Control of a motor vehicle
By using machine learning technology to identify users and boundary conditions, and training models to optimize vehicle presets, the problem of users having difficulty managing multiple presets is solved, and more intelligent vehicle control that meets user expectations is achieved.
Patent Information
- Application Number
- CN202580025425.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2024-04-05
- Filing Date
- 2025-02-26
- Publication Date
- 2026-11-10
AI Technical Summary
In modern motor vehicles, users find it difficult to effectively manage a large number of adjustable presets, and existing pattern recognition technologies cannot accurately predict user expectations, resulting in unclear or difficult user interfaces.
Machine learning technology is used to train machine learning models to predict and control motor vehicles by identifying users and determining boundary conditions. This includes using artificial neural networks (KNN, CNN, DNN, GAN) and biosensors, combined with user expressions and vehicle status, to optimize preset control.
It improves the accuracy of preset controls and user experience, and can dynamically adjust vehicle functions according to user preferences and environmental conditions, simplifying the operation process.
Smart Images

Figure CN123076530A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the control of motor vehicles. In particular, this invention relates to the preset control of motor vehicles. Background Technology
[0002] Motor vehicles have a number of functions to control their movement—particularly in the longitudinal or lateral directions. Other functions relate to, for example, the safety of occupants inside or outside the vehicle, in-vehicle entertainment or information, and comfort features (such as seat adjustment or air conditioning). Functions not directly related to driving can be set according to the user's preferences or intentions. Thus, the user can, for example, set the ventilation to a comfortable temperature or select the volume of a hands-free calling device.
[0003] The number of presets that can be changed by the user can be high in modern vehicles. Often, it is nearly impossible for the average user to review all the presets or remember how they can all be changed. It has been shown that many users frequently change certain presets. Therefore, the interface for changing presets may become slightly unclear or difficult to operate.
[0004] It has been proposed to identify user preset patterns. However, existing pattern recognition methods often fail to accurately predict user presets, forcing users to ultimately violate the pattern recognition to implement the presets. Summary of the Invention
[0005] The objective of this invention is to provide an improved technique for controlling onboard presets in motor vehicles. This invention achieves this objective through the technical solutions provided in the independent claims. The dependent claims provide preferred embodiments.
[0006] According to a first aspect of the present invention, a first method for controlling a motor vehicle includes the steps of: identifying a user of the motor vehicle; determining existing boundary conditions; and determining a preset of the motor vehicle based on a machine learning technique trained for this purpose, the preset being configured with the boundary conditions and the user; and controlling the determined preset.
[0007] Compared to known techniques, machine learning can improve the approximation of preset control to user expectations. Machine learning can also map a large number of input parameters to a small number of output parameters. Here, machine learning can improve interpolation between learned data points, thus enabling meaningful control even in unlearned situations.
[0008] According to a second aspect of the present invention, a second method for training a machine learning technique includes the steps of: identifying a user of a motor vehicle; determining a user-controlled preset of the motor vehicle; determining boundary conditions for the existence of the preset; and training the machine learning technique to associate the preset with the boundary conditions and the user.
[0009] The second method can set up techniques for training the machine learning, which can be used within the scope of the first method. These two methods can be integrated, nested, or implemented sequentially, such that the machine learning techniques are further trained on the user during operation. This also allows for further improvement or adaptation of already good initial control to the user over time.
[0010] It should be noted that the technology described herein can also assist different users of the same vehicle. Therefore, methods can be implemented for different users. A dedicated machine learning technique can be configured for a user, and this dedicated machine learning technique is preferably trained only for that user. The technique may, for example, include a model or network exclusively configured for the user.
[0011] The user can refer to the driver or other persons protruding from the vehicle in a similar manner, and the preset can relate to functions concerning guiding the vehicle. Multiple methods can also be implemented independently or in parallel for different persons in the vehicle. A preset determined for the driver can have priority over a preset determined for another person in the vehicle.
[0012] Preferably, the technique is pre-trained with respect to multiple additional users. For this purpose, the second method can be implemented with different users, and the machine learning technique can determine the consistency or deviation of preferences among the individual users. Training data can be collected from any user in the vehicle while the user is driving. Based on detection on each vehicle in the fleet, sufficient training data for training the machine learning technique can be easily obtained at a low cost.
[0013] Therefore, it is recommended to observe users over a longer period, such as a year, to also detect the impact on preset climatic or weather-based conditions controlled by the users. Furthermore, it is recommended to train the technology based on users who are representative of users who should be driving a given motor vehicle. Motor vehicle manufacturers can easily determine which types of users they manufacture for and which user groups they need to consider for this purpose. In practice, it may be sufficient to pre-train the technology based on dozens to hundreds of representative users.
[0014] The pre-trained technique can be further trained on the vehicle and, in particular, in parallel with the implementation of the first method, with respect to the corresponding user. Generally, there is no upper limit to the amount of training data for techniques used in machine learning, so the second method can also be implemented persistently alongside the first method. Slow changes, such as the age of the vehicle or the age of the user, can also be tracked, and preset controls can be adapted accordingly.
[0015] Boundary conditions can be determined, in particular, based on control or measurement values on the vehicle. This can be done, for example, by querying sensors or evaluating messages from the control system. Exemplary values or messages may relate to external temperature, precipitation, daytime, day of the week, route traveled, road grade, speed, or existing lighting conditions.
[0016] Boundary conditions can also be determined based on messages received from outside the vehicle. These messages preferably relate to the vehicle's operation, particularly its mode of travel or route guidance. For example, the messages may relate to existing or anticipated weather or traffic conditions.
[0017] Boundary conditions can include user expressions. Users can express themselves in different ways, including not only explicit expressions that the user intends to make, but also implicit or tacit expressions that are observable from the user, even if the user does not intend to convey them.
[0018] In one implementation, the expression includes the user's control intervention. The user can, in particular, manually change or influence presets. To do this, the user can manipulate control elements, such as levers, switches, or buttons.
[0019] In another implementation, the user's expression is determined based on their facial expressions or gaze direction. This allows identification of where the user's attention is directed or to determine their emotional state. Both parameters can be used independently as input parameters to determine preset parameters.
[0020] In another implementation, the expression can be determined based on the user's posture. The posture can be determined in particular with respect to the head, arms, and / or legs. The frequency or extent of the user's movements or changes in posture can also be determined.
[0021] More preferably, the pattern or sequence of the user's expressions is determined. This also allows for the identification and evaluation of complex or varied expressions made by the user.
[0022] A motor vehicle may have a system configured to control predetermined functions of the vehicle. The preset may relate to the type and manner of control. The function itself may not be the object of the preset. Thus, for example, lateral control of the vehicle may be performed by the user, but the degree of steering assist may be preset according to applicable scenarios. The control may relate to the vehicle's movement and, for example, to the adjustment of the running gear, external lighting, or control programs for the drive motor. The control may also relate to functions other than the vehicle's motion functions, such as air conditioning, seat adjustment, or entertainment or communication systems for occupants. Furthermore, it may control, for example, safety functions, welcome functions, comfort functions, or ambient functions, such as interior or exterior lighting or aromatherapy. The preset can be controlled according to the user's preferences or physical condition.
[0023] The technology preferably includes a kinematic neural network (KNN), wherein different types of KNNs can be used. In a particularly preferred embodiment, a first KNN is used to determine the user's stance, and a second KNN is trained to determine a preset based on the stance. Multiple machine learning techniques or combinations of different techniques can also be used to control the preset. The machine learning techniques used can include GAN (Generative Adversarial Network), CNN (Convolutional Neuronal Network), or DNN (Deep Neuronal Network).
[0024] According to another aspect of the present invention, a control device for controlling a motor vehicle includes: a device for identifying a user of the motor vehicle; a first interface for detecting information indicating the presence of boundary conditions; a processing device; and a preset second interface for controlling the motor vehicle. Here, the processing device is configured to implement machine learning techniques, wherein the techniques are trained to determine presets for the motor vehicle that are compatible with the boundary conditions and the user.
[0025] The control device, and in particular the processing device, can be configured to partially or completely implement the methods described herein. For this purpose, the processing device can be implemented electronically and, for example, include integrated switching circuitry, programmable logic modules, or programmable microcomputers. The method can be implemented in the form of a configuration or as a computer program product having program code segments for the processing device. The configuration or the computer program product can be stored on a computer-readable data carrier. Features or advantages of the method can be transferred to the device, or vice versa.
[0026] The device used to identify a user may include, for example, a camera or a biological sensor. The user may also be identified by means of a password, the input of which is detected. More preferably, the user may be identified by means of a token assigned to them. The token may preferably be wirelessly readable. The token may be used to control the vehicle's security functions, for example, in the form of a digital vehicle key. A digital car key (DCK) may be stored on a handheld device or a mobile device assigned to the user. To utilize the DCK, the user can be authenticated relative to the mobile device, for example, by entering a password or by presenting biometric data.
[0027] According to another aspect of the invention, a motor vehicle includes the control devices described herein. The motor vehicle preferably includes a motorcycle or a passenger car. Attached Figure Description
[0028] The invention will now be described in more detail with reference to the accompanying drawings, in which:
[0029] Figure 1 Showing the control devices on a motor vehicle;
[0030] Figure 2 A flowchart illustrating the method is shown;
[0031] Figure 3 This illustrates an exemplary machine learning technique for recognizing user behavior. Detailed Implementation
[0032] Figure 1 A control device 100 on a motor vehicle 105 is shown. The control device 100 has a processing device 110 configured to implement, form, or execute learnable techniques or machine learning techniques. An exemplary such technique is shown in the form of an artificial neural network KNN 115.
[0033] The processing device 110 is connected to one or more sources of information indicating boundary conditions present on or in areas of the vehicle 105. Sensor 120, interior space camera 125, ambient environment camera 130, and receiving device 135 are exemplarily shown. An interface 140 is also provided, through which presets of devices on the vehicle 105 can be controlled.
[0034] User 150 can control the presets of the motor vehicle or the presets of the control system on the motor vehicle 105. The presets preferably influence the type and manner in which predetermined control functions are implemented by the control system. Such presets can also be referred to as preferences.
[0035] The control device 100 is configured to identify the user 150 and determine boundary conditions present in the vehicle or in the surrounding environment of the motor vehicle 105. A KNN 115 can be trained to establish a correlation between the existing boundary conditions and a preset set by the user 150. The trained KNN can then control the preset based on the identified boundary conditions. Preferably, however, the training of the KNN 115 continues further by determining the user 150's expressions or behavior. This allows determination of whether the user is satisfied with the set preset. The determined satisfaction level can be introduced into the control as a correction. If necessary, the preset can be modified by adjusting the control based on the detected satisfaction level.
[0036] User 150 can be identified, for example, biologically using interior space camera 125. Other feasible solutions include identifying a token assigned to user 150, such as (preferably wirelessly) a readable handheld device (Fob, Keyfob). Interior space camera 125 can also be used to identify user 150’s facial expressions, gaze direction, or posture.
[0037] The environmental camera 130 can be used to determine the surrounding environmental conditions of the motor vehicle 105, which are related to the operation or driving of the motor vehicle 105, such as the weather, current traffic density, traffic conditions, or existing lighting conditions.
[0038] Information indicating boundary conditions affecting the operation of motor vehicle 105 can be received from an external location using receiving device 135. Such boundary conditions may include, for example, weather, precipitation, temperature, or traffic information.
[0039] Sensor 120 can detect any parameters relating to the operation of vehicle 105, such as the volume of the vehicle, the open or closed state of doors, hoods, or roof, the driving status of vehicle 105, location, road traveled, date, day of the week, time, number of occupants, speed, planned route, and past or remaining driving time. Sensor 120 can also be configured as a control device or control system, providing suitable information in the form of messages.
[0040] Figure 2 A schematic flowchart illustrates a preset method 200 for a preset or control system for controlling a motor vehicle 105 based on existing boundary conditions. The control is preferably relative to a predetermined user 150.
[0041] In step 205, information is collected from sources 120-135, wherein the information indicates the boundary conditions present on vehicle 105. This can be done by performing measurements, requesting messages, or receiving messages.
[0042] In step 210, the received information may be organized, grouped, and / or pre-filtered. It should be noted that this step may already include the use of trained machine learning techniques. The information may be correlated and / or fused. Unreliable information may be discarded, while reliable information may be enhanced or refined.
[0043] Subsequently, in step 215, one or more presets can be determined using machine learning techniques, particularly using KNN 115, regarding information about the boundary conditions instructing the operation of the vehicle 105. Preferably, the presets are determined as an associated group and can be controlled on the vehicle 105 in step 220. For this purpose, one or more control systems of the vehicle 105 can be set or configured accordingly with presets.
[0044] The presets act on the vehicle 105 and thus on the user 150 in a certain way. In particular, the functions of the vehicle 105 can be controlled according to the presets. Multiple presets can work in a coordinated manner and influence the user 150's experience with the vehicle 105.
[0045] In step 225, an expression from user 105 can be determined, indicating how satisfied the user is with the controlled preset or what changes the user wishes to make. This expression can be identified at user 150, or user 150's actions regarding the preset can be detected and evaluated. This allows, for example, determining whether preset automatic control (e.g., increasing the temperature in the interior space of vehicle 105) is acknowledged or strengthened by user 150, or whether user 150 is instead weakening, reverting, or even reversing the control.
[0046] In step 230, the evaluation can be used to examine or adjust the correlation between the KNN 115 and the preset conditions. This process can be understood as training or learning.
[0047] Figure 3 Exemplary machine learning techniques are shown for recognizing user 150's behavior regarding boundary conditions and preset parameters. Different techniques are interconnected and linked in an exemplary manner to ultimately control preset parameters on vehicle 105. The association is... Figure 3 The techniques are graphically represented by arrows and should be understood as exemplary as described. It is not necessary to set up all the described techniques or connect them to each other in the manner shown.
[0048] In step 305, facial expressions (facial features) of user 150 can be detected. In step 310, user 150 can be identified, for example, by means of biological features. This identification can be performed, for example, with respect to their face. The identification can be performed, for example, by means of a convolutional neural network (CNN). A known algorithm for this is known under the name ReLU.
[0049] In step 315, the weather in the area of vehicle 105 can be predicted. This can be done using a deep neural network (DNN), such as the Sigmoid algorithm. In step 320, the date and / or time can be determined. In step 325, the music preferences of user 150 can be determined. This can also be done using a CNN, such as the Adam algorithm.
[0050] In step 330, the user 150's mood can be determined, for example, using a CNN based on the ReLU algorithm. In step 335, the driving route of the vehicle 105 can be determined. Here, the location on the driving route or driving conditions within the vehicle's area based on the driving route (e.g., the type of road traveled, its slope or gradient, or nearby attractions) can also be determined.
[0051] In step 340, the navigation destination of vehicle 105 can be determined based on observation. This can be done, for example, by identifying previously traveled segments of a known route. Frequently traveled routes, such as commuting routes, shopping routes, or school routes, can be identified in this way.
[0052] In step 345, the state of the motor vehicle 105 can be determined. The state may include, for example, driving status and / or traffic conditions.
[0053] Regarding the parameters determined, particularly those determined in steps 330 to 345, different presets can then be determined. For example, preset 350 can be used for the sound effects of the vehicle's audio equipment. Preset 355 can also be controlled for seat adjustment. Preset 360 for aromatherapy and preset 365 for driving route can also be controlled.
[0054] Other presets that can be determined or controlled include interior lighting 370, the arrangement of operating elements 375, vehicle dynamics 380, shift characteristics 385 of the gearbox, or characteristics of the drivetrain 390 of the motor vehicle 105. Shift characteristics may include, for example, the selection of shift programs or shift features, such as Eco, Sport, or Winter. Drivetrain characteristics may include similar programs.
[0055] List of reference numerals
[0056] 100 control device
[0057] 105 Motor Vehicles
[0058] 110 processing equipment
[0059] 115 Artificial Neural Networks (KNN)
[0060] 120 sensor
[0061] 125 interior space camera
[0062] 130 Surrounding Environment Camera
[0063] 135 receiving equipment
[0064] 140 interface
[0065] 150 users
[0066] 200 methods
[0067] 205 Inspection Information
[0068] 210 Information Compilation
[0069] 215 Determine boundary conditions and derive presets
[0070] 220 control preset
[0071] 225 Inspection, Adaptation, and Training
[0072] 305 Facial Expression Recognition
[0073] 310 Facial Recognition
[0074] 315 Weather Forecast
[0075] 320 date and time
[0076] 325 Music Preferences
[0077] 330 Emotional Confirmation
[0078] 335 route model
[0079] 340 Predictive Navigation Destination
[0080] 345 vehicle status
[0081] 350 sound presets
[0082] 355 seat adjustment preset
[0083] 360° Aroma Presets
[0084] 365 route preset
[0085] 370 Interior Lighting Preset
[0086] 375 Operating element layout preset
[0087] 380 vehicle dynamics preset
[0088] 385 Shift Characteristics Preset
[0089] 390 drivetrain preset
Claims
1. A method (200) for controlling a motor vehicle (105), wherein, The method (200) includes the following steps: - Identify (210) the user (150) of the motor vehicle (105); - Determine the boundary conditions for the existence of (205); and - Based on the machine learning technique (115) trained for this purpose, determine (215) a preset for the motor vehicle (105), the preset being configured with the boundary conditions and the user (150); and - The preset determined by control (220).
2. Methods (200) for techniques (115) used to train machine learning, wherein, The method (200) includes the following steps: - Identify (210) the user (150) of the motor vehicle (105); - Determine (205) the user-controlled preset of the motor vehicle (105); - Determine (215) the boundary conditions for the existence of the preset condition; and - Train (225) the machine learning technique (115) to associate the preset with the boundary conditions and the user (150).
3. The method (200) according to claim 1 or 2, wherein, The technique (115) is pre-trained with multiple additional users (150).
4. The method (200) according to any one of the preceding claims, wherein, Boundary conditions are determined based on messages received from outside the motor vehicle (105).
5. The method (200) according to any one of the preceding claims, wherein, The boundary conditions include the expression of the user (150).
6. The method (200) according to claim 5, wherein, The expression includes the user's (150) control intervention.
7. The method (200) according to claim 5 or 6, wherein, The expression is determined based on the user's (150) facial expression.
8. The method (200) according to any one of claims 5 to 7, wherein, The expression is determined based on the posture of the user (150).
9. The method (200) according to any one of the preceding claims, wherein, Determine the pattern or order of the user's (150) expression.
10. The method (200) according to any one of the preceding claims, wherein, The motor vehicle (105) includes a system configured to control predetermined functions of the motor vehicle (105), and the preset relates to the type and manner of the control.
11. The method (200) according to any one of the preceding claims, wherein, The technology (115) includes artificial neural networks.
12. The method (200) according to claim 11, wherein, The technology (115) includes one of GAN, CNN and DNN.
13. A control device (100) for controlling a motor vehicle (105), wherein, The device includes the following components: - Device (125) for identifying the user (150) of the motor vehicle (105); - A first interface (120-135) for detecting information indicating the presence of boundary conditions; and - A processing device (110) configured to implement a machine learning technique (115), wherein the technique (115) is trained to determine a preset for the motor vehicle (105) in conjunction with the boundary conditions and the user (150); and - A preset second interface (140) for controlling the motor vehicle (105).
14. A motor vehicle (105) having a control device (100) according to claim 13.
