustment parameters are obtained, and then the target seat is further adjusted to reduce the fatigue of the current user; based on this, the present solution can improve the intelligence of seat adjustment to improve the user's driving experience.

[0170] Figure 4 It is a schematic flow chart of another seat adjustment method provided in an embodiment of the present application.

[0171] For example, Figure 4 The executor of the seat adjustment method shown can be a vehicle, a vehicle-side controller, or a seat adjustment device (for example, a chip) used in a vehicle; wherein the vehicle-side controller can be: a vehicle controller, a cockpit domain controller or a central domain controller.

[0172] like Figure 4 As shown, the seat adjustment method includes S401 to S409, which are described in detail below.

[0173] For example, S401 to S409 can be divided into three stages: passenger size detection, seat adjustment, and target reinforcement learning model optimization. The passenger size detection stage includes S401 to S405, the seat adjustment stage includes S406 to S407, and the target reinforcement learning model optimization stage includes S408 to S409. S401 to S409 are described in detail below.

[0174] Phase 1: Passenger body shape detection phase:

[0175] S401. Detect passengers entering the vehicle.

[0176] For example, a camera installed in the vehicle cabin may be used to detect when a passenger enters the vehicle.

[0177] Optionally, after the passenger enters the vehicle, it is possible to detect that the passenger is sitting on the target seat through a pressure sensor installed on the target seat.

[0178] S402. Collect passenger data through sensors.

[0179] For example, the sensors may include a camera deployed inside the cabin (e.g., an RGB camera, an IR camera, etc.), a depth sensor (e.g., a time-of-flight sensor, a lidar, etc.), and a pressure sensor in the target seat. Passenger data may include a color image, a depth image, and pressure values. Specifically, the color image may be obtained by the camera, the depth image by the depth sensor, and the pressure value by the pressure sensor.

[0180] For example, the camera can capture the driver's facial and body images, and the depth sensor can obtain the passenger's three-dimensional depth image to avoid the problem of unrecognition caused by partial occlusion in the passenger's facial and body images. The pressure sensor in the target seat can record the contact pressure distribution of the passenger after taking the seat, further improving the accuracy of inferring body characteristics. Optionally, the passenger data can also include driving behavior data (such as historical seat adjustment records, driving style, long-term driving posture changes, etc.) and biometric information (such as historical seat adjustment records, driving style, long-term driving posture changes, etc.).

[0181] S403. Input the passenger data into the body shape detection model so that the body shape detection model outputs a body shape feature vector.

[0182] In practical applications, the body shape detection model can be a Transformer-Pose model. It is understood that compared to traditional neural network models, the Transformer-Pose model can improve its adaptability to occlusion and complex poses when extracting body shape feature vectors.

[0183] Optionally, before inputting the passenger data into the body shape detection model, the passenger's facial image and body image may be subjected to image enhancement processing (e.g., random rotation, scaling, brightness adjustment, etc.) to improve the accuracy of the body shape detection model in detecting body shape feature vectors.

[0184] Optionally, the body shape detection model may also output the passenger's posture key points; the passenger's posture key points may then be subjected to coordinate normalization processing, so that the body shape detection model can determine the body shape feature vector based on the normalized posture key points.

[0185] Optionally, before inputting the passenger data into the body shape detection model, the passenger's depth image may be subjected to denoising to filter out noise points. Specifically, edge information may be enhanced through bilateral filtering to improve the efficiency of filtering out noise points.

[0186] Optionally, before inputting the passenger data into the body shape detection model, the pressure value may be converted into a low-dimensional feature vector to facilitate the body shape detection model to perform recognition processing on the pressure value.

[0187] For the detailed description of S401 to S403 , please refer to the relevant description of S201 above, which will not be repeated here.

[0188] S404. Input the body shape feature vector into the body shape clustering model so that the body shape clustering model performs cluster matching.

[0189] In actual application scenarios, the body shape clustering model may be a DBSCAN body shape clustering model.

[0190] Exemplarily, the clustering process of the DBSCAN body shape clustering model may include dimensionality reduction, clustering, outlier detection, etc. Among them, dimensionality reduction can be used to reduce the dimension of the body shape feature vector to improve the computational efficiency of the subsequent clustering process. Clustering can be used to calculate the similarity between the body shape feature vector and the historical body shape categories stored in the DBSCAN body shape clustering model to determine the body shape category of the passenger, and then obtain the initial seat adjustment parameters based on the body shape category. Outlier detection can be used to trigger the target reinforcement learning model below to determine the initial seat adjustment parameters corresponding to the body shape feature vector if no initial seat adjustment parameters matching the body shape category corresponding to the body shape feature vector are found in the historical body shape categories.

[0191] S405. Output initial seat adjustment parameters through the body shape clustering model.

[0192] Illustratively, the body shape clustering model can determine the cluster label that matches the passenger's body shape feature vector by calculating the similarity between the passenger and the historical data stored in the body shape clustering model, and further determine the initial seat adjustment parameters corresponding to the cluster label.

[0193] For example, if it is detected that the passenger's body shape feature vector cannot find matching data in the historical data stored in the body shape clustering model (i.e., the data is an outlier), the target reinforcement learning model can be used to determine the initial seat adjustment parameters corresponding to the passenger's body shape feature vector; and the body shape feature vector and the initial seat adjustment parameters corresponding to the body shape feature vector are stored in the body shape clustering model.

[0194] For example, the body type clustering model can select the body type category (i.e., cluster label) closest to the cluster result corresponding to the body type feature vector based on the cluster result, and provide initial seat adjustment parameters based on the closest body type category, such as seat back angle, distance between seat and gear position, lumbar support strength, etc.

[0195] For the detailed description of S404 to S405 , please refer to the relevant description of S202 above, which will not be repeated here.

[0196] Stage 2: Seat adjustment stage:

[0197] S406. Obtain seat adjustment parameters through the target reinforcement learning model and the initial seat adjustment parameters.

[0198] In actual application scenarios, the target reinforcement learning model can be a DQN reinforcement learning model.

[0199] For example, the passenger's identification information can be input into the target reinforcement learning model so that the target reinforcement learning model outputs the passenger's seat adjustment parameters, and then based on the seat adjustment parameters and the initial seat adjustment parameters, the seat adjustment parameters (i.e., the first seat adjustment parameters mentioned above) are obtained.

[0200] For a detailed description of S406 , please refer to the above-mentioned descriptions of S202 and S203 , which will not be repeated here.

[0201] S407. Adjust the backrest angle, height, and position of the target seat based on the seat adjustment parameters.

[0202] For a detailed description of S407 , please refer to the above-mentioned description of S204 , which will not be repeated here.

[0203] Phase 3: Optimization of the target reinforcement learning model:

[0204] S408. If a passenger's feedback operation is detected, data corresponding to the feedback operation is obtained.

[0205] Exemplarily, the feedback operation is an adjustment operation of the target seat by the passenger. For example, the feedback operation may refer to a fine adjustment operation of the seat adjustment parameter of the target seat by the passenger (e.g., adjustment of the seat back angle) after the target seat is adjusted to the seat adjustment parameter.

[0206] S409. Optimize the target reinforcement learning model based on the data corresponding to the feedback operation.

[0207] For the description of S408 to S409, please refer to the above description of the feedback operation, which will not be repeated here.

[0208] As a possible embodiment, the method provided in the embodiment of the present application also includes a training stage of the target reinforcement learning model. This stage can be after stage three, before stage one, or between the other two stages. The embodiment of the present application does not impose any specific restrictions on this.

[0209] For example, the initial seat adjustment parameters can be used as initial parameters, and the initial parameters can be adjusted using any adjustment amount in the seat adjustment set to obtain target parameters. When the target seat is adjusted using the target parameters, a reward value for any adjustment amount in the seat adjustment set is determined based on passenger feedback on the target seat. Based on the reward value for each adjustment amount in the seat adjustment set, a target reinforcement learning model is trained to obtain a trained target reinforcement learning model. This trained target reinforcement learning model can be the target reinforcement learning model described in S406 above.

[0210] Exemplarily, during the training process of a target reinforcement learning model, the target reinforcement learning model can be trained using a state space (State), an action space (Action), and a reward function (Reward). The state space can include the passenger's body type, the initial backrest angle of the target seat, the initial distance between the target seat and the gear position, and the passenger's fine-tuning feedback (e.g., the number of fine-tunings); optionally, the state space can also include the passenger's physiological signals (e.g., heart rate, fatigue monitoring, etc.). The action space can include adjustment parameters for the target seat, such as adjusting the backrest of the target seat based on the initial backrest angle; adjusting the position of the target seat based on the initial distance between the target seat and the gear position, etc. The reward function can include positive reward values (high reward values) and negative reward values (low reward values). If the passenger adjusts the target seat infrequently, or if the passenger has not adjusted the target seat for an extended period of time, this can indicate that the initial seat adjustment parameters are accurate and can be assigned a positive reward value. If the passenger frequently adjusts the target seat length, this can indicate that the initial seat adjustment parameters are inappropriate and need to be optimized, and can be assigned a negative reward value.

[0211] For example: A specific scenario is used to illustrate the information interaction between multiple models. Among them, multiple models may include a multimodal perception model, a Transformer-Pose model, a DBSCAN body shape clustering model, and a DQN reinforcement learning model. Among them, the multimodal perception model can be connected to multiple sensors in the vehicle (for example, cameras, depth sensors, pressure sensors), and then collect color images, depth images, pressure data and other data of the passenger; the multimodal perception model inputs the passenger's color image, depth image, pressure data and other data into the Transformer-Pose model, and then extracts the passenger's posture key points and body shape feature vector, and inputs the body shape feature vector into the DBSCAN body shape clustering model, so that the DBSCAN body shape clustering model can identify the passenger's body shape category and match the initial seat adjustment parameters corresponding to the body shape category, and output the passenger's seat adjustment parameters through the DQN reinforcement learning model, and then adjust the passenger's seat based on the initial seat adjustment parameters and the passenger's seat adjustment parameters. After adjusting the seat occupied by the passenger, it is also possible to detect whether the passenger provides feedback on the seat; if the passenger provides feedback on the seat, the seat adjustment parameters output by the DQN reinforcement learning model can be optimized based on the data corresponding to the feedback operation.

[0212] In an embodiment of the present application, when a passenger enters the vehicle, the passenger's data is collected and input into a body shape detection model to obtain a body shape feature vector; and the body shape feature vector is input into a body shape clustering model to obtain initial seat adjustment parameters; seat adjustment parameters are obtained through a target reinforcement learning model and the initial seat adjustment parameters, so as to adjust the backrest angle, height and position of the target seat based on the seat adjustment parameters; therefore, the present solution can provide personalized seat adjustment parameters for the passenger according to the passenger's body shape characteristics, thereby improving the intelligence of the seat adjustment; and, through the data corresponding to the passenger's feedback operation, the target reinforcement learning model can be optimized, further improving the intelligent adjustment of the seat adjustment; based on this, the present solution can improve the intelligence of the seat adjustment to improve the user's driving experience.

[0213] It should be understood that the above examples are intended to help those skilled in the art understand the embodiments of the present application, and are not intended to limit the embodiments of the present application to the specific numerical values or specific scenarios illustrated. Those skilled in the art can obviously make various equivalent modifications or variations based on the above examples, and such modifications or variations also fall within the scope of the embodiments of the present application.

[0214] Combined with the above Figures 2 to 4 The seat adjustment method provided by the embodiment of the present application is described in detail; Figure 5 and Figure 6 The device embodiments of the present application are described in detail. It should be understood that the devices in the embodiments of the present application can execute the various methods of the aforementioned embodiments of the present application, that is, the specific working processes of the following various products can refer to the corresponding processes in the aforementioned method embodiments.

[0215] The following combination Figure 5 The seat adjustment device provided in the embodiment of the present application is described in detail.

[0216] like Figure 5 As shown, Figure 5 It is a structural schematic diagram of a seat adjustment device provided in an embodiment of the present application.

[0217] For example, Figure 5 As shown, the seat adjustment device 500 includes:

[0218] An acquisition module 510 is configured to acquire information about the current user on the target seat, where the information includes a body shape feature vector and identification information of the current user.

[0219] a determination module 520 for obtaining initial seat adjustment parameters based on the target body shape feature vector of the current user, and obtaining seat adjustment parameters for the current user based on the identification information of the current user;

[0220] The determination module 520 is further configured to adjust the initial seat adjustment parameter based on the seat adjustment parameter to obtain a first seat adjustment parameter for the current user;

[0221] The adjustment module 530 is configured to adjust the target seat based on the first seat adjustment parameter.

[0222] As a possible implementation, the apparatus 500 provided in the embodiment of the present application further includes:

[0223] The optimization module is used to determine whether a feedback operation is detected, where the feedback operation is an adjustment operation of the target seat by the current user; if a feedback operation is detected, the seat adjustment parameters of the current user are optimized based on the adjustment parameters corresponding to the feedback operation; and the optimized seat adjustment parameters are used as the seat adjustment parameters of the current user.

[0224] As a possible implementation, the determination module 520 is further configured to detect the current state of the current user; obtain a second seat adjustment parameter of the target seat based on the current state; and adjust the target seat based on the second seat adjustment parameter.

[0225] As a possible implementation, the determination module 520 is specifically configured to: determine whether the current state is a fatigue state; if the current state is a fatigue state, obtain a second seat adjustment parameter based on the current posture of the current user.

[0226] As a possible implementation method, the determination module 520 is specifically used to: if the vehicle's driving time is greater than or equal to a preset time, detect whether the current user's posture is an abnormal posture; if the current user's posture is an abnormal posture, determine that the current state is a fatigue state; if the current user's posture is a posture other than an abnormal posture, determine that the current state is not a fatigue state.

[0227] As a possible implementation manner, the determination module 520 is specifically configured to obtain a second seat adjustment parameter based on the identification information of the current user and the current posture of the current user.

[0228] As a possible implementation, the apparatus 500 provided in the embodiment of the present application further includes:

[0229] A training module is used to use the initial seat adjustment parameters as the initial parameters, adjust the initial parameters with a first adjustment amount, and obtain the target parameters, where the first adjustment amount is any adjustment amount in the seat adjustment amount set; when the target seat is adjusted by the target parameters, the reward value of the first adjustment amount is determined based on the user's feedback information on the target seat and the first adjustment amount; based on the reward value of each adjustment amount in the seat adjustment amount set, the target reinforcement learning model is trained to obtain the trained target reinforcement learning model; wherein the trained target reinforcement learning model is used to output the seat adjustment parameters of the user, and the user includes the current user.

[0230] As a possible implementation, the determination module 520 is specifically configured to: input the target body shape feature vector into the body shape clustering model to obtain initial seat adjustment parameters; wherein the body shape clustering model is configured to obtain seat adjustment parameters corresponding to the body shape feature vector based on the input body shape feature vector.

[0231] It should be noted that, when the seat adjustment device 500 provided in the above embodiment executes the seat adjustment method, the division of the above-mentioned functional modules is only used as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the equipment can be divided into different functional modules to complete all or part of the functions described above.

[0232] In addition, the seat adjustment device and the seat adjustment method provided in the above embodiments belong to the same concept. Therefore, for details not disclosed in the device embodiments of this specification, please refer to the seat adjustment method embodiments mentioned above in this specification, and no further details will be given here.

[0233] Figure 6 It is a structural schematic diagram of a vehicle provided in an embodiment of the present application.

[0234] For example, Figure 6 As shown, the vehicle 600 includes: a memory 601 and a processor 602, wherein the memory 601 stores an executable program code 603, and the processor 602 is used to call and execute the executable program code 603 to perform a seat adjustment method.

[0235] In addition, an embodiment of the present application also protects a seat adjustment device, which may include a memory and a processor, wherein the memory stores executable program code, and the processor is used to call and execute the executable program code to perform a seat adjustment method provided in an embodiment of the present application.

[0236] In this embodiment, the device can be divided into functional modules based on the above-described method examples. For example, each functional module can be mapped to a specific functional module, or two or more functions can be integrated into a single processing module. The integrated module can be implemented in hardware. It should be noted that the module division in this embodiment is illustrative and represents only a logical functional division. In actual implementation, other division methods may be used.

[0237] In the case of dividing the functional modules into corresponding functional modules, the device may further include an acquisition module, a detection module, a processing module, a control module, etc. It should be noted that all relevant contents of each step involved in the above method embodiment can be referred to the functional description of the corresponding functional module and will not be repeated here.

[0238] It should be understood that the device provided in this embodiment is used to execute the above-mentioned seat adjustment method, and thus can achieve the same effect as the above-mentioned implementation method.

[0239] In the case of an integrated unit, the device may include a processing module and a storage module. When the device is applied to a vehicle, the processing module may be used to control and manage the movement of the vehicle.

[0240] The processing module may be a processor or controller that implements or executes the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. The processor may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a digital signal processing system (DSP) and a microprocessor, and the storage module may be a memory.

[0241] In addition, the device provided in the embodiments of the present application can specifically be a chip, component or module, and the chip may include a connected processor and memory; wherein the memory is used to store instructions, and when the processor calls and executes the instructions, the chip can execute a seat adjustment method provided in the above embodiment.

[0242] This embodiment also provides a computer-readable storage medium, which stores computer program code. When the computer program code runs on a computer, the computer executes the above-mentioned related method steps to implement a seat adjustment method provided in the above embodiment.

[0243] This embodiment further provides a computer program product. When the computer program product is run on a computer, the computer is caused to execute the above-mentioned related steps to implement a seat adjustment method provided in the above embodiment.

[0244] Among them, the device, computer-readable storage medium, computer program product or chip provided in this embodiment are all used to execute the corresponding methods provided above. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0245] Through the description of the above implementation methods, technical personnel in the relevant field can understand that for the convenience and simplicity of description, only the division of the above-mentioned functional modules is used as an example. In actual applications, the above-mentioned functions can be distributed and completed by different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0246] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0247] The above content is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of the present application. Therefore, the scope of protection of the present application should be based on the scope of protection of the claims.

Claims

1. A seat adjustment method, characterized in that: The method comprises: Acquiring information about a current user on a target seat, the information about the current user including a target body shape feature vector of the current user and identification information of the current user; obtaining initial seat adjustment parameters based on the target body shape feature vector, and obtaining seat adjustment parameters of the current user based on the identification information; adjusting the initial seat adjustment parameter based on the seat adjustment parameter to obtain a first seat adjustment parameter for the current user; The target seat is adjusted based on the first seat adjustment parameter.

2. The method according to claim 1, characterized in that After adjusting the target seat based on the first seat adjustment parameter, the method further includes: determining whether a feedback operation is detected, the feedback operation being an adjustment operation of the target seat by the current user; If the feedback operation is detected, optimizing the seat adjustment parameters for the current user based on the adjustment parameters corresponding to the feedback operation; The optimized seat adjustment parameters are used as the seat adjustment parameters of the current user.

3. The method according to claim 1, characterized in that After adjusting the target seat based on the first seat adjustment parameter, the method further includes: Detecting the current status of the current user; Based on the current state, obtaining a second seat adjustment parameter of the target seat; The target seat is adjusted based on the second seat adjustment parameter.

4. The method according to claim 3, characterized in that The obtaining, based on the current state, a second seat adjustment parameter of the target seat includes: determining whether the current state is a fatigue state; If the current state is the fatigue state, the second seat adjustment parameter is obtained based on the current posture of the current user.

5. The method according to claim 4, characterized in that The determining whether the current state is a fatigue state includes: If the driving time of the vehicle is greater than or equal to the preset time, detecting whether the current user's posture is abnormal; If the current user posture is the abnormal posture, determining that the current state is the fatigue state; If the current user's posture is a posture other than the abnormal posture, it is determined that the current state is not the fatigue state.

6. The method according to claim 4, characterized in that The obtaining of the second seat adjustment parameter based on the current posture of the current user includes: The second seat adjustment parameter is obtained based on the identification information and the current posture.

7. The method according to any one of claims 1 to 6, characterized in that The method further comprises: Taking the initial seat adjustment parameter as an initial parameter, adjusting the initial parameter by a first adjustment amount to obtain a target parameter, where the first adjustment amount is any adjustment amount in a seat adjustment amount set; In a case where the target seat is adjusted according to the target parameter, determining a reward value for the first adjustment amount based on user feedback information on the target seat and the first adjustment amount; Training a target reinforcement learning model based on a reward value of each adjustment amount in the seat adjustment amount set to obtain a trained target reinforcement learning model; The trained target reinforcement learning model is used to output seat adjustment parameters of the user, including the current user.

8. The method according to any one of claims 1 to 6, characterized in that The obtaining of the current user's information includes: Acquire a target image of the current user and a pressure value of a pressure sensor in the target seat, wherein the target image includes a color image and a depth image; Inputting the color image, the depth image, and the pressure value into a body shape detection model to obtain the target body shape feature vector; The body shape detection model is used to obtain the posture key points of the current user based on the color image and the depth image, and to obtain the target body shape feature vector based on the posture key points and the pressure value.

9. The method according to any one of claims 1 to 6, characterized in that The initial seat adjustment parameters are obtained based on the target body shape feature vector, including: Inputting the target body shape feature vector into a body shape clustering model to obtain the initial seat adjustment parameters; The body shape clustering model is used to obtain the seat adjustment parameters corresponding to the input body shape feature vector based on the input body shape feature vector.

10. A vehicle, characterized in that: The vehicle comprises: a memory for storing executable program code; A processor is configured to call and run the executable program code from the memory, so that the vehicle executes the method according to any one of claims 1 to 9.

Images