A vehicle active suspension control method considering passenger motion sickness based on road surface information

Abstract

The present application relates to a kind of vehicle active suspension control method based on road information considering passenger motion sickness, comprising the following steps: establishing car-seat active suspension coupling dynamics model;Random road information in the process of vehicle travel is identified using binocular vision recognition algorithm, and road information is classified, and road grade information is input into car-seat active suspension system model as excitation signal input using preview control algorithm;Using model predictive control (MPC) algorithm, the performance index of active suspension is optimized, the vertical acceleration value of active suspension system is reduced, the influence that vehicle suspension vertical acceleration generates vehicle vertical vibration and then causes passenger motion sickness occurs is fully considered, the key performance indicators of research are controlled respectively, the vertical acceleration size of vehicle is improved by improving algorithm, finally effectively improve the smoothness and steering stability of automatic driving vehicle, reduce the incidence of passenger motion sickness in vehicle.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent control of autonomous vehicles, and particularly relates to a vehicle active suspension control method considering passenger motion sickness based on road surface information. BACKGROUND

[0002] At present, with the development of intelligent vehicle technology, the automatic driving technology of the vehicle has been greatly improved, and the ride comfort and handling stability of the autonomous vehicle are also directly related to the promotion and application of the automatic driving technology. Since the role of the driver changes during the driving process of the autonomous vehicle, the passengers in the vehicle generally engage in a series of leisure activities during the driving process of the vehicle, however, the occurrence of passenger motion sickness during the driving process of the vehicle will cause the passengers to feel uncomfortable and the ride experience to be poor. Therefore, how to improve the stability of the autonomous vehicle and improve the ride comfort of the vehicle can effectively improve the motion sickness effect of the passengers and reduce the motion sickness rate, which is a problem to be solved in the current autonomous driving industry.

[0003] The development of the drive-by-wire chassis technology brings the possibility of improving the ride comfort and handling stability of the autonomous vehicle, and the drive-by-wire chassis technology is continuously promoted and becomes the standard configuration of more and more medium and high-grade vehicles. Based on the characteristics of the drive-by-wire chassis, the vehicle active suspension technology has also been widely applied. The active suspension can adaptively adjust the damping characteristics of the suspension based on the input of the road surface information, thereby improving the ride comfort of the vehicle. The occurrence of passenger motion sickness is not only affected by the lateral and longitudinal acceleration of the vehicle, but also affected by the vertical acceleration of the vehicle. If the vertical acceleration of the vehicle can be reduced during the driving process of the vehicle, the passenger motion sickness rate can be effectively reduced. Therefore, the use of the active suspension plays an important role in the improvement of the autonomous vehicle technology.

[0004] In the rapidly developing field of intelligent vehicles, vehicle-mounted sensors such as laser radars and binocular cameras are used to provide more abundant sensing information for the vehicle. The development of sensing technology enables intelligent vehicles to have their own "eyes" and accurately judge the surrounding things. The application of sensing technology plays an irreplaceable role in the development of autonomous vehicles. By using the vehicle driving road unevenness information perceived by the visual recognition technology in the design of the active suspension optimization algorithm, the suspension system preview control can be realized by fully utilizing the visual perception, thereby effectively improving the ride comfort and handling stability of the vehicle and increasing the comfort of the vehicle. At present, the intelligent vehicle cannot improve the ride comfort during the driving process based on the road surface characteristics, further improve the driving comfort of the vehicle, and increase the probability of the occurrence of passenger motion sickness. SUMMARY

[0005] The purpose of this invention is to provide a vehicle active suspension control method that takes into account passenger motion sickness based on road surface information. Addressing the issue of motion sickness in passengers of autonomous vehicles, this method uses binocular vision recognition technology to identify road surface information and inputs it into the vehicle-seat active suspension system. The onboard controller can adjust the damping characteristics of the vehicle-seat active suspension system in a timely manner based on road surface unevenness signals, ensuring optimal comfort when the vehicle travels on the corresponding road surface.

[0006] The technical solution adopted in this invention is as follows:

[0007] The present invention proposes a vehicle active suspension control method based on road surface information and considering occupant motion sickness, comprising the following steps:

[0008] S1. Establish a dynamic model of the vehicle-seat active suspension system; by analyzing the model, it can be concluded that under the input of road information, the active suspension can adjust the damping magnitude of the CDC shock absorber in the active suspension in real time according to different road information.

[0009] S2. Use a binocular camera above the front of the vehicle to extract road information, and use a binocular vision recognition algorithm to identify random road information during the vehicle's driving process. Then, classify the road information into levels and use a pre-aiming control algorithm to input the road level information into the vehicle-seat active suspension system model as an excitation signal input.

[0010] S3. Establish a motion sickness model system to perceive the vehicle's motion state, including the vehicle's motion excitation signals and the excitation signals from the seat suspension to the occupants' bodies.

[0011] S4. The Model Predictive Control (MPC) algorithm is used to perform rolling optimization on the performance indicators of the active suspension, reducing the vertical acceleration value of the active suspension system. The influence of the vertical acceleration of the vehicle suspension on the vertical vibration of the vehicle and the resulting motion sickness of the occupants is fully considered. The key performance indicators are controlled separately, and the magnitude of the vehicle's vertical acceleration is improved by improving the algorithm, which ultimately effectively improves the ride comfort and handling stability of the autonomous vehicle.

[0012] Furthermore, in step S1, the differential equations of the vehicle-seat active suspension system dynamics model are as follows:

[0013]

[0014]

[0015]

[0016] Where: m d Let m be the unsprung mass. cLet m be the sprung mass. b Total mass of people and vehicle seats, z g Z represents the input displacement of the road surface. b For the displacement of the seat, z c For the displacement of the vehicle body, z d k is the displacement of the tire. b k c k d These are the damping coefficients of the corresponding springs, and F1 and F2 are the control forces;

[0017] Rewriting the above differential equation as a state-space equation yields:

[0018]

[0019] We obtain the state vector z(t), state matrix A, control input U(t), control input matrix B, noise input matrix F, and Gaussian white noise W(t);

[0020]

[0021]

[0022] U(t)=[F1 F2] T W(t) = [w t ].

[0023] Furthermore, step S4 specifically includes: designing the MPC algorithm for the vehicle-seat active suspension system dynamic model derived in step S1; using the MPC algorithm to predict the future state of the controlled object using a discrete model; and obtaining the optimal control quantity by solving the optimization problem in the finite time domain, thereby discretizing the continuous vehicle active suspension dynamic equations.

[0024]

[0025] In the formula: T is the control step size; x(k|k) and ω(k|k) are the measured quantities at time K, representing the actual system state and road excitation at time k; y(k|k), u(k|k) and x(k+1|k) are the predicted quantities at time k, representing the predicted system output at time k, the control quantity at time k, and the system state at time k+1.

[0026] Furthermore, in order to ensure that the vehicle has good ride comfort and handling stability, the objective function of the MPC controller optimization problem is set as min J(y,u), that is, to minimize the output and damping force of the system and reduce the impact of road excitation on the human body.

[0027]

[0028] In the formula: Q and R are the weight matrices, respectively.

[0029] Furthermore, in step S2, during the extraction of road surface information by the binocular camera, a database is constructed using a large number of images containing different road surface information. Based on the VGGNet structure, the entire network uses the same size convolutional kernel and max pooling size. A VGG16 neural network is used to add labels to the images in the database and train the network. The temporal perturbation curves for different road conditions vary according to international standards and are described in the form of road surface PSD. The power spectrum of the road surface excitation is typically fitted using the following formula:

[0030]

[0031] Where: the spatial frequency is n(m) -1 The reference spatial frequency is n0; typically, n0 = 0.1 (m). -1 The road surface roughness coefficient is G. q (n0)(m 3 The frequency index is generally chosen to be 2.

[0032] Furthermore, the motion sickness model system includes an occupant body stimulation perception module, a vehicle state perception module, and a user ride comfort feedback module; the vehicle state perception module is used to perceive the vehicle's motion state, including the vehicle's motion stimulation signals; the occupant body stimulation perception module is used to perceive the stimulation signals from the seat suspension to the occupant's body; and the user ride comfort feedback module is used to adjust the vehicle's operating mode according to the occupant's subjective feelings and switch the vehicle suspension system's operating mode in a timely manner.

[0033] Compared with the prior art, the present invention has the following advantages:

[0034] This invention addresses the current limitations of intelligent vehicles in using real-time road information to improve ride smoothness and thus enhance passenger comfort. Compared to traditional methods for improving ride smoothness, this invention combines real-time road information with real-time optimization of the active suspension's vertical acceleration, effectively improving the actual comfort of occupants. Previous inventions mostly relied on front axle anti-aliasing and rear axle damping. This invention directly extracts road information from the front of the vehicle using visual recognition technology as the target input, achieving optimal damping after optimization and reducing the probability of motion sickness. Attached Figure Description

[0035] Figure 1 This is a schematic diagram illustrating the principle of the method of the present invention;

[0036] Figure 2 This is a schematic diagram illustrating the working principle of the vehicle-seat active suspension system in this invention;

[0037] Figure 3 This is a schematic diagram of a binocular camera collecting road surface information;

[0038] Figure 4 These are schematic diagrams of different levels of road surface excitation signals;

[0039] Figure 5 A diagram illustrating the process of motion sickness. Detailed Implementation

[0040] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0041] The present invention proposes a vehicle active suspension control method that considers occupant motion sickness based on road surface information, such as... Figures 1-5 As shown, the specific implementation process is as follows:

[0042] S1. Establish a dynamic model of the vehicle-seat active suspension system; the vehicle-seat active suspension system includes a vehicle mass module, a variable damping shock absorber (CDC), an unsprung mass module, a sprung mass module, sensors, an ECU electronic control unit, a drive motor, air springs, etc.

[0043] The differential equations of the dynamic model of the vehicle-seat active suspension system are as follows:

[0044]

[0045]

[0046]

[0047] Where: m d Let m be the unsprung mass. c Let m be the sprung mass. b Total mass of people and vehicle seats, z g Z represents the input displacement of the road surface. b For the displacement of the seat, z c For the displacement of the vehicle body, z d k is the displacement of the tire. b k c k d These are the damping coefficients of the corresponding springs, and F1 and F2 are the control forces;

[0048] Rewriting the above differential equation as a state-space equation yields:

[0049]

[0050] We obtain the state vector z(t), state matrix A, control input U(t), control input matrix B, noise input matrix F, and Gaussian white noise W(t);

[0051]

[0052]

[0053] U(t)=[F1 F2] T W(t) = [w t ].

[0054] This invention uses Newton's laws of motion to establish the differential equations of the dynamic model of the vehicle-seat active suspension system. To facilitate controller design, the differential equations are converted into state-space equations. The state-space equations are then discretized to design the corresponding model predictive controller, including key parameters such as control step size and prediction step size. Finally, the control objective of the vehicle-seat active suspension system is optimized by rolling, with reducing the vehicle's vertical acceleration as the primary performance indicator.

[0055] Analysis of this model reveals that, given road surface information, the active suspension can adjust the damping of the CDC shock absorbers in real time according to different road conditions. When the vehicle encounters speed bumps or potholes, the suspension can quickly restore the vehicle's traction. When driving on long inclines or roads with significant curvature, the chassis controls wheel slippage, reducing peak lateral and longitudinal acceleration. Since motion sickness is caused by large variations in lateral, longitudinal, and vertical acceleration during vehicle movement, with acceleration frequencies within a certain range, the active suspension enhances vehicle stability and effectively reduces motion sickness.

[0056] S2. Use a binocular camera above the front of the vehicle to extract road information, and use a binocular vision recognition algorithm to identify random road information during the vehicle's driving process. Then, classify the road information into levels and use a pre-aiming control algorithm to input the road level information into the vehicle-seat active suspension system model as an excitation signal input.

[0057] In the process of extracting road surface information using a binocular camera, a database is constructed using a large number of images containing different road surface information. Based on the VGGNet structure, the entire network uses the same size convolutional kernel and max pooling size. A VGG16 neural network is used to add labels to the images in the database and train the network. The temporal perturbation curves for different road conditions vary according to international standards and are described in the form of road surface PSD. The power spectrum of the road surface excitation is typically fitted using the following formula:

[0058]

[0059] Where: the spatial frequency is n(m) -1 The reference spatial frequency is n0; typically, n0 = 0.1 (m). -1 The road surface roughness coefficient is G. q (n0)(m 3 The frequency index is generally chosen to be 2.

[0060] The GB-7031-1986 standard classifies road surface roughness coefficients into eight levels, from A to H. Level A represents highways and their corresponding road conditions, indicating the best road condition and good vehicle passability. Level E represents unpaved roads, and Level H represents the worst road condition, with poor vehicle passability. During the road surface information training and extraction process, road surface images under different labels are clustered and analyzed. Based on various road surface conditions, they are divided into eight categories according to the national standard, ranging from best to worst road condition. Using road surface information from different classification levels and corresponding road surface excitation signals, the road surface excitation signals are input into the vehicle-seat active suspension system as system input information.

[0061] By employing image recognition technology and utilizing the vehicle's dual-lens cameras, data is extracted from the ground ahead of the vehicle, and the road surface grade is identified in real time. This road surface grade information is transmitted to the onboard controller, which then sends commands to adjust the suspension system stiffness in real time. This forms a pre-aiming control strategy in the vehicle's suspension control, effectively improving the vehicle's smoothness and comfort. The controller combines the road surface identification results with information such as vehicle speed and distance to calculate the system's activation timing, allowing the system to operate at more precise moments, raising or lowering the suspension, and increasing or decreasing damping, thereby achieving better comfort and handling.

[0062] Image recognition algorithms are used to train on various types of images in a road image dataset, with the VGG-16 algorithm employed to improve recognition accuracy. After a series of preliminary training steps, the algorithm achieves relatively accurate recognition of vehicle road types. By using an onboard binocular camera to identify different types of road surfaces, different levels of road surface information are clustered and analyzed, classifying all road surfaces according to their corresponding levels. The different road surface levels are then input to the onboard controller. During visual recognition, if the identified road surface level is A, this information is input in real-time to the vehicle-seat active suspension system, forming a suspension pre-aiming control strategy framework. Upon receiving the road excitation signal, the active suspension ECU of the autonomous vehicle can adjust the suspension damping stiffness online in real-time based on prior experience and road input information, thereby better navigating the road surface and improving the comfort and smoothness of the autonomous vehicle's ride.

[0063] S3. Establish a motion sickness model system to perceive the vehicle's motion state, including the vehicle's motion excitation signals and the excitation signals from the seat suspension to the occupants' bodies. The motion sickness model system mainly includes an occupant body excitation perception module, a vehicle state perception module, and a user ride comfort feedback module. The vehicle state perception module is used to perceive the vehicle's motion state, including the vehicle's motion excitation signals. The occupant body excitation perception module is used to perceive the excitation signals from the seat suspension to the occupants' bodies. The user ride comfort feedback module is used to adjust the vehicle's operating mode according to the occupants' subjective feelings and switch the vehicle suspension system's operating mode in a timely manner.

[0064] Since motion sickness is primarily caused by subjective vertical conflict, the Subjective Vertical Conflict (SVC) model mainly considers the passive movement of passengers, neglecting the influence of vestibular function and visual stimuli on motion sickness. If the vehicle's longitudinal and lateral movements are unstable during operation, and the vertical movement is affected by road surface input excitation signals, motion sickness will occur. If the human body is exposed to a vibrating environment for a prolonged period, motion sickness symptoms will gradually worsen, and the likelihood of vomiting will increase. Although the human body is slightly more sensitive to horizontal movement than to vertical and pitch movements, and the contribution of motion sickness is mainly due to horizontal vibration, vertical vibration also plays a significant role. This invention mainly considers how to reduce the occurrence of motion sickness symptoms in passengers by ensuring that the vertical excitation frequency is within the range where motion sickness occurs. Here, the Motion Sickness Dose Value (MSDV) is mainly used as an indicator of motion sickness occurrence, and the frequency with the greatest influence on MSDV, i.e., the highest weight, is approximately 0.16 Hz.

[0065]

[0066] Where: aw Represents the vibration-weighted acceleration, where T is not limited by time.

[0067] During vehicle operation, road surface excitation signals are continuously transmitted to passengers through the vehicle body. Based on the adaptive damping characteristics of the vehicle-seat active suspension system, passengers can select the suspension system operating model, such as the suspension stiffness, according to their subjective riding experience through the user ride comfort feedback module. If, according to the occupant body excitation perception module, the passenger's motion sickness dose value is higher than normal, the passenger's ride comfort is poor, and the MSDV value is high. The vehicle state perception module should then adjust the vehicle driving mode in real time to mitigate the occurrence of motion sickness.

[0068] During the operation of autonomous vehicles, occupants are typically seated inside the vehicle. Their bodies continuously receive stimulation signals transmitted from the wheels, which are then processed by the vestibular system in the brain. If the frequency of these stimulation signals falls within the range of human sensitivity, it can cause discomfort or even vomiting, resulting in a poor passenger experience. To overcome this problem, occupants can switch driving modes in real-time based on their subjective feelings, adjusting the damping characteristics of the vehicle-seat active suspension system accordingly. Various sensing modules within the vehicle can also calculate whether the motion sickness dose (MSDV) exceeds the normal tolerance level through a series of transmitted signals. If it does, the system feeds back to the vehicle controller via the vehicle's state perception module, allowing the vehicle to further adjust its operating state and reduce the MSDV.

[0069] S4. Model predictive control (MPC) algorithms are used to perform rolling optimization of active suspension performance indicators, reducing the vertical acceleration value of the active suspension system. The impact of vertical acceleration on vehicle suspension, leading to vertical vibration and potentially causing motion sickness in passengers, is fully considered. Key performance indicators are controlled separately, and the magnitude of the vehicle's vertical acceleration is improved through algorithm refinement, ultimately effectively improving the ride comfort and handling stability of autonomous vehicles. Specifically, this includes:

[0070] The MPC algorithm is used to design the dynamic model of the vehicle-seat active suspension system derived in step S1. The MPC algorithm uses a discrete model to predict the future state of the controlled object, and obtains the optimal control quantity by solving the optimization problem in the finite time domain. The continuous vehicle active suspension dynamic equations are discretized as follows:

[0071]

[0072] In the formula: A d =e AT , T is the control step size; x(k|k) and ω(k|k) are the measured quantities at time K, representing the actual system state and road excitation at time k; y(k|k), u(k|k) and x(k+1|k) are the predicted quantities at time k, representing the predicted system output at time k, the control quantity at time k, and the system state at time k+1.

[0073] To ensure good ride comfort and handling stability of the vehicle, the objective function of the MPC controller optimization problem is set as minJ(y,u), which means minimizing the system output and damping force to reduce the impact of road excitation on the human body.

[0074]

[0075] In the formula: Q and R are the weight matrices, respectively.

[0076] All matters not covered in this invention are common knowledge.

[0077] The specific embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A vehicle active suspension control method considering passenger motion sickness based on road surface information, characterized by: It comprises the following steps: S1, a vehicle-seat active suspension system dynamics model is established; by analyzing the model, it can be concluded that under the input of road information, the active suspension can adjust the damping size of the CDC damper in the active suspension in real time with different road information; S2, the road information is extracted above the front side of the vehicle using a binocular camera, and a binocular vision recognition algorithm is used to identify the random road information during vehicle driving, and the road information is classified, and the road classification information is input into the vehicle-seat active suspension system model as an excitation signal input using a preview control algorithm; S3, a motion sickness model system is established to perceive the vehicle motion state, including the vehicle motion excitation signal and the excitation signal of the seat suspension to the passenger body; S4, a model predictive control algorithm is used to rollingly optimize the performance indicators of the active suspension, reduce the vertical acceleration value of the active suspension system, fully consider the influence of the vertical acceleration of the vehicle suspension on the occurrence of passenger motion sickness caused by vehicle vertical vibration, control the key performance indicators respectively, and improve the vertical acceleration size of the vehicle through the improved algorithm, and finally effectively improve the smoothness and handling stability of the autonomous vehicle; In the step S1, the differential equation of the vehicle-seat active suspension system dynamics model is as follows: In the formula: is the unsprung mass, is the sprung mass, is the total mass of the person and the vehicle seat, is the input displacement of the road surface, is the displacement of the seat, is the displacement of the vehicle body, is the displacement of the tire, is the spring coefficient of the seat, is the spring coefficient of the vehicle suspension, is the spring coefficient of the tire, F1 is the input actuation control force of the vehicle suspension, and F2 is the input actuation control force of the seat suspension; is the damping coefficient of the vehicle suspension, is the damping coefficient of the seat suspension; The above differential equation can be written as a state space equation as follows: state vector , state matrix , control input , control input matrix , noise input matrix , gaussian white noise ; ， ， 。 2. The vehicle active suspension control method based on road information considering passenger motion sickness according to claim 1, characterized in that: In the step S2, during the road information extraction by the binocular camera, a database is established using pictures containing different road information, based on the VGGNet structure, the entire network uses the same size of convolution kernel and maximum pooling size, and the VGG16 neural network is used to add labels to the pictures in the database and train and learn; according to the different time domain disturbance curves of different road conditions in international standards, and described in the form of road PSD; the power spectrum of the road excitation is fitted by the following formula: where: the spatial frequency is n(m -1 ); the reference spatial frequency is n0; take n0=0.1(m -1 ); the road roughness coefficient is G q (n0)(m 3 ), and the frequency exponent is 2.

3. The vehicle active suspension control method based on road information considering passenger motion sickness according to claim 1, characterized in that: The motion sickness model system comprises a passenger body excitation perception module, a vehicle state perception module and a user ride comfort feedback module; the vehicle state perception module is used to perceive the motion state of the vehicle, including the motion excitation signal of the vehicle; the passenger body excitation perception module is used to perceive the excitation signal of the seat suspension to the passenger body; and the user ride comfort feedback module is used to adjust the mode of vehicle operation according to the subjective feeling of the passenger, and timely switch the operation mode of the vehicle suspension system.

Images