Semi-active suspension multi-target control method and device based on vehicle-road cooperation
Through a deep reinforcement learning method based on vehicle-road collaboration, road surface and vehicle status information are obtained, weights are adjusted and suspension control current is generated, the vibration instability of semi-active suspension on complex road surfaces is solved, and the comfort of the vehicle is improved. Safety enables real-time interaction between vehicles and road surfaces and multi-objective optimization.
Patent Information
- Application Number
- CN202411152009.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-21
- Publication Date
- 2024-11-08
AI Technical Summary
The current semi-active suspension control method cannot be learned independently according to different situations and cannot adapt to complex and changeable road conditions, resulting in unstable vibration of vehicles on uneven roads, affecting riding comfort and safety.
Adopt deep reinforcement learning method based on vehicle-road collaboration, by obtaining road state information and vehicle state parameters, adjusting weights and inputting them to deep reinforcement learning model, generating suspension control current, adjusting suspension in real time to adapt to road changes, combined with multi-objective evaluation Model and Rainbow algorithm optimization control strategy.
Real-time interaction between the vehicle and the road surface is realized, the comfort, handling and safety of the vehicle is improved, the user experience is enhanced, the complex road surface conditions is adapted to, and the precision and efficiency of suspension control is improved.
Smart Images

Figure CN118906726A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a multi-objective control method and device for a semi-active suspension based on vehicle-road collaboration. Background Art
[0002] With the evolution of vehicle networking technology, autonomous driving has a new development form based on single-vehicle intelligence - vehicle-road cooperative autonomous driving. The fundamental development of this development is to improve the safety and comfort of passengers, which mainly depends on the control of the underlying dynamics of the vehicle. As an important part of the vehicle chassis, the vehicle suspension can be used to buffer the vehicle vibration caused by uneven roads and improve ride comfort and safety. When the vehicle encounters an uneven road during driving, the body will rise and fall with the road surface, causing the vehicle to bump, and the suspension may even hit the limit block. The uneven road surface will reduce the contact area between the road surface and the tire, thereby reducing the adhesion of the tire. In response to the above problems, passive suspension is less helpful, while semi-active suspension and active suspension can adjust their hardness and softness to reduce vehicle bumps, and can adjust the dynamic load of the tire within a certain range, which can improve the tire grip. Semi-active suspension consumes a small amount of energy to achieve variable damping control to improve vehicle dynamics. In the context of the development of automobile energy conservation and emission reduction, it has received widespread attention. In the precise regulation of vehicle dynamic control and ride comfort, control algorithms play a vital role, especially in the intelligent management of semi-active suspension systems. Scholars around the world have conducted in-depth research in this field and have proposed a series of innovative control strategies, such as classic PID control, forward-looking ceiling and floor shed control, and more adaptive sliding mode, LQG, fuzzy logic, model prediction and other diversified means. These methods have significantly reduced body vibration and significantly improved driving stability in their unique ways. In recent years, deep reinforcement learning, as a bright new star in the science and technology community, has demonstrated powerful learning and optimization capabilities. It generates massive data through environmental interaction and explores the most optimal decision-making plan based on it, and has shown outstanding results in many fields. Therefore, applying deep reinforcement learning to suspension control systems, especially optimizing suspension damping through self-learning, has become the focus of current scientific research.
[0003] However, current control methods do not have the ability to autonomously learn according to different situations, and thus cannot adapt to complex and changing road conditions. Summary of the invention
[0004] The purpose of the present invention is to solve the problem that the current control method does not have the ability to autonomously learn according to different situations and cannot adapt to complex and changeable road conditions. The present invention provides a semi-active suspension multi-objective control method, device and equipment based on vehicle-road collaboration.
[0005] The technical solution of the embodiment of the present application is implemented as follows: A first aspect of an embodiment of the present application provides a multi-objective control method for a semi-active suspension based on vehicle-road collaboration, comprising: Acquire current road surface status information and vehicle status parameter information; the road surface status information includes road surface type and road surface roughness value; the vehicle status parameter information includes vehicle body acceleration, suspension dynamic deflection, tire dynamic load, vehicle body roll angle and vehicle body pitch angle; Adjusting the weight of each parameter in the vehicle state parameter information based on the road state information to obtain a target weight corresponding to each of the vehicle state parameter information; The road surface state information, the vehicle state parameter information and the target weight are input into a deep reinforcement learning model to generate a suspension control current; the suspension control current is used to adjust the vehicle suspension in real time.
[0006] Optionally, the obtaining of current road state information and vehicle state parameter information includes: Obtain road surface image information through a binocular camera, and use a deep learning ResNet classifier to obtain road surface types, including concrete road surface, asphalt road surface, sandy soil road surface, gravel road surface, and crushed stone road surface; Point cloud information is obtained through a binocular camera, and the road surface roughness value is measured and obtained in combination with the SGBM semi-global stereo matching algorithm. The road surface roughness value includes the root mean square value of the roughness and the local peak value.
[0007] Optionally, after obtaining the current road state information and vehicle state parameter information, the method includes: The missing data values and abnormal data values in the road surface state information are supplemented or deleted, and the data is discretized.
[0008] Optionally, adjusting the weight of each parameter in the vehicle state parameter information based on the expected index and the road state information to obtain a target weight corresponding to each of the vehicle state parameter information includes: Creating a multi-objective evaluation model; the multi-objective evaluation model uses road surface status information as a variable and constructs a functional relationship with vehicle status parameter weights as output; The expected index and the road state information are input into the multi-objective evaluation model to determine the target weight corresponding to each of the vehicle state parameter information; the expected index includes at least one of a comfort index and a safety index.
[0009] Optionally, before inputting the road state information, the vehicle state parameter information and the target weight into the deep reinforcement learning model to generate the suspension control current, the method further includes: training the deep Q network using a Rainbow algorithm combined with a reward function of an expected indicator to obtain a deep reinforcement learning model; The reward function of the expected indicator is R, which is expressed as ; in, is the acceleration at the center of mass of the vehicle body, is the mean value of suspension dynamic deflection, is the mean dynamic load of the tire, is the pitch angle at the center of mass of the vehicle body, is the roll angle at the center of mass of the vehicle body, The suspension dynamic deflection when the limit is exceeded. ~ are the reward weights of these six state spaces respectively.
[0010] Optionally, after inputting the road state information, the vehicle state parameter information and the target weight into the deep reinforcement learning model to generate a suspension control current, the method further includes: performing current limiting processing on the suspension control current.
[0011] Optionally, after inputting the road state information, the vehicle state parameter information and the target weight into the deep reinforcement learning model to generate the suspension control current, the method further includes: migrating and deploying the deep reinforcement learning model to the target vehicle.
[0012] A second aspect of the embodiment of the present application provides a multi-objective control device for a semi-active suspension based on vehicle-road collaboration, comprising: an acquisition module, an adjustment module and a generation module, wherein: The acquisition module is configured to acquire current road surface state information and vehicle state parameter information; the road surface state information includes road surface type and road surface roughness value; the vehicle state parameter information includes vehicle body acceleration, suspension dynamic deflection, tire dynamic load, vehicle body roll angle and vehicle body pitch angle; The adjustment module is configured to adjust the weight of each parameter in the vehicle state parameter information based on the expected index and the road state information to obtain a target weight corresponding to each of the vehicle state parameter information; The generation module is configured to input the road state information, the vehicle state parameter information and the target weight into a deep reinforcement learning model to generate a suspension control current; the suspension control current is used to adjust the vehicle suspension in real time.
[0013] A third aspect of an embodiment of the present application provides an electronic device, comprising a processor and a memory; the memory stores a computer program, wherein the computer program, when executed by the processor, implements the multi-objective control method of the semi-active suspension based on vehicle-road collaboration described in the first aspect.
[0014] A fourth aspect of an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in the first aspect.
[0015] Compared with the prior art, the technical solution provided by this application has the following beneficial effects: The present invention provides a semi-active suspension multi-objective control method, device and equipment based on vehicle-road collaboration, which obtains the current road state information and vehicle state parameter information, adjusts the weight of each parameter in the vehicle state parameter information based on the road state information, and obtains the target weight corresponding to each vehicle state parameter information; the road state information, vehicle state parameter information and target weight are input into the deep reinforcement learning model to generate the suspension control current to adjust the vehicle suspension in real time. By introducing real-time road surface information into the suspension control, the vehicle and the road surface can interact in real time, the feedback of the road surface to the vehicle body is strengthened, and the vehicle-road collaboration is realized; in addition, multiple expected indicators are comprehensively considered to adjust the weight of each parameter in the vehicle state parameter information, and the comfort, handling stability and safety of the vehicle are weighed, which greatly improves the user experience. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 A flowchart of a multi-objective control method for a semi-active suspension based on vehicle-road collaboration provided in an embodiment of the present application; Figure 2 A schematic diagram of a suspension control process provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of a multi-objective control device for a semi-active suspension based on vehicle-road collaboration provided in an embodiment of the present application; Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0017] Below, embodiments of the present application will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present application. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present application. However, it is apparent that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present application. The terms used herein are only for describing specific embodiments and are not intended to limit the present application. The terms "include", "comprising", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components. All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.
[0018] Some block diagrams and / or flow charts are shown in the accompanying drawings. It should be understood that some blocks or combinations thereof in the block diagrams and / or flow charts may be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that these instructions, when executed by the processor, may create a device for implementing the functions / operations described in these block diagrams and / or flow charts.
[0019] In some embodiments, see Figure 1 , Figure 1 A flow chart of a multi-objective control method for a semi-active suspension based on vehicle-road collaboration provided in an embodiment of the present application; The multi-objective control method for a semi-active suspension based on vehicle-road collaboration provided in an embodiment of the present application includes: S110, obtaining current road surface status information and vehicle status parameter information; the road surface status information includes road surface type and road surface roughness value; the vehicle status parameter information includes vehicle body acceleration, suspension dynamic deflection, tire dynamic load, vehicle body roll angle and vehicle body pitch angle.
[0020] Here, the current road surface status information can be obtained in real time through sensors such as binocular cameras installed on the top of the vehicle, and the road surface status information can include unevenness information. In addition, body acceleration, suspension dynamic deflection, tire dynamic load, body roll angle, body pitch angle and other parameters can be measured through sensors such as inertial measurement units and force sensors installed on the vehicle body. It should be noted that the number and installation positions of sensors are not fixed and can be flexibly adjusted according to measurement requirements.
[0021] In some embodiments, S110, obtaining current road surface status information and vehicle status parameter information includes: Obtain road surface image information through a binocular camera, and use the deep learning ResNet classifier to obtain the road surface type, which includes concrete road surface, asphalt road surface, sandy soil road surface, gravel road surface and crushed stone road surface; The point cloud information is obtained by using a binocular camera, and the road roughness value is measured and obtained in combination with the SGBM semi-global stereo matching algorithm. The road roughness value includes the root mean square value of the roughness and the local peak value.
[0022] In some embodiments, after obtaining current road surface status information and vehicle status parameter information at S110, the method includes: The missing data values and data outliers in the road condition information are supplemented or deleted, and the data is discretized.
[0023] In this embodiment, after obtaining the current road state information and vehicle state parameter information, these data need to be cleaned. The data cleaning method includes processing the missing data values and data outliers in the road state information. First, for the missing data values, a supplement and deletion strategy is adopted. The missing data values are supplemented by interpolation methods including mean, median, nearest neighbor interpolation, etc. to fill in the missing data points to maintain data integrity. If there are too many missing values or they are not suitable for filling, these missing values are deleted. Secondly, for data outliers, it can be achieved by correcting them to reasonable values or deleting them as appropriate. In addition, data discretization is also part of data cleaning, especially for continuous variables. Discretization can convert continuous data into a limited number of discrete categories or intervals, which helps to simplify the analysis and modeling process, while reducing the noise and complexity of the data and improving the interpretability and comprehensibility of the data.
[0024] S120, adjusting the weight of each parameter in the vehicle state parameter information based on the expected index and the road state information to obtain a target weight corresponding to each vehicle state parameter information. Different road surface status information has different effects on parameters such as vehicle acceleration, suspension dynamic deflection, tire dynamic load, vehicle roll angle, and vehicle pitch angle. Expected indicators can be comfort indicators, stability indicators, and safety indicators. The expected indicators and road surface status information are used to adjust the weights of each parameter to obtain the target weights and formulate a multi-objective decision-making strategy.
[0025] In some embodiments, S120, adjusting the weight of each parameter in the vehicle state parameter information based on the expected index and the road state information to obtain the target weight corresponding to each vehicle state parameter information includes: Create a multi-objective evaluation model; the multi-objective evaluation model takes road condition information as a variable and constructs a functional relationship with vehicle condition parameter weights as output; The expected index and road condition information are input into a multi-objective evaluation model to determine the target weight corresponding to each vehicle condition parameter information; the expected index includes at least one of a comfort index and a safety index.
[0026] In this embodiment, the road surface state information includes the road surface type and the road surface roughness value, wherein the road surface type is divided into paved roads (concrete roads and asphalt roads) and unpaved roads (sand roads, gravel roads and crushed stone roads, etc.) according to the construction material. The road surface roughness of various roads is different, and different vibration effects are produced on the vehicle suspension. For example, since the road surface roughness of concrete roads and asphalt roads is small, the vibration and posture deflection produced on the vehicle are small. When weighing comfort and safety, the strategy of the multi-objective decision model is set to take comfort as the goal, then the body acceleration in the evaluation index will be given a higher weight at this time; while slightly rugged roads such as gravel and crushed stone not only bring greater vibration and posture deflection to the vehicle, but also reduce the contact area between the tire and the road surface, thereby reducing the tire adhesion. In order to ensure the adhesion performance of the tire and the stability of the vehicle body, the strategy of the multi-objective evaluation model here is set to take safety as the goal, and the tire dynamic load, body roll angle and body pitch angle in the parameters will be given a higher weight.
[0027] In some embodiments, at S130, before inputting the road state information, the vehicle state parameter information and the target weight into the deep reinforcement learning model to generate the suspension control current, the method further includes: training the deep Q network using the Rainbow algorithm in combination with the reward function of the expected index to obtain the deep reinforcement learning model; The reward function of the expected indicator is R, and its expression is ; in, is the acceleration at the center of mass of the vehicle body, is the mean value of suspension dynamic deflection, is the mean dynamic load of the tire, is the pitch angle at the center of mass of the vehicle body, is the roll angle at the center of mass of the vehicle body, The suspension dynamic deflection when the limit is exceeded. ~ are the reward weights of these six state spaces respectively.
[0028] In this embodiment, when the road surface is asphalt or cement, will be given a larger weight, and the specific assignment can be combined with expert experience and multiple training tests; when the road surface is muddy and off-road, ~ A larger weight will be assigned, and the specific value can be combined with expert experience and multiple training tests. To ensure vehicle safety and prevent the suspension from colliding with the limit block, set When the suspension dynamic deflection exceeds the limit threshold, the current round of training is stopped.
[0029] In one example, a functional relationship related to the target weight is constructed by combining the road surface type and the road surface roughness value.
[0030] ; The pavement correlation coefficients are concrete pavement, asphalt pavement, sandy soil pavement, gravel pavement and crushed stone pavement respectively; the pavement roughness value needs to take into account the root mean square value of the roughness With local peak , m and n are the weight coefficients of the two respectively; is the initial weight of each state parameter of the vehicle.
[0031] S130, inputting road state information, vehicle state parameter information and target weight into a deep reinforcement learning model to generate a suspension control current; the suspension control current is used to adjust the vehicle suspension in real time.
[0032] The control current can adjust the damping of the semi-active suspension and act on the vehicle. When the vehicle is driving on the road, the road excitation acts on both the vehicle and the suspension. Therefore, the vehicle and the road interact with each other. The vehicle information and the road information can be used at the same time to optimize the vehicle and suspension performance and realize suspension control based on vehicle-road collaboration.
[0033] In an example, see Figure 2 , Figure 2 A schematic diagram of the suspension control process provided for an embodiment of the present application; wherein, the process of training a deep Q network may be: initializing the main network and the target network, and the parameters of the two networks are the same initially. Initialize the experience replay buffer for storing experience tuples; initialize specific components including priority experience replay, multi-step learning parameters, etc. Select an action based on the selected reward function and the current strategy, observe the reward and next state returned by the environment, and store the experience tuple (including the current state, current action, returned reward, and next state) in the experience replay buffer. Here, a batch of experience tuples can be randomly sampled for experience replay, from the experience replay buffer. On this basis, calculate the Q value of the target network and the Q value of the main network. If the agent is based on the state , choose actions according to the strategy , get rewarded , and enter the next state , the differential target is constructed by the main network and the target network to update the Q value, The Q value obtained for the target network, is the Q value obtained by the main network, and the difference between the two is used as the learning update target, then its expression is: ; Where γ is the discount factor, are the parameters of the main network, are the parameters of the target network.
[0034] Repeat the above steps until convergence or the preset number of training steps is reached to complete the network training.
[0035] In another example, the deep reinforcement learning control method is deployed on a real vehicle by using a Simulink virtual training simulation based on a seven-degree-of-freedom model of the whole vehicle. Specifically, the seven-degree-of-freedom model of the whole vehicle is built according to the following formula: Assume the car body is a rigid body and the center of mass moves , , The values are all small. From the geometric relationship, the displacements on the springs at the four corners of the vehicle body can be obtained as follows: ; The vertical motion equation at the center of mass of the vehicle body is: ; The vehicle body pitch motion equation is: ; The body roll motion equation is: ; The equations of motion for the four unsprung masses are: ; Among the above five formulas, , , , The displacement of the vehicle body at the four corners, front, back, left, and right. is the vertical displacement of the center of mass of the vehicle body, , is the distance from the center of mass to the front and rear axles, is the vehicle body pitch angle, is the body roll angle, , is the front and rear axle track width, For the body quality, , , , is the damping coefficient of the front, rear, left and right shock absorbers, , , , is the front, rear, left, and right spring stiffness, , , , is the displacement of the center of mass of the four front, rear, left and right wheels, , , , The four unsprung masses are front, rear, left, and right. , , , is the stiffness of the four tires, , , , Provide road excitation for the four tires, front, rear, left and right.
[0036] In an optional embodiment, the suspension control current is limited, and the vehicle-road cooperative technology in the vertical direction of the vehicle is verified after the current is limited to a safety threshold.
[0037] In an optional embodiment, the method further includes migrating and deploying the deep reinforcement learning model to the target vehicle.
[0038] In one example, in order to realize vehicle-road collaborative technology and meet its safety performance, a seven-degree-of-freedom model of the entire vehicle is used for Simulink virtual training simulation, and its training model is migrated to the actual vehicle to complete the vibration reduction task in the real scene.
[0039] Specifically, the binocular camera is first calibrated. The Zhang Zhengyou calibration method is used to obtain the internal and external parameters and distortion coefficients of the camera using OpenCV to achieve automatic calibration. Secondly, the point cloud area of interest in front of the wheel driving track is extracted in combination with the installation position of the binocular camera, and the point cloud is simplified to reduce the amount of calculation. Thirdly, the SGBM semi-global stereo matching algorithm is used to match the left and right camera images and calculate the parallax to extract the three-dimensional information of the road area of interest, supplement or delete the missing data values and data outliers in the three-dimensional information, and perform data discretization. Then, the virtual training model of deep reinforcement learning is transplanted to the actual vehicle suspension controller for verification. In order to ensure actual safety, the output suspension control current is limited to avoid hardware accidents caused by excessive or insufficient current to the vehicle and suspension. Finally, the entire system is time-aligned, combining the distance of the binocular camera to measure the roughness information of the road ahead and the vehicle speed, and the system delay time is obtained through experiments. By adjusting the suspension response time, the suspension can achieve the vibration reduction function when the vehicle drives to the measured road surface.
[0040] The embodiment of the present application obtains the current road state information and vehicle state parameter information, adjusts the weight of each parameter in the vehicle state parameter information based on the expected index and the road state information, and obtains the target weight corresponding to each vehicle state parameter information; the road state information, vehicle state parameter information and target weight are input into the deep reinforcement learning model to generate the suspension control current to adjust the vehicle suspension in real time. By introducing real-time road surface information into the suspension control, the vehicle and the road surface can interact in real time, the feedback of the road surface to the vehicle body is strengthened, and the vehicle-road collaboration is realized; in addition, multiple expected indexes are comprehensively considered to adjust the weight of each parameter in the vehicle state parameter information, and the comfort, handling stability and safety of the vehicle are weighed, which greatly improves the user experience.
[0041] In some embodiments, see Figure 3 , Figure 3 The schematic diagram of the structure of a multi-objective control device for a semi-active suspension based on vehicle-road collaboration provided in an embodiment of the present application. The multi-objective control device for a semi-active suspension based on vehicle-road collaboration provided in an embodiment of the present application 300 includes: an acquisition module 310, an adjustment module 320 and a generation module 330, wherein: The acquisition module 310 is configured to acquire current road surface state information and vehicle state parameter information; the road surface state information includes road surface type and road surface roughness value; the vehicle state parameter information includes vehicle body acceleration, suspension dynamic deflection, tire dynamic load, vehicle body roll angle and vehicle body pitch angle; An adjustment module 320 is configured to adjust the weight of each parameter in the vehicle state parameter information based on the expected index and the road state information to obtain a target weight corresponding to each vehicle state parameter information; The generation module 330 is configured to input road state information, vehicle state parameter information and target weight into the deep reinforcement learning model to generate a suspension control current; the suspension control current is used to adjust the vehicle suspension in real time.
[0042] In some embodiments, the acquisition module 310 is specifically configured as follows: Obtain road surface image information through a binocular camera, and use the deep learning ResNet classifier to obtain the road surface type, which includes concrete road surface, asphalt road surface, sandy soil road surface, gravel road surface and crushed stone road surface; The point cloud information is obtained by using a binocular camera, and the road roughness value is measured and obtained in combination with the SGBM semi-global stereo matching algorithm. The road roughness value includes the root mean square value of the roughness and the local peak value.
[0043] In some embodiments, a data cleaning module is also included; the data cleaning module is specifically configured as follows: The missing data values and data outliers in the road condition information are supplemented or deleted, and the data is discretized.
[0044] In some embodiments, the adjustment module 320 is specifically configured as follows: Create a multi-objective evaluation model; the multi-objective evaluation model takes road condition information as a variable and constructs a functional relationship with vehicle condition parameter weights as output; The expected index and road condition information are input into a multi-objective evaluation model to determine the target weight corresponding to each vehicle condition parameter information; the expected index includes at least one of a comfort index and a safety index.
[0045] In some embodiments, a training module is also included; the training module is specifically configured as follows: The Rainbow algorithm is combined with the reward function to train the deep Q network to obtain the deep reinforcement learning model of the expected indicators; The reward function of the expected indicator is R, and its expression is ; in, is the acceleration at the center of mass of the vehicle body, is the mean value of suspension dynamic deflection, is the mean dynamic load of the tire, is the pitch angle at the center of mass of the vehicle body, is the roll angle at the center of mass of the vehicle body, The suspension dynamic deflection when the limit is exceeded. ~ are the reward weights of these six state spaces respectively.
[0046] In some embodiments, a current limiting module is also included; the current limiting module is specifically configured to perform current limiting processing on the suspension control current.
[0047] In some embodiments, a migration module is also included; the migration module is specifically configured to migrate and deploy the deep reinforcement learning model to the target vehicle.
[0048] The multi-objective control device for semi-active suspension based on vehicle-road collaboration provided in the embodiment of the present application can realize each process in the embodiment corresponding to the multi-objective control method for semi-active suspension based on vehicle-road collaboration mentioned above. To avoid repetition, they will not be described here.
[0049] It should be noted that the semi-active suspension multi-objective control device based on vehicle-road collaboration provided in the embodiment of the present application and the semi-active suspension multi-objective control method based on vehicle-road collaboration provided in the embodiment of the present application are based on the same application concept. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned semi-active suspension multi-objective control method based on vehicle-road collaboration, and the repeated parts will not be repeated.
[0050] In some embodiments, see Figure 4 , Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. An electronic device 400 provided in an embodiment of the present application includes a processor 410 and a memory 420; the memory 420 stores a computer program, wherein the computer program implements the above-mentioned multi-objective control method of semi-active suspension based on vehicle-road collaboration when executed by the processor. Specifically, the processor 410 may include, for example, a general-purpose microprocessor, an instruction set processor and / or a related chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 410 may also include an onboard memory for cache purposes. The processor 410 may be a single processing unit or multiple processing units for executing different actions of the method flow according to the embodiment of the present application. The memory 420 may be any medium capable of containing, storing, conveying, propagating or transmitting instructions. For example, the memory 420 may include, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, device, component or propagation medium. Specific examples of the memory 420 include: a magnetic storage device, such as a magnetic tape or a hard disk (HDD); an optical storage device, such as a compact disk (CD-ROM); a random access memory (RAM) or a flash memory; and / or a wired / wireless communication link. The present application also provides a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned multi-objective control method for semi-active suspension based on vehicle-road collaboration. The computer-readable medium may be included in the device / apparatus / system described in the above-mentioned embodiment; or it may exist independently without being assembled into the device / apparatus / system. The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed, the method according to the embodiment of the present application is implemented. According to an embodiment of the present application, a computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium or any combination of the two. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a computer-readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wired, optical cable, radio frequency signal, etc., or any suitable combination of the above. Those skilled in the art will appreciate that the features described in the various embodiments and / or claims of the present application may be combined in a variety of ways and / or combinations, even if such combinations or combinations are not explicitly described in the present application. In particular, without departing from the spirit and teachings of the present application, the features described in the various embodiments and / or claims of the present application may be combined in a variety of ways and / or combinations. All of these combinations and / or combinations fall within the scope of the present application. Therefore, the scope of the present application should not be limited to the above-described embodiments, but should be determined not only by the appended claims, but also by the equivalents of the appended claims.
Claims
1. A multi-objective control method for semi-active suspension based on vehicle-road collaboration, characterized in that: include: Acquire current road surface status information and vehicle status parameter information; the road surface status information includes road surface type and road surface roughness value; the vehicle status parameter information includes vehicle body acceleration, suspension dynamic deflection, tire dynamic load, vehicle body roll angle and vehicle body pitch angle; Adjusting the weight of each parameter in the vehicle state parameter information based on the expected index and the road state information to obtain a target weight corresponding to each of the vehicle state parameter information; Inputting the road surface state information, the vehicle state parameter information and the target weight into a deep reinforcement learning model to generate a suspension control current; The suspension control current is used to make real-time adjustments to the vehicle suspension.
2. The multi-objective control method for semi-active suspension based on vehicle-road collaboration according to claim 1 is characterized in that: The obtaining of current road state information and vehicle state parameter information includes: Obtain road surface image information through a binocular camera, and use a deep learning ResNet classifier to obtain road surface types, including concrete road surface, asphalt road surface, sandy soil road surface, gravel road surface, and crushed stone road surface; Point cloud information is obtained through a binocular camera, and the road surface roughness value is measured and obtained in combination with the SGBM semi-global stereo matching algorithm. The road surface roughness value includes the root mean square value of the roughness and the local peak value.
3. The multi-objective control method for semi-active suspension based on vehicle-road collaboration according to claim 1 is characterized in that: After obtaining the current road state information and vehicle state parameter information, the method includes: The missing data values and abnormal data values in the road surface state information are supplemented or deleted, and the data is discretized.
4. The multi-objective control method for semi-active suspension based on vehicle-road collaboration according to claim 1, characterized in that: The step of adjusting the weight of each parameter in the vehicle state parameter information based on the expected index and the road state information to obtain the target weight corresponding to each of the vehicle state parameter information includes: Creating a multi-objective evaluation model; the multi-objective evaluation model uses road surface status information as a variable and constructs a functional relationship with vehicle status parameter weights as output; The expected index and the road state information are input into the multi-objective evaluation model to determine the target weight corresponding to each of the vehicle state parameter information; the expected index includes at least one of a comfort index and a safety index.
5. The multi-objective control method for semi-active suspension based on vehicle-road collaboration according to claim 4 is characterized in that: Before inputting the road state information, the vehicle state parameter information and the target weight into the deep reinforcement learning model to generate the suspension control current, the method further includes: using the Rainbow algorithm in combination with the reward function of the expected index to train the deep Q network to obtain the deep reinforcement learning model; The reward function of the expected indicator is R, which is expressed as ; in, is the acceleration at the center of mass of the vehicle body, is the mean value of suspension dynamic deflection, is the mean dynamic load of the tire, is the pitch angle at the center of mass of the vehicle body, is the roll angle at the center of mass of the vehicle body, The suspension dynamic deflection when the limit is exceeded. ~ are the reward weights of these six state spaces respectively.
6. The multi-objective control method for semi-active suspension based on vehicle-road collaboration according to claim 1 is characterized in that: After the road state information, the vehicle state parameter information and the target weight are input into the deep reinforcement learning model to generate a suspension control current, the method further includes: performing current limiting processing on the suspension control current.
7. The multi-objective control method for semi-active suspension based on vehicle-road collaboration according to claim 1 is characterized in that: After inputting the road state information, the vehicle state parameter information and the target weight into the deep reinforcement learning model to generate the suspension control current, the method further includes: migrating and deploying the deep reinforcement learning model to the target vehicle.
8. A semi-active suspension multi-objective control device based on vehicle-road collaboration, characterized in that: include: Acquisition module, adjustment module and generation module, among which, The acquisition module is configured to acquire current road surface state information and vehicle state parameter information; the vehicle state parameter information includes vehicle body acceleration, suspension dynamic deflection, tire dynamic load, vehicle body roll angle and vehicle body pitch angle; The adjustment module is configured to adjust the weight of each parameter in the vehicle state parameter information based on the expected index and the road state information to obtain a target weight corresponding to each of the vehicle state parameter information; The generation module is configured to input the road state information, the vehicle state parameter information and the target weight into a deep reinforcement learning model to generate a suspension control current; the suspension control current is used to adjust the vehicle suspension in real time.
9. An electronic device comprising a processor and a memory; the memory stores a computer program, wherein: When the computer program is executed by the processor, the multi-objective control method of the semi-active suspension based on vehicle-road collaboration described in any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.
Citation Information
Cited By
Control method considering safety of intelligent suspension, medium and equipment
CN120269980A
Cooperative transportation method, control system and cooperative detection device
CN120636140A
