Vehicle control method and device based on passenger comfort, storage medium and equipment

Abstract

The application provides a vehicle control method and device based on passenger comfort, a storage medium and equipment, integrates vehicle motion characteristics and path characteristics to obtain vehicle motion multi-dimensional characteristics, determines the contribution of each dimension to passenger comfort based on the vehicle motion multi-dimensional characteristics, automatically adjusts the specific weight of the MPC cost function based on the feature contribution degree affecting passenger comfort, and dynamically adjusts the vehicle control strategy or parameter weight in real time, thereby improving the poor adaptability of the vehicle control caused by artificial experience or offline fixed weight in the prior art; meanwhile, the soft constraint condition and the hard constraint condition are determined based on the specific comfort baseline of the passenger in the process of determining the MPC cost function, the vehicle control can be dynamically adjusted according to the comfort requirements of specific passengers, so as to overcome the differences in different passenger comfort requirements and different factors inducing motion sickness of different passengers, and realize the effect of individualized relief of user motion sickness in the automatic driving vehicle.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of vehicle control, in particular to a vehicle control method and device based on passenger comfort, a storage medium and equipment. BACKGROUND

[0002] With the rapid development of automatic driving technology, its application in the fields of travel service, logistics transportation, public transportation, etc. is increasingly widespread. However, with the popularization of the application range, there are higher requirements for vehicle motion control in the process of automatic driving for passenger experience, especially in terms of vehicle motion smoothness, acceleration change rate and steering stability, etc. In the automatic driving environment, passengers lose control of the vehicle and engage in non-driving tasks (such as reading, using electronic devices, etc.), and the incidence of motion sickness is significantly higher than that in the traditional driving mode, which seriously affects the acceptance and application promotion of the automatic driving vehicle.

[0003] To solve the above problems, the prior art provides various ways to alleviate the motion sickness of passengers in the automatic driving vehicle, including visual stimulation, active suspension adjustment, path planning and driving style adjustment, and comfort optimization, etc. Among them, visual stimulation is to provide light stimulation or virtual picture simulation of the passenger's observation of the road surface around the passenger's vision, to eliminate the conflict between the vestibule and the visual input, but both light stimulation and virtual picture need to increase the hardware cost of the automatic driving vehicle, and cannot solve the discomfort caused by the vehicle motion itself. Active suspension adjustment is to suppress the vibration transmission in the frequency range of 0.1-0.5 Hz through an active suspension system, thereby reducing the low-frequency vibration energy transmitted to the passengers and reducing the probability of motion sickness induction, but this scheme requires the vehicle to be equipped with an active suspension system, which needs to sacrifice a certain road feel of the vehicle in exchange for comfort, the suppression of high-frequency vibration is poor, and it cannot be adapted to individual differences, affecting the overall riding experience. Path planning and driving style adjustment select relatively gentle road sections in the path planning stage and reduce the probability of sudden acceleration and sudden braking in the driving style, but for passengers, the optimized path or driving style may cause the user's trip to be extended, affecting the passenger's travel efficiency, and lacks systematic control in the adjustment of driving style. For passengers with different sensitivities to motion sickness, the same strategy is used. Comfort optimization dynamically adjusts the vehicle speed curve of MPC or the vehicle control behavior that causes passenger discomfort based on genetic algorithm or reinforcement learning, but this adjustment process is executed offline and cannot be adjusted according to the actual driving situation of the vehicle, and this method lacks consideration of the comfort of specific passengers in the process of adjusting the vehicle speed curve or adjusting the vehicle driving behavior. SUMMARY

[0004] Based on this, the application provides a vehicle control method, device, storage medium and equipment based on passenger comfort, a vehicle motion model is constructed based on real-time motion characteristics of the vehicle, and the specific relevance of each parameter dimension to user comfort is determined in combination with vehicle motion characteristics and path planning characteristics, so as to construct a cost function for adjusting the vehicle motion model, wherein the cost function adjusts the control strategy of the vehicle motion by adjusting the specific parameter weight according to the relevance of the user comfort, and the constraint condition of the cost function is determined based on the comfort baseline of the specific user, so that the vehicle control can be adjusted and constrained in combination with the individual differences of the specific user for the vehicle ride comfort in the process of adjusting the vehicle motion control strategy, the comfort of the specific user is improved in real time and individually, and thus the motion sickness of the passengers riding the autonomous vehicle is relieved.

[0005] In a first aspect, the application provides a vehicle control method based on passenger comfort, which is applied to a vehicle comfort control system, the vehicle comfort control system comprising a vehicle data acquisition unit, a passenger comfort feedback unit and a vehicle motion control unit, the vehicle motion control unit being in communication connection with the vehicle data acquisition unit and the passenger comfort feedback unit, the vehicle data acquisition unit being used to acquire real-time motion characteristics of the vehicle and path characteristics of a vehicle planning path, the passenger comfort feedback unit being used to acquire actual comfort feedback of a target passenger, and the vehicle motion control unit being used to generate a control instruction of the corresponding vehicle according to the received vehicle real-time motion characteristics, path characteristics of the vehicle planning path and real-time comfort feedback of the target passenger;

[0006] The vehicle control method based on passenger comfort is cyclically executed according to a preset control period, and each preset control period comprises the following steps:

[0007] Real-time motion data of the vehicle and path data of the vehicle planning path are acquired, and the real-time motion data and the path data are fused to obtain vehicle motion multi-dimensional characteristics;

[0008] The vehicle motion multi-dimensional characteristics are input into a trained motion sickness prediction model to obtain a predicted comfort of the target passenger;

[0009] The vehicle motion multi-dimensional characteristics are input into a motion sickness relevance model to obtain a weight of the MPC model;

[0010] The soft constraint threshold and the hard constraint threshold of the target passenger comfort are determined according to a comfort baseline of the target passenger in a standard driving scene, and the soft constraint threshold is lower than the hard constraint threshold;

[0011] The weight of the MPC model is adjusted according to the predicted comfort, the soft constraint threshold and the hard constraint threshold.

[0012] updating the MPC model by combining the adjusted weight, inputting the vehicle motion model and the vehicle motion multi-dimensional feature into the updated MPC model, and obtaining the optimal control sequence of the vehicle;

[0013] executing the first control instruction of the optimal control sequence of the vehicle, and obtaining the actual comfort degree of the target passenger feedback;

[0014] adjusting the MPC model weight by combining the deviation of the predicted comfort degree and the actual comfort degree, and taking the adjusted MPC model weight as the MPC model weight obtained in the next preset control period;

[0015] updating the control instruction of the vehicle according to the adjusted MPC model weight, the vehicle motion model and the first control instruction.

[0016] Further, the motion sickness correlation model includes a Pearson correlation analysis model and a random forest model.

[0017] The inputting of the vehicle motion multi-dimensional feature into the motion sickness correlation model to obtain the weight of the MPC model is specifically:

[0018] The vehicle motion multi-dimensional feature is inputted into the Pearson correlation analysis model to obtain the correlation coefficient of each vehicle motion multi-dimensional feature and the motion sickness.

[0019] The correlation coefficient of each vehicle motion multi-dimensional feature and the motion sickness is inputted into the random forest model to obtain the importance proportion of the vehicle motion multi-dimensional feature.

[0020] The weight of the MPC model is determined according to the importance proportion of each vehicle motion multi-dimensional feature.

[0021] Further, the adjusting of the weight of the MPC model according to the predicted comfort degree, the soft constraint threshold and the hard constraint threshold is specifically:

[0022] If the predicted comfort degree is lower than the soft constraint threshold, the MPC model weight is maintained unchanged.

[0023] If the predicted comfort degree is higher than the soft constraint threshold and lower than the hard constraint threshold, the MPC model weight is increased.

[0024] If the predicted comfort degree is higher than the hard constraint threshold, an emergency protection mechanism is triggered, and the MPC model weight is increased.

[0025] Further, the MPC model weight comprises a jerk-related feature corresponding penalty term and an angular motion feature corresponding penalty term, and the MPC model weight is adjusted according to the deviation between the predicted comfort degree and the actual comfort degree.

[0026] If the deviation between the predicted comfort degree and the actual comfort degree is greater than a preset first error threshold, and the predicted comfort degree is lower than the actual comfort degree, the weight of the jerk-related feature corresponding penalty term and the angular motion feature corresponding penalty term is increased.

[0027] If the deviation between the predicted comfort degree and the actual comfort degree is less than a preset second error threshold, the weight of the jerk-related feature corresponding penalty term and the angular motion feature corresponding penalty term is decreased.

[0028] If the deviation between the predicted comfort degree and the actual comfort degree is greater than the preset second error threshold, and the deviation between the predicted comfort degree and the actual comfort degree is less than the preset first error threshold, the weight of the jerk-related feature corresponding penalty term and the angular motion feature corresponding penalty term is maintained unchanged.

[0029] Further, the process of determining the comfort degree baseline of the target passenger in the standard driving scene is specifically as follows:

[0030] Obtain vehicle motion data in the standard driving scene and comfort degree scores of the target passenger;

[0031] Extract multi-dimensional motion features according to the vehicle motion data;

[0032] Input the multi-dimensional motion features and the comfort degree scores of the target passenger into the trained comfort degree prediction model to obtain the comfort degree baseline of the target passenger in the standard driving scene.

[0033] Further, the vehicle control method based on passenger comfort degree further comprises:

[0034] If the number of target passengers in the vehicle is greater than 1, the comfort degree baseline of the target passenger in the standard driving scene is the average of the comfort degree baselines of each passenger in the vehicle in the standard driving scene.

[0035] The predicted comfort degree of the target passenger is the average of the predicted comfort degrees of each target passenger in the vehicle.

[0036] In a second aspect, the present application further provides a vehicle control device based on passenger comfort, which is applied to a vehicle comfort control system, the vehicle comfort control system comprising a vehicle data acquisition unit, a passenger comfort feedback unit and a vehicle motion control unit, the vehicle motion control unit being in communication connection with the vehicle data acquisition unit and the passenger comfort feedback unit respectively, the vehicle data acquisition unit being configured to acquire real-time motion characteristics of a vehicle and path characteristics of a vehicle planning path, the passenger comfort feedback unit being configured to acquire actual comfort degree fed back by a target passenger, and the vehicle motion control unit being configured to generate a control instruction of a corresponding vehicle according to the received real-time motion characteristics of the vehicle, the path characteristics of the vehicle planning path and the real-time comfort degree fed back by the target passenger.

[0037] The vehicle control device based on passenger comfort is cyclically executed according to preset control periods, and each preset control period comprises the following modules:

[0038] a vehicle characteristic acquisition module, configured to acquire real-time motion data of a vehicle and path data of a vehicle planning path, and to fuse the real-time motion data and the path data to obtain vehicle motion multi-dimensional characteristics;

[0039] a comfort degree prediction module, configured to input the vehicle motion multi-dimensional characteristics into a trained motion sickness prediction model to obtain a predicted comfort degree of a target passenger;

[0040] a weight determination module, configured to input the vehicle motion multi-dimensional characteristics into a motion sickness correlation model to obtain a weight of an MPC model;

[0041] a constraint determination module, configured to determine a soft constraint threshold and a hard constraint threshold of the comfort degree of the target passenger according to a comfort degree baseline of the target passenger in a standard driving scenario, the soft constraint threshold being lower than the hard constraint threshold;

[0042] a weight first adjustment module, configured to adjust the weight of the MPC model according to the predicted comfort degree, the soft constraint threshold and the hard constraint threshold;

[0043] a weight update instruction module, configured to update the MPC model in combination with the adjusted weight of the MPC model, input a vehicle motion model and the vehicle motion multi-dimensional characteristics into the updated MPC model, and obtain an optimal control sequence of the vehicle;

[0044] an update instruction execution module, configured to execute a first control instruction of the optimal control sequence of the vehicle, and acquire an actual comfort degree fed back by the target passenger;

[0045] a weight second adjusting module, configured to adjust the MPC model weight in combination with the deviation between the predicted comfort degree and the actual comfort degree, and take the adjusted MPC model weight as the MPC model weight obtained in a next preset control period;

[0046] a control updating module, configured to update the control instruction of the vehicle according to the adjusted MPC model weight, the vehicle motion model and the first control instruction.

[0047] Further, the motion sickness correlation model comprises a Pearson correlation analysis model and a random forest model.

[0048] The weight determining module comprises:

[0049] a correlation analysis component, configured to input the vehicle motion multi-dimensional features into the Pearson correlation analysis model to obtain a correlation coefficient of each vehicle motion multi-dimensional feature and the motion sickness;

[0050] an importance proportion component, configured to input the correlation coefficient of each vehicle motion multi-dimensional feature and the motion sickness into the random forest model to obtain an importance proportion of the vehicle motion multi-dimensional features;

[0051] a weight determining component, configured to determine the weight of the MPC model according to the importance proportion of each vehicle motion multi-dimensional feature.

[0052] In a third aspect, the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the vehicle control method based on passenger comfort degree in any of the first aspect.

[0053] In a fourth aspect, the present application further provides a computer device, which comprises a memory and a processor, the memory stores a computer program, and the processor executes the computer program to execute the vehicle control method based on passenger comfort degree in any of the first aspect.

[0054] The beneficial effects of the above technical solutions are that: the vehicle motion multi-dimensional features are constructed by integrating the vehicle motion data and the path data, and the contribution of each dimension to the passenger comfort is determined based on the vehicle motion multi-dimensional features, so that the specific weights of the MPC cost function are automatically adjusted based on the feature contribution degree affecting the passenger comfort, so that the vehicle control strategy or parameter weight can be dynamically adjusted in real time, and the poor adaptability of the vehicle control caused by artificial experience or offline fixed weight in the prior art is improved; meanwhile, the soft constraint condition and the hard constraint condition are determined based on the specific comfort baseline of the passenger in the process of determining the MPC cost function, so that the vehicle control can be dynamically adjusted according to the comfort requirements of the specific passenger, so as to overcome the differences in the comfort requirements of different passengers and the factors inducing motion sickness of different passengers, and realize the personalized effect of relieving motion sickness of the user in the automatic driving vehicle. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced as follows.

[0056] Figure 1 A vehicle control method based on passenger comfort in one embodiment of the present application is shown in the schematic diagram.

[0057] Figure 2 A comparison between the MPC optimized vehicle control and the artificial baseline optimized vehicle control in one embodiment of the present application is shown in the schematic diagram.

[0058] Figure 3 The average jerk and the comfort correlation degree in one embodiment of the present application are shown in the schematic diagram.

[0059] Figure 4 A vehicle control device based on passenger comfort in one embodiment of the present application is shown in the schematic diagram. DETAILED DESCRIPTION

[0060] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all the other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the present application. In order to more specifically describe the present application, the vehicle control method, device, storage medium and equipment based on passenger comfort provided by the present application will be described in detail in combination with the drawings.

[0061] Unless otherwise defined, technical terms or scientific terms used in the present disclosure shall have the ordinary meaning as understood by a person having ordinary skill in the art to which the present disclosure pertains. The terms "first", "second", and similar terms are used herein to distinguish one element from another, but do not necessarily require or imply these elements to be in any certain order, number, or importance. Similarly, the terms "one", "a", or "the" do not require or imply a quantity of at least one. The terms "comprise", "comprising", and similar terms are intended to encompass the elements listed thereafter, and equivalents thereof, without excluding other elements. The terms "connected" or "coupled" are not limited to physical or mechanical connections or couplings, but can include electrical connections or couplings, whether direct or indirect. The terms "upper", "lower", "left", "right", and the like are used to indicate relative positions, and can change when the absolute positions of the described objects change.

[0062] With the rapid development of autonomous driving technology, its application in the fields of travel service, logistics transportation, public transportation, etc. is increasingly widespread. However, as the scope of application of autonomous driving is promoted, passengers' experience puts higher requirements on vehicle motion control, especially motion smoothness, acceleration change rate and steering stability. Studies have shown that in the autonomous driving environment, due to the loss of control over the vehicle by passengers and the engagement in non-driving tasks (such as reading, using electronic devices), the incidence of motion sickness is significantly higher than that in the traditional driving mode, which seriously affects the acceptance and application of autonomous driving cars by users.

[0063] Currently, the problem of motion sickness in autonomous vehicles is mainly affected by the following factors:

[0064] (1) Too large jerk: Jerk is the time rate of change of acceleration, and studies have shown that it is the main factor inducing motion sickness. The sharp acceleration change of the vehicle under the working conditions of starting, braking, turning, etc. will stimulate the vestibular system of the passenger, leading to sensory conflict and discomfort.

[0065] (2) Angular motion irregularity: The yaw angular velocity and lateral acceleration generated by the vehicle when driving on a curve, changing lanes and U-turning will further aggravate the symptoms of motion sickness of the passenger, especially in high-speed turning scenarios.

[0066] (3) Significant individual differences: The sensitivity of different passengers to motion stimuli varies significantly, related to factors such as age, gender, vestibular sensitivity, etc. Existing control methods usually use comfort models based on population averages (such as the ISO 2631 standard), which cannot meet individualized needs.

[0067] (4) Real-time computation and personalized balance challenge: Although the traditional Model Predictive Control (MPC) method can optimize multiple objective functions within the prediction time domain, the weight parameters thereof are usually dependent on experience setting or offline tuning, and it is difficult to realize online personalized adaptation.

[0068] Among them, Model Predictive Control (MPC) is considered as a suitable framework for designing path tracking controllers for autonomous vehicles. This technology solves an optimization problem at each time step and can handle multiple objectives simultaneously. In addition, it can adapt to the physical limitations of actuators and vehicle states to ensure safety and other required behaviors.

[0069] Based on this, the embodiment provides a vehicle control method based on passenger comfort, which is applied to a vehicle comfort control system, the vehicle comfort control system comprising a vehicle data acquisition unit, a passenger comfort feedback unit and a vehicle motion control unit, and the vehicle motion control unit is in communication connection with the vehicle data acquisition unit and the passenger comfort feedback unit respectively.

[0070] The vehicle data acquisition unit is used to acquire real-time motion data (such as vehicle three-axis acceleration , and vehicle three-axis angular velocity , and vehicle position, vehicle speed, vehicle heading angle and vehicle steering angle) and path data (such as whether the planned path is provided with a corner, a U-turn, a corner radius, etc.) of a vehicle planning path. Among them, the vehicle data acquisition unit can include an inertial measurement unit, a positioning unit, a gyroscope, an offline navigation map, etc.

[0071] The passenger comfort feedback unit is used to acquire the actual comfort degree of the target passenger feedback. Among them, the passenger comfort feedback unit in the embodiment can be a communication device with information display and input function, such as a vehicle interactive screen, a mobile phone and the like. The actual comfort degree of user feedback can be represented by MISC (Motion Illness Symptoms Classification, Motion Sickness Symptoms Classification) score, wherein the value range of MISC score is [0, 10], when MISC score is 0, it means that the target passenger has no discomfort, when MISC score is 10, it means that the target passenger's motion sickness performance is the most serious.

[0072] The vehicle motion control unit is configured to generate control instructions for the corresponding vehicle according to the received real-time motion characteristics of the vehicle, path characteristics of the vehicle planning path, and real-time comfort feedback of the target passenger. In the present embodiment, the vehicle motion control unit can be a device terminal for on-board or remote control of the vehicle, including but not limited to a smartphone and a computer device, wherein the computer device can be at least one of a desktop computer, a portable computer, a laptop computer, a mainframe computer, a tablet computer, etc. When the device terminal receives the real-time motion data and path data of the vehicle, the MPC cost function of the adjusted vehicle model can be adjusted in combination with the comfort baseline of the specific passenger, so as to adjust the control strategy of the vehicle and reduce the degree of motion sickness of the specific passenger when riding the vehicle. The specific implementation steps of the embodiment of the vehicle control method based on passenger comfort are described in detail with reference to the accompanying Figure 1 The vehicle control method based on passenger comfort is shown in the schematic diagram of the vehicle control method based on passenger comfort. The specific implementation steps of the embodiment of the vehicle control method based on passenger comfort are described in detail with reference to the accompanying

[0073] It should be noted that the vehicle control method based on passenger comfort in the present embodiment is executed cyclically according to a preset control period. Each preset control period includes the following steps of the vehicle control method based on passenger comfort:

[0074] Step S100, obtaining real-time motion data of the vehicle and path data of the vehicle planning path, and fusing the real-time motion data and the path data to obtain vehicle motion multi-dimensional characteristics.

[0075] The real-time motion data of the vehicle is measured in real time by the vehicle data acquisition unit during vehicle driving, and at least includes real-time longitudinal acceleration of the vehicle, real-time lateral acceleration of the vehicle, real-time yaw angular velocity of the vehicle, real-time lateral acceleration of the vehicle, and statistical values (such as mean value, peak value and standard deviation) thereof, for representing the real-time motion state of the vehicle. The path data of the vehicle planning path can be obtained by statistical analysis of the specific planning path data of the vehicle, mainly reflecting whether there are road sections causing motion sickness of passengers, such as curves, turns, U-turns, etc., including but not limited to real-time path curvature and statistical values (such as mean value, peak value and standard deviation) thereof.

[0076] Since whether the motion sickness of the passenger can be alleviated is not only related to the motion data in the driving process of the vehicle, but also related to the specific planned path passed by the vehicle, in order to make the control of the vehicle more accurate in the stage of alleviating the motion sickness, the embodiment comprehensively considers the motion characteristics of the vehicle and the path characteristics of the planned path, and integrates all the above characteristics into vehicle motion multi-dimensional characteristics. Among them, the vehicle motion multi-dimensional characteristics include jerk related characteristics, angular motion characteristics and path tracking characteristics, such as the vehicle lateral jerk mean value, lateral jerk standard deviation, jerk amplitude mean value, jerk amplitude standard deviation, etc. The jerk amplitude and its statistical characteristics calculated based on the three-axis jerk, the angular motion characteristics calculated based on the vehicle speed and the steering angle, and the path tracking characteristics calculated based on the vehicle position coordinates and the planned trajectory.

[0077] Step S200, input the vehicle motion multi-dimensional characteristics into the trained motion sickness prediction model to obtain the predicted comfort level of the target passenger.

[0078] Specifically, the motion sickness prediction model of the embodiment can be implemented by using a random forest model, a support vector machine model (SVM) or a gradient boosting tree model (such as XGBoost, LightGBM, etc.) or a deep neural network model (such as LSTM, Transformer). The training process of the motion sickness prediction model is not described in detail here. Among them, the support vector machine model and the gradient boosting tree model are suitable for predicting passenger comfort under different data size and computing resource conditions; the deep neural network is suitable for capturing the time sequence dependence relationship of the motion characteristics under the condition of sufficient data amount, and improving the prediction accuracy.

[0079] Further, considering that the number of vehicle motion multi-dimensional characteristics is large, if all vehicle motion multi-dimensional characteristics are input into the motion sickness prediction model to obtain the predicted comfort level of the user, a large amount of operation on the above data is inevitable. In view of this, the embodiment screens the vehicle motion multi-dimensional characteristics input into the motion sickness prediction model according to the correlation between each vehicle motion multi-dimensional characteristic and the passenger comfort level, specifically:

[0080] Step S201, sort the correlation between each vehicle motion multi-dimensional characteristic and the passenger comfort level from high to low, and cut off the vehicle motion multi-dimensional characteristics with high correlation according to the preset number, to obtain the screened vehicle motion multi-dimensional characteristics.

[0081] Specifically, the screening of the vehicle motion multi-dimensional features in the embodiment can be performed according to a set number, for example, the top 10 vehicle motion multi-dimensional features with the highest correlation degree can be intercepted; a screening correlation degree threshold can also be set, and only the vehicle motion multi-dimensional features with a correlation degree greater than the correlation degree threshold are retained. In the embodiment, the screening correlation degree threshold can be dynamically updated based on historical data (for example, if the vehicle motion multi-dimensional features with a correlation degree less than a have little effect on vehicle control in the historical control records of the vehicle, the correlation degree threshold can be adjusted to a, and the correlation degree threshold can be further adjusted in combination with the correlation degrees of other vehicle motion multi-dimensional features as the control records of the vehicle are continuously updated, if it is found that the vehicle motion multi-dimensional features with a correlation degree less than a play a role in the subsequent n times of vehicle control); the screening correlation degree threshold can also be set in combination with the sensitivity of motion sickness of the passenger, for example, more vehicle motion multi-dimensional features can be retained for passengers who are more sensitive to motion sickness, and fewer vehicle motion multi-dimensional features can be retained for passengers who are less sensitive to motion sickness, and only the key vehicle motion multi-dimensional features (such as the jerk of the vehicle, the angular velocity, the path curvature radius of the current vehicle control period, etc.) can be retained.

[0082] Step S202, inputting the screened vehicle motion multi-dimensional features into the trained motion sickness prediction model to obtain the predicted comfort degree of the passenger.

[0083] It should be noted that when the vehicle motion multi-dimensional features are screened, the 5-fold cross-validation can also be used to evaluate the generalization of the trained motion sickness prediction model to improve the reliability of data prediction.

[0084] In addition, the vehicle motion multi-dimensional feature data input into the motion sickness prediction model can also be optimized, including:

[0085] The vehicle motion multi-dimensional features are subjected to multi-stage filtering and noise reduction processing, including resampling and sliding average filtering processing of the vehicle motion multi-dimensional features, which reduces the calculation burden while retaining the vehicle motion features in the 0.1-5Hz frequency band related to motion sickness;

[0086] The vehicle motion multi-dimensional features are subjected to outlier rejection, including identifying and rejecting statistical outliers of the vehicle motion multi-dimensional features by using the quartile method to ensure the quality of the vehicle motion feature data;

[0087] The filtered vehicle acceleration related signals are subjected to central difference method to calculate the jerk, and the mean, standard deviation, and root mean square value of the jerk are extracted.

[0088] Step S300, inputting the vehicle multi-dimensional features into the motion sickness correlation degree model to obtain the weight of the MPC model.

[0089] Specifically, the weight of the MPC model discussed in the embodiment refers to the penalty weight of the MPC model, a coefficient for suppressing the state of vehicle motion or controlling the state of vehicle motion, including but not limited to the penalty weight corresponding to the jerk-related feature, the angular motion feature, and the path tracking feature. The motion sickness correlation model in the embodiment includes a Pearson correlation analysis model and a random forest model, which in turn processes the vehicle multi-dimensional features and their output quantities to obtain the weight of the specific MPC model, including the following steps:

[0090] In step S301, the vehicle motion multi-dimensional features are input into the Pearson correlation analysis model to obtain the correlation coefficients of each vehicle motion multi-dimensional feature and motion sickness.

[0091] Specifically, the Pearson correlation analysis model is used to evaluate the linear correlation between each vehicle motion multi-dimensional feature and the perception of motion sickness. For example, the lateral jerk mean correlation coefficient is the highest (r = 0.748), followed by the lateral jerk standard deviation (r = 0.672), the jerk amplitude standard deviation (r = 0.637), and the jerk amplitude mean (r = 0.633).

[0092] In step S302, the correlation coefficients of each vehicle motion multi-dimensional feature and motion sickness are input into the random forest regression model to obtain the importance proportion of the vehicle motion multi-dimensional features.

[0093] Specifically, after obtaining the correlation coefficients of each vehicle motion multi-dimensional feature and motion sickness, the correlation coefficients are input into the random forest regression model to obtain the feature importance analysis corresponding to each vehicle motion multi-dimensional feature. For example, the jerk-related feature accounts for 71.2% of the total importance, the angular motion feature accounts for 24.5% of the importance, and the path tracking feature accounts for only 4.3% of the importance.

[0094] In step S303, the weight of the MPC model is determined according to the importance proportion of each vehicle motion multi-dimensional feature.

[0095] Specifically, the importance proportion of each vehicle motion multi-dimensional feature is mapped to the weight coefficient of the corresponding penalty term in the MPC model. The mapping method is as follows: taking the path tracking term weight as the reference value 1.0, dividing the importance percentage of each comfort-related feature by 10 as the weight coefficient of the corresponding penalty term. For example, the importance of the jerk-related feature accounts for 71.2%, so the weight coefficient of the penalty term corresponding to the jerk-related feature is 7.1; the importance of the angular motion feature accounts for 24.5%, so the weight coefficient of the penalty term corresponding to the angular motion feature is 2.4; the importance of the path tracking feature accounts for 4.3%, so the weight coefficient of the penalty term corresponding to the path tracking feature is 0.2. The form of the MPC cost function is:

[0096] ,

[0097] a penalty term corresponding to the jerk-related feature, a weight coefficient of the penalty term corresponding to the jerk-related feature, a penalty term corresponding to the angular motion feature, a weight coefficient of the penalty term corresponding to the angular motion feature, a penalty term corresponding to the path tracking feature, a weight coefficient of the penalty term corresponding to the path tracking feature. In step S400, a soft constraint threshold and a hard constraint threshold of the target passenger comfort are determined according to a comfort baseline of the target passenger in a standard driving scene, the soft constraint threshold being lower than the hard constraint threshold.

[0098] In this embodiment, different comfort baselines are set for different target passengers, and this setting makes the subsequent adjustment of the MPC model to control the vehicle motion have obvious passenger individualization setting, avoiding the prior art of simply dividing passengers into different sensitivity types and ignoring the different sensitivity characteristics of different passengers to different factors of the vehicle, so that the motion sickness of the specific target passenger can be more effectively relieved in the vehicle adjustment process.

[0099] Specifically, in one of the embodiments, the determination process of the comfort baseline of the target passenger in the standard driving scene is specifically as follows:

[0100] In step S401, vehicle motion data in a standard driving scene and a comfort score of the target passenger are obtained.

[0101] The vehicle motion data in the standard driving scene and the comfort score of the target passenger can be divided into a data collection stage, and the objective vehicle motion data and the subjective comfort score of the target passenger are associated in the specific standard driving scene. In order to obtain a more complete target passenger comfort evaluation, the above standard driving scene includes one or more combinations of starting, straight acceleration, turning, corner driving, U-turn and parking; the comfort score of the target passenger adopts the MISC scale (Motion Illness Symptoms Classification), and the score range is 0-10 points, 0 points indicating no symptoms and 10 points indicating severe discomfort.

[0102] In step S402, multi-dimensional motion features are extracted according to the vehicle motion data.

[0103] Specifically, the multi-dimensional motion features extracted from the vehicle motion data include jerk-related features (such as lateral jerk mean, lateral jerk standard deviation, jerk amplitude, and jerk amplitude standard deviation), angular motion features (such as three-axis angular velocity), and path tracking features.

[0104] In step S403, the multi-dimensional motion features and the comfort score of the target passenger are input into the trained comfort prediction model to obtain the comfort baseline of the target passenger in the standard driving scenario.

[0105] Specifically, the multi-dimensional motion features and the comfort score of the target passenger are input into the trained comfort prediction model (such as a random forest regression model) to output the predicted comfort score of the passenger in the standard driving scenario. The parameters of the comfort prediction model are continuously adjusted so that the predicted comfort score approaches the comfort score of the target passenger. At this time, the predicted comfort scores in all standard driving scenarios are the comfort baseline of the passenger.

[0106] It should be noted that different passengers have different sensitivities to the same motion stimulus, so under the same vehicle and path conditions, the comfort baselines of different passengers are different, which is the basis for realizing personalized control. Different suspension systems, wheelbase, body stiffness, and other parameters of different vehicles will result in differences in motion features (jerk, angular velocity, etc.) generated under the same driving operation, thereby affecting the comfort baseline of the passenger. Different features such as curvature of bends, slope, and road roughness of different paths will also result in differences in vehicle motion features, thereby affecting the comfort baseline.

[0107] Further, the comfort baseline of the passenger makes it possible to ensure the comfort of all passengers in the vehicle and reduce the incidence of motion sickness. The comfort baseline of the passenger is the average of the comfort baselines of all passengers in the vehicle, and the predicted comfort of the passenger is the average of the predicted comfort of all passengers in the vehicle. Specifically,

[0108] If the number of target passengers in the vehicle is greater than 1, the comfort baseline of the target passenger in the standard driving scenario is the average of the comfort baselines of all passengers in the vehicle in the standard driving scenario, and the predicted comfort of the target passenger is the average of the predicted comfort of all target passengers in the vehicle.

[0109] In addition, the MPC model is provided with a soft constraint threshold and a hard constraint threshold. The soft constraint threshold refers to a constraint condition that can be violated in a vehicle control process according to a specific scene. For example, in a climbing scene, even if the maximum throttle speed still cannot start, the speed constraint is violated. The soft constraint threshold is mainly used to avoid conflicts between the output and the input of the vehicle motion model. The hard constraint threshold refers to a constraint threshold that cannot be violated, such as the movement amount of the vehicle throttle, the maximum speed of the motor, and the like. After the comfort baseline of the target passenger is determined, the specific soft constraint threshold and the hard constraint threshold can be determined based on the comfort baseline. The threshold of the soft constraint condition is the comfort baseline plus a first preset value, and the threshold of the hard constraint condition is the comfort baseline plus a second preset value, wherein the second preset value is greater than the first preset value.

[0110] In step S500, the weight of the MPC model is adjusted according to the predicted comfort, the soft constraint threshold, and the hard constraint threshold.

[0111] Specifically, adjusting the weight of the MPC model according to the predicted comfort, the soft constraint threshold, and the hard constraint threshold includes:

[0112] In step S501, if the predicted comfort is lower than the soft constraint threshold, the weight of the MPC model is maintained unchanged.

[0113] In step S502, if the predicted comfort is higher than the soft constraint threshold and lower than the hard constraint threshold, the weight of the MPC model is increased.

[0114] In step S503, if the predicted comfort is higher than the hard constraint threshold, an emergency protection mechanism is triggered, and the weight of the MPC model is increased.

[0115] The adjustment mode of the emergency protection mechanism is that the weight of the jerk-related feature corresponding to the penalty term is increased to 1.2 times of the original weight, and the weight of the angular motion feature corresponding to the penalty term is increased to 1.1 times of the original weight. The adjustment range of the jerk-related feature corresponding to the penalty term is greater than that of the angular motion feature, because the contribution of the jerk-related feature to motion sickness (71.2%) is significantly higher than that of the angular motion feature (24.5%). The adjustment range is controlled within 10%-20% to avoid control shock or system instability caused by sudden weight changes.

[0116] For example, the comfort baseline of the passenger is denoted as The expression of the soft constraint condition can be denoted as The expression of the hard constraint condition can be denoted as Thus, the configuration of the double-threshold personalized constraint boundary is realized.

[0117] Step S600, updating the MPC model combined with the adjusted MPC model weight, inputting the vehicle motion model and the vehicle motion multi-dimensional feature into the updated MPC model to obtain the optimal control sequence of the vehicle.

[0118] Specifically, since the MPC model is used to control the vehicle motion model, after updating the weight of the MPC model, the control strategy of the MPC model for the vehicle motion model will also change, thereby obtaining the optimal control sequence (such as the front wheel steering angle sequence, the acceleration sequence, etc.) of the vehicle in the future control time domain (such as the future 1 second, the future 1 control step, etc.).

[0119] In this embodiment, the vehicle motion model adopts a standard kinematic bicycle model, which contains five state quantities, including vehicle position coordinates, heading angle, vehicle speed, vehicle acceleration, and steering angle. The control input quantity of the vehicle motion model is the vehicle acceleration and the steering angle speed. The vehicle wheelbase parameter in the vehicle motion model can be set to 2.7 meters, and the time step is 0.1 seconds. For a specific vehicle model, a data-driven vehicle model can be calculated by a Gaussian process regression or a neural network model.

[0120] Step S700, executing the first control instruction of the optimal control sequence of the vehicle, and obtaining the actual comfort degree of the target passenger feedback.

[0121] Specifically, when the vehicle is controlled according to the first control instruction (such as the front wheel steering angle, acceleration, etc.) of the optimal control sequence, the actual comfort degree of the current target passenger feedback is collected, and the model can be corrected or compensated based on the deviation between the actual comfort degree and the predicted comfort degree.

[0122] Step S800, adjusting the MPC model weight combined with the deviation between the predicted comfort degree and the actual comfort degree, and taking the adjusted MPC model weight as the MPC model weight obtained in the next preset control period.

[0123] In one embodiment, the MPC model weight includes a jerk-related feature corresponding penalty term and an angular motion feature corresponding penalty term, and the MPC model weight is adjusted combined with the deviation between the predicted comfort degree and the actual comfort degree, specifically as follows:

[0124] Step S801, if the deviation between the predicted comfort degree and the actual comfort degree is greater than a preset first error threshold, and the predicted comfort degree is lower than the actual comfort degree, increasing the weight of the jerk-related feature corresponding penalty term and the angular motion feature corresponding penalty term.

[0125] For example, the weight of the jerk-related feature corresponding penalty term is the jerk-related feature, the weight of the penalty term corresponding to the angular motion feature is also increased by a ratio of the jerk-related feature, the weight of the penalty term corresponding to the angular motion feature is also increased by a ratio of is a preset learning rate.

[0126] If the deviation between the predicted comfort level and the actual comfort level is less than a preset second error threshold, the weight of the penalty term corresponding to the jerk-related feature and the weight of the penalty term corresponding to the angular motion feature are reduced.

[0127] For example, the weight of the penalty term corresponding to the jerk-related feature is reduced by a ratio of the weight of the penalty term corresponding to the angular motion feature is also reduced by a ratio of is a preset learning rate.

[0128] If the deviation between the predicted comfort level and the actual comfort level is greater than the preset second error threshold and the deviation between the predicted comfort level and the actual comfort level is less than a preset first error threshold, the weight of the penalty term corresponding to the jerk-related feature and the weight of the penalty term corresponding to the angular motion feature are maintained unchanged.

[0129] The first error threshold is greater than the second error threshold.

[0130] According to the adjusted MPC model weight, the vehicle motion model and the first control instruction, the control instruction of the vehicle is updated.

[0131] In addition, the method for training the motion sickness prediction model of step S200 of the embodiment specifically includes the following steps:

[0132] A standardized test route is selected, which is a closed test route without traffic signals and interference from other vehicles. Passengers of different ages and genders are recruited by stratified sampling, and the test vehicles are classified according to the experience of the drivers to control the difference in driving style. The comfort level of the passengers in the test vehicle can be obtained by discrete scoring of the MISC scale, or by collecting continuous physiological signals (such as heart rate variability, skin conductance, eye movement tracking frequency, etc.) of the passengers to obtain the comfort level of the passengers, or by combining the above subjective scoring and objective physiological indicators to construct a multi-modal passenger comfort level score. The relevant data of each test measurement and the corresponding driver of the test vehicle are combined into complete training data. The training data is divided into a data training set and a data validation set in proportion. The data training set is input into the motion sickness prediction model to adjust the specific parameters in the motion sickness prediction model, and the data validation set is used to verify the trained motion sickness prediction model until the difference between the predicted comfort level of the passengers and the historical comfort level of the passengers is less than a set threshold, and then the motion sickness prediction model is trained and output.

[0133] ​To better illustrate the passenger comfort-based vehicle control method of the present embodiment, a sedan of brand A is taken as an experimental platform, and a multi-channel synchronous data acquisition system is realized by combining the Sensor Logger application as an example for illustration:

[0134] The sampling frequency of the inertial measurement unit configured on the vehicle is 100 Hz (actually 99.8 Hz), including a three-axis accelerometer (range ±8g) and a three-axis gyroscope (range ±2000° / s), and the GPS module samples at a frequency of 1 Hz, with a horizontal positioning accuracy of 5.63±3.79m. In addition, the vehicle-mounted sensor records the steering wheel angle and brake pedal position at 50Hz. All the above data streams are synchronized with a unified timestamp, with an alignment accuracy of ±1ms.

[0135] For the training of the motion sickness prediction model, a 6.9 km closed standardized test road is selected for testing, with no traffic lights, no other vehicle interference, and flat road surface. The route contains 12 consecutive standardized driving scenes, in the order of: starting, straight acceleration, turning, bend, turning, straight driving, U-turn, straight driving, turning, and parking. The scene design covers typical motion patterns in daily driving, including longitudinal acceleration and deceleration, lateral turning, and combined motion, with a total test duration controlled within 10 minutes. 96 participants (age 18-65, average 34.3 years old) are recruited using stratified sampling principles (male and female each 48), with 3 participants in each experimental vehicle, and the drivers are divided into three levels according to experience: novice (1-3 years), skilled (3-10 years), and experienced (more than 10 years), to control the influence of driving style differences on motion characteristics. The MISC scale (0-10 points) is used to evaluate the symptoms of motion sickness. After the end of each driving scene, three raters independently give the MISC score, and the interclass correlation coefficient (ICC) is used to evaluate the consistency of the scores. When any rater gives MISC≥7 or the participant actively requests, the safety parking procedure is immediately executed.

[0136] The collected vehicle motion multi-dimensional features are optimized and pre-processed, including data resampling, filtering and noise reduction processing, outlier rejection processing, jerk calculation, and feature statistics. Among them, the original inertial sensor data is resampled from 99.8 Hz to 10 Hz, and the average based on time and linear interpolation method is used. This resampling reduces the computational burden while retaining the 0.1-5 Hz frequency band motion features related to motion sickness. A 5-point center sliding average filter is applied to the three-axis acceleration signal and its synthesized amplitude, that is, the average value of each two sampling points before and after the current time is taken as the filter output, and the time interval is 0.1 seconds. The interquartile range (IQR) method is used to identify and eliminate statistical outliers, and the effective data range is defined as the first quartile minus 1.5 times the interquartile range to the third quartile plus 1.5 times the interquartile range. Jerk is calculated by central difference method, that is, the difference between the filtered acceleration of adjacent time points is divided by twice the time interval. The following statistical features are extracted for each driving scene: jerk-related features, angular velocity features, and path tracking features.

[0137] The pre-processed vehicle motion multi-dimensional features are input into the trained correlation model (Pearson correlation analysis model) to evaluate the relationship between vehicle motion multi-dimensional features and motion sickness perception, among which the mean value of lateral jerk has the highest correlation coefficient (r = 0.748), followed by the standard deviation of lateral jerk (r = 0.672), the standard deviation of jerk amplitude (r = 0.637), and the mean value of jerk amplitude (r = 0.633), all reaching statistical significance. According to the above correlation, the top 10 motion features with the strongest correlation are selected as input variables, and a personalized comfort prediction model is trained using random forest regression algorithm. The data set is divided into training set and test set (7:3) according to the group, and 5-fold cross-validation is used to evaluate the generalization performance. The model achieves a prediction accuracy of R² = 0.838 on the test set of 84 participants.

[0138] The above vehicle multi-dimensional features and motion sickness perception correlation are input into the random forest regression model to obtain the importance analysis results of each vehicle multi-dimensional feature: the jerk-related parameters account for 71.2% of the total importance, the angular motion features contribute 24.5%, and the path tracking features only account for 4.3%. This result confirms the central role of jerk in comfort optimization, as shown in the average jerk and comfort correlation diagram. Figure 3

[0139] ​Based on this, a standard kinematic bicycle model is adopted. The state vector includes five state variables: position coordinates, heading angle, velocity, and steering angle. The control inputs are acceleration and steering rate. The model parameters are set with a wheelbase of 2.7 meters and a time step of 0.1 seconds. An MPC cost function is designed based on feature importance analysis results, including a trajectory tracking term, a penalty term corresponding to jerk-related features, a penalty term corresponding to angular motion features, and a penalty term corresponding to path deviation. The weight of the penalty term corresponding to jerk-related features is set to 7.1 times the baseline (corresponding to 71.2% feature contribution), the weight of the penalty term corresponding to angular motion features is 2.4 times the baseline (corresponding to 24.5% contribution), and the weight of the penalty term corresponding to path deviation is 0.2 times the baseline (corresponding to 4.3% contribution). Two levels of constraints are automatically generated based on the individual comfort baseline: soft constraints are set to the baseline plus 1.0 point, increasing the comfort term weight when exceeded; hard constraints are set to the baseline plus 2.0 point, forcibly implementing comfort-priority control when exceeded. For example, for an occupant with a comfort baseline of 2.19, the soft constraint is 3.19 and the hard constraint is 4.19. The system monitors the deviation between predicted and actual comfort in real time. When the deviation exceeds a high threshold of 1.0 and the actual comfort is lower than the predicted value, the system increases motion smoothing penalties; when the error is below a low threshold of 0.5, the weights are gradually reduced to avoid overly conservative control. The learning rate is set to 0.1. When the occupant comfort falls below the protection threshold of 5.0, an emergency protection mechanism is triggered, immediately increasing the Jerk weight to 1.2 times and the angular motion weight to 1.1 times.

[0140] Furthermore, this embodiment can also use the Sequential Quadratic Programming (SLSQP) algorithm to solve the MPC optimization problem in real time, minimizing the cost function while satisfying safety constraints and actuator boundaries.

[0141] Combined with appendix Figure 2 The diagram shows a comparison between MPC-optimized vehicle control and manual baseline-optimized vehicle control. The example experiment results indicate that all 21 driving groups showed improvement. The weighted average score of the manual baseline was 2.05 points, which decreased to 1.26 points after MPC optimization, an average improvement of 0.79 points, representing a relative improvement of 38.5%. Group 19 showed the greatest improvement (1.21 points), while Group 17 showed the smallest improvement (0.29 points). 85.7% of the groups (18 / 21) showed an improvement exceeding 0.5 points. The MPC system's average solution time was 3.6 milliseconds, with a maximum of 5.3 milliseconds, far below the 100 millisecond real-time control requirement, resulting in a 100% optimization success rate. The improvement was slightly greater in high-jerk conditions (curves, turns, U-turns), with the average solution time for each scenario below 5.5 milliseconds.

[0142] It should be understood that, although attached Figure 1The steps in the flowchart are shown in sequence according to the arrows, but the steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the steps are not strictly limited in sequence, and the steps can be executed in other sequences. Moreover, the steps in the flowchart can include multiple sub-steps or sub-phases, which are not necessarily executed at the same time, but can be executed at different times. The execution sequence of the sub-steps or sub-phases is not necessarily sequential, but can be alternated or alternated with at least part of other steps or sub-steps or sub-phases of other steps. Figure 1 The steps in the flowchart are shown in sequence according to the arrows, but the steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the steps are not strictly limited in sequence, and the steps can be executed in other sequences. Moreover, the steps in the flowchart can include multiple sub-steps or sub-phases, which are not necessarily executed at the same time, but can be executed at different times. The execution sequence of the sub-steps or sub-phases is not necessarily sequential, but can be alternated or alternated with at least part of other steps or sub-steps or sub-phases of other steps.

[0143] The vehicle control method based on passenger comfort degree is described in detail in the above embodiments of the present disclosure. The above method disclosed by the present disclosure can be implemented in various forms of equipment, and therefore the present disclosure also discloses a vehicle control device based on passenger comfort degree. The vehicle control device based on passenger comfort degree is shown in the accompanying drawings, and specific embodiments are given below for detailed description. Figure 4 The vehicle control device based on passenger comfort degree is shown in the accompanying drawings, and specific embodiments are given below for detailed description.

[0144] The vehicle control device based on passenger comfort degree is applied to a vehicle comfort control system, which includes a vehicle data acquisition unit, a passenger comfort feedback unit, and a vehicle motion control unit. The vehicle motion control unit is in communication connection with the vehicle data acquisition unit and the passenger comfort feedback unit. The vehicle data acquisition unit is used to acquire real-time motion characteristics of the vehicle and path characteristics of the vehicle planning path. The passenger comfort feedback unit is used to acquire actual comfort degree feedback from the target passenger. The vehicle motion control unit is used to generate a control instruction for the corresponding vehicle according to the received real-time motion characteristics of the vehicle, path characteristics of the vehicle planning path, and real-time comfort degree feedback from the target passenger.

[0145] The vehicle control device based on passenger comfort degree is executed cyclically according to a preset control period. Each preset control period includes the following modules:

[0146] The vehicle feature acquisition module 1001 is used to acquire real-time motion data of the vehicle and path data of the vehicle planning path, and fuse the real-time motion data and the path data to obtain vehicle motion multi-dimensional characteristics.

[0147] The comfort degree prediction module 1002 is used to input the vehicle motion multi-dimensional characteristics into a trained motion sickness prediction model to obtain a predicted comfort degree of the target passenger.

[0148] The weight determination module 1003 is used to input the vehicle motion multi-dimensional characteristics into a motion sickness correlation model to obtain a weight of the MPC model.

[0149] The constraint determination module 1004 is configured to determine a soft constraint threshold and a hard constraint threshold of the target passenger comfort according to a comfort baseline of the target passenger in a standard driving scene, the soft constraint threshold being lower than the hard constraint threshold;

[0150] The weight first adjustment module 1005 is configured to adjust the weight of the MPC model according to the predicted comfort, the soft constraint threshold and the hard constraint threshold;

[0151] The weight update instruction module 1006 is configured to update the MPC model in combination with the adjusted weight of the MPC model, input the vehicle motion model and the vehicle motion multi-dimensional features into the updated MPC model, and obtain the optimal control sequence of the vehicle;

[0152] The update instruction execution module 1007 is configured to execute the first control instruction of the optimal control sequence of the vehicle, and obtain the actual comfort feedback by the target passenger;

[0153] The weight second adjustment module 1008 is configured to adjust the weight of the MPC model in combination with the deviation between the predicted comfort and the actual comfort, and take the adjusted weight of the MPC model as the weight of the MPC model obtained in the next preset control period;

[0154] The control update module 1009 is configured to update the control instruction of the vehicle according to the adjusted weight of the MPC model, the vehicle motion model and the first control instruction.

[0155] Further, the motion sickness correlation model includes a Pearson correlation analysis model and a random forest model.

[0156] The weight determination module includes:

[0157] The correlation analysis component is configured to input the vehicle motion multi-dimensional features into the Pearson correlation analysis model to obtain a correlation coefficient of each vehicle motion multi-dimensional feature and the motion sickness;

[0158] The importance proportion component is configured to input the correlation coefficient of each vehicle motion multi-dimensional feature and the motion sickness into the random forest model to obtain an importance proportion of the vehicle motion multi-dimensional features;

[0159] The weight determination component is configured to determine the weight of the MPC model according to the importance proportion of each vehicle motion multi-dimensional feature.

[0160] The vehicle control device for passenger comfort can refer to the above-mentioned method, which will not be repeated here. Each module in the above-mentioned device can be implemented by software, hardware and their combination. The above-mentioned modules can be embedded in the processor of the terminal device in hardware form or independent of the processor, or stored in the memory of the terminal device in software form, so that the processor calls and executes the operations corresponding to the above-mentioned modules.

[0161] In one embodiment, the present application also provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the steps of the above-mentioned vehicle control method based on passenger comfort.

[0162] The computer readable storage medium can be an electronic storage such as a flash memory, an EEPROM (Electrically Erasable Programmable Read-Only Memory), an EPROM (Erasable Programmable Read-Only Memory), a hard disk or a ROM. Alternatively, the computer readable storage medium includes a non-transitory computer readable medium. The computer readable storage medium has a storage space for storing program codes for executing any method steps of the above-mentioned methods. The program codes can be read from or written into one or more computer program products, and the program codes can be compressed in a suitable form.

[0163] In one embodiment, the present application provides a computer device, which includes a memory and a processor, the memory stores a computer program, and the processor executes the computer program to execute the above-mentioned vehicle control method based on passenger comfort.

[0164] The computer device includes a memory, a processor and one or more computer programs, wherein the one or more computer programs can be stored in the memory and configured to be executed by the one or more processors, and the one or more application programs are configured to execute the above-mentioned vehicle control method based on passenger comfort.

[0165] The processor can include one or more processing cores. The processor connects various parts within the entire computer device by various interfaces and lines, performs various functions of the computer device and processes data by running or executing instructions, programs, code sets or instruction sets stored in the memory, and calling data stored in the memory. Optionally, the processor can be implemented in at least one of a hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), programmable logic array (PLA). The processor can integrate a combination of one or several of central processing unit (CPU), graphics processing unit (GPU) and modem. Among them, the CPU mainly processes operating systems, user interfaces and application programs, etc.; the GPU is responsible for rendering and drawing display content; and the modem is used for processing wireless communication. It can be understood that the above-mentioned modem can also not be integrated into the processor, but be realized by a separate communication chip.

[0166] The memory can include random access memory (RAM) and can also include read-only memory (ROM). The memory can be used to store instructions, programs, codes, code sets or instruction sets. The memory can include a program storage area and a data storage area, wherein the program storage area can store instructions for implementing an operating system, instructions for implementing at least one function (such as a touch function, a sound playing function, an image playing function, etc.), instructions for implementing the above-mentioned various method embodiments, etc. The data storage area can also store data created by the terminal device in use, etc.

[0167] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A vehicle control method based on passenger comfort, characterized by, The passenger comfort-based vehicle control method is applied to a vehicle comfort control system, the vehicle comfort control system comprising a vehicle data acquisition unit, a passenger comfort feedback unit, and a vehicle motion control unit, the vehicle motion control unit being in communication connection with the vehicle data acquisition unit and the passenger comfort feedback unit respectively, the vehicle data acquisition unit being configured to acquire real-time motion data of a vehicle and path data of a vehicle planning path, the passenger comfort feedback unit being configured to acquire actual comfort degree fed back by a target passenger, and the vehicle motion control unit being configured to generate a control instruction of a corresponding vehicle according to the received vehicle real-time motion data, path data of the vehicle planning path, and real-time comfort degree fed back by the target passenger; The passenger comfort-based vehicle control method is executed cyclically according to a preset control period, and each preset control period comprises the following steps: acquiring real-time motion data of a vehicle and path data of a vehicle planning path, fusing the real-time motion data and the path data to obtain vehicle motion multi-dimensional features, inputting the vehicle motion multi-dimensional features into a trained motion sickness prediction model to obtain a predicted comfort degree of a target passenger; inputting the vehicle motion multi-dimensional features into a motion sickness correlation model to obtain a weight of an MPC model; determining a soft constraint threshold and a hard constraint threshold of the target passenger comfort degree according to a comfort degree baseline of the target passenger in a standard driving scenario, the soft constraint threshold being lower than the hard constraint threshold; adjusting the weight of the MPC model according to the predicted comfort degree, the soft constraint threshold, and the hard constraint threshold, updating the MPC model in combination with the adjusted weight of the MPC model, inputting a vehicle motion model and the vehicle motion multi-dimensional features into the updated MPC model to obtain an optimal control sequence of the vehicle; executing a first control instruction of the optimal control sequence of the vehicle, and acquiring an actual comfort degree fed back by the target passenger; adjusting the weight of the MPC model in combination with a deviation between the predicted comfort degree and the actual comfort degree, and taking the adjusted weight of the MPC model as the weight of the MPC model in a next preset control period; updating the control instruction of the vehicle according to the adjusted weight of the MPC model, the vehicle motion model, and the first control instruction; the motion sickness correlation model comprises a Pearson correlation analysis model and a random forest model; the inputting of the vehicle motion multi-dimensional features into the motion sickness correlation model to obtain the weight of the MPC model is specifically as follows: inputting the vehicle motion multi-dimensional features into the Pearson correlation analysis model to obtain a correlation coefficient between each vehicle motion multi-dimensional feature and motion sickness; inputting the correlation coefficient between each vehicle motion multi-dimensional feature and motion sickness into the random forest model to obtain an importance proportion of the vehicle motion multi-dimensional features; determining the weight of the MPC model according to the importance proportion of each vehicle motion multi-dimensional feature.

2. The vehicle control method based on passenger comfort according to claim 1, characterized by, the adjusting of the weight of the MPC model according to the predicted comfort degree, the soft constraint threshold, and the hard constraint threshold is specifically as follows: if the predicted comfort degree is lower than the soft constraint threshold, maintaining the weight of the MPC model unchanged; if the predicted comfort level is higher than the soft constraint threshold and the predicted comfort level is lower than the hard constraint threshold, increasing the MPC model weight; if the predicted comfort level is higher than the hard constraint threshold, triggering an emergency protection mechanism and increasing the MPC model weight.

3. The vehicle control method based on passenger comfort according to claim 1, characterized by, The MPC model weight includes a jerk-related feature corresponding penalty term and an angular motion feature corresponding penalty term, and the adjustment of the MPC model weight based on the deviation between the predicted comfort level and the actual comfort level is specifically: if the deviation between the predicted comfort level and the actual comfort level is greater than a preset first error threshold and the predicted comfort level is lower than the actual comfort level, increasing the weight of the jerk-related feature corresponding penalty term and the angular motion feature corresponding penalty term; if the deviation between the predicted comfort level and the actual comfort level is less than a preset second error threshold, reducing the weight of the jerk-related feature corresponding penalty term and the angular motion feature corresponding penalty term; if the deviation between the predicted comfort level and the actual comfort level is greater than the preset second error threshold and the deviation between the predicted comfort level and the actual comfort level is less than the preset first error threshold, maintaining the weight of the jerk-related feature corresponding penalty term and the angular motion feature corresponding penalty term unchanged.

4. The vehicle control method based on passenger comfort according to claim 1, characterized by, The process of determining the comfort level baseline of the target passenger in the standard driving scene is specifically: obtaining vehicle motion data and comfort level scores of the target passenger in the standard driving scene; extracting multi-dimensional motion features from the vehicle motion data; inputting the multi-dimensional motion features and the comfort level scores of the target passenger into the trained comfort level prediction model to obtain the comfort level baseline of the target passenger in the standard driving scene.

5. The vehicle control method based on passenger comfort according to claim 1, characterized by, Further comprising: if the number of target passengers in the vehicle is greater than 1, the comfort level baseline of the target passenger in the standard driving scene is the average of the comfort level baselines of each passenger in the vehicle in the standard driving scene; the predicted comfort level of the target passenger is the average of the predicted comfort levels of each target passenger in the vehicle.

6. A vehicle control device based on passenger comfort, characterized by, The vehicle control device based on passenger comfort level is applied to a vehicle comfort control system, which includes a vehicle data acquisition unit, a passenger comfort feedback unit and a vehicle motion control unit. The vehicle motion control unit is in communication connection with the vehicle data acquisition unit and the passenger comfort feedback unit. The vehicle data acquisition unit is used to acquire real-time motion features of the vehicle and path features of the vehicle planning path. The passenger comfort feedback unit is used to acquire the actual comfort level feedback by the target passenger. The vehicle motion control unit is used to generate a control instruction of the corresponding vehicle according to the received vehicle real-time motion features, vehicle planning path features and target passenger real-time comfort level feedback; The vehicle control device based on passenger comfort level is executed cyclically according to a preset control period. Each preset control period includes the following modules: a vehicle feature acquisition module, which is used to acquire real-time motion data of the vehicle and path data of the vehicle planning path, and fuse the real-time motion data and the path data to obtain vehicle motion multi-dimensional features; The comfort degree prediction module is configured to input the vehicle motion multi-dimensional features into the trained motion sickness prediction model to obtain a predicted comfort degree of the target passenger. The weight determination module is configured to input the vehicle motion multi-dimensional features into the motion sickness correlation model to obtain the weight of the MPC model. The constraint determination module is configured to determine a soft constraint threshold and a hard constraint threshold of the target passenger comfort degree according to a comfort degree baseline of the target passenger in a standard driving scenario, the soft constraint threshold being lower than the hard constraint threshold. The weight first adjustment module is configured to adjust the weight of the MPC model according to the predicted comfort degree, the soft constraint threshold, and the hard constraint threshold. The weight update instruction module is configured to update the MPC model in combination with the adjusted weight of the MPC model, input the vehicle motion model and the vehicle motion multi-dimensional features into the updated MPC model, and obtain an optimal control sequence of the vehicle. The update instruction execution module is configured to execute a first control instruction of the optimal control sequence of the vehicle, and obtain an actual comfort degree fed back by the target passenger. The weight second adjustment module is configured to adjust the weight of the MPC model in combination with a deviation between the predicted comfort degree and the actual comfort degree, and use the adjusted weight of the MPC model as the weight of the MPC model obtained in a next preset control period. The control update module is configured to update a control instruction of the vehicle according to the adjusted weight of the MPC model, the vehicle motion model, and the first control instruction. The motion sickness correlation model includes a Pearson correlation analysis model and a random forest model. The weight determination module includes: The correlation analysis component is configured to input the vehicle motion multi-dimensional features into the Pearson correlation analysis model to obtain a correlation coefficient of each vehicle motion multi-dimensional feature and motion sickness. The importance proportion component is configured to input the correlation coefficient of each vehicle motion multi-dimensional feature and motion sickness into the random forest model to obtain an importance proportion of the vehicle motion multi-dimensional features. The weight determination component is configured to determine the weight of the MPC model according to the importance proportion of each vehicle motion multi-dimensional feature.

7. A computer-readable storage medium having stored thereon a computer program, characterized in that The computer program is executed by the processor to implement the steps of the vehicle control method based on passenger comfort degree in any one of claims 1-5.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the vehicle control method based on passenger comfort degree in any one of claims 1-5.

Images