Interactive multiple model extremum seeking control method and system for vehicle damping

Abstract

The application provides an interactive multi-model extreme value search control method and system for vehicle damping, and the method comprises the following steps: acquiring vehicle running state data to obtain vehicle state parameters; establishing a system model based on a discrete space model, and setting vehicle motion state models in the system model for not less than two road uneven excitation conditions; combining an interactive multi-model (IMM) and an extreme value search control model (ESC) to construct and utilize an interactive multi-model extreme value search control (IMM-ESC) algorithm as a control strategy. The application solves the technical problems of high dependence on the number and precision of sensors, heavy system calculation burden caused by complex data processing of multi-sensor data fusion, restriction of vehicle damping effect, and high vehicle cost.

Description

Technical Field

[0001] The present invention relates to the fields of control engineering and information fusion, and in particular to an interactive multi-model extreme value search control method and system for vehicle shock absorption. Background Art

[0002] With increasing vehicle speeds and increasingly complex traffic environments, demands for vehicle safety are becoming increasingly stringent. Improving vehicle stability can better ensure safe driving on various road conditions. Numerous studies have shown that improving vehicle stability can effectively reduce traffic accidents, particularly in slippery road conditions or in emergency avoidance situations. Advances in sensor technology, electronic control units, and actuators provide the technical foundation for achieving improved vehicle stability. Improving vehicle stability is also linked to reducing energy consumption and emissions, as unstable vehicle driving can lead to unnecessary fuel consumption. Globalization requires automakers to consider road conditions and driving habits in different countries and regions, and improving vehicle stability can help accommodate these differences. Advanced sensors monitor vehicle status and automatically adjust braking and engine output when necessary to help drivers maintain vehicle stability and control. With technological advancements, these systems are becoming increasingly complex and intelligent, improving not only safety but also the driving experience. Vibration reduction can significantly enhance vehicle stability. Therefore, improving the effectiveness of vehicle vibration reduction has become a core research issue.

[0003] Traditional technologies typically control wheel dampers by collecting vehicle status information to achieve effective vibration reduction. However, this approach relies heavily on the number and accuracy of sensors, often resulting in high costs. Furthermore, multi-sensor data fusion involves complex data processing and synchronization, which can lead to excessive computational overhead for the system, making real-time vibration reduction control difficult.

[0004] The invention patent application document, "Vehicle Shock Absorber Control Method, Device, Storage Medium, and Vehicle," with publication number CN116638908A, describes a conventional method that collects vehicle data from a target vehicle to determine multiple parameters of the target vehicle's wheel shock absorbers. The method then calculates a target damping force using different weighting coefficients combined with the wheel shock absorber parameters, and controls the wheel shock absorbers using the target damping force. However, this conventional solution requires the collection of a large amount of vehicle data and relies heavily on the number of sensors. It also requires processing a large amount of data and may not provide real-time shock absorption control. Furthermore, this conventional technology focuses on vehicle data and neglects road condition data, potentially preventing the shock absorption system from responding optimally and effectively absorbing impacts and vibrations caused by the road surface.

[0005] The invention patent application document with publication number CN116653525A, "A method, device, on-board terminal and vehicle for optimizing the control of vehicle shock absorbers", states that the existing method realizes the mutual transmission of collected vehicle and road data through communication between the vehicle and the cloud, including upload and download functions. The vehicle shock absorber is controlled by combining vehicle data with road data. However, the aforementioned existing solution requires more sensors with higher precision and is overly dependent on the accuracy of the sensors. It also requires real-time processing of large amounts of data and may involve complex calculations, which places higher requirements on the performance of the on-board computing system and may increase the cost of the vehicle itself. At the same time, the aforementioned existing technology has high performance requirements for the communication module, and there may be delays or inaccuracies in data acquisition or transmission.

[0006] Patent application CN117382676A, "An Intelligent Driving Method and System Based on Multi-Sensor Data Fusion and Enhancement," combines data from various sensors, such as cameras and shock sensors, to adjust the vehicle's shock absorption parameters using road surface data, vehicle shock absorption data, and vehicle speed. This approach improves the comfort and safety of autonomous vehicles under various road conditions. However, the multi-sensor data fusion in this existing solution involves complex data processing and synchronization issues, which may increase the system's computational burden; its ability to resist external interference still needs to be improved.

[0007] In summary, existing technologies have technical problems such as high dependence on the number and accuracy of sensors, heavy system computing burden caused by processing complex data of multi-sensor data fusion, restricting the vehicle's shock absorption effect, and high vehicle cost. Summary of the Invention

[0008] The technical problem to be solved by the present invention is: how to solve the technical problems in the existing technology that are highly dependent on the number and accuracy of sensors, the processing of complex data of multi-sensor data fusion leads to a heavy system calculation burden, restricts the vehicle's shock absorption effect, and has high vehicle costs.

[0009] The present invention solves the above technical problems by adopting the following technical solutions: an interactive multi-model extreme value search control method for vehicle shock absorption includes:

[0010] S1. Obtain vehicle operating status data to obtain vehicle status parameters;

[0011] S2. Establishing a system model based on the discrete space model, and setting a vehicle motion state model in the system model for at least two road surface unevenness excitation scenarios, wherein the vehicle motion state model includes: a state estimation prediction equation, a covariance estimation prediction equation, a measurement equation, and an eight-degree-of-freedom vehicle model;

[0012] S3. Combining the interactive multi-model IMM and the extreme value search control model ESC, construct and utilize the interactive multi-model extreme value search control IMM-ESC algorithm as a control strategy, wherein the interactive multi-model extreme value search control (IMM-ESC) algorithm adopts no less than two vehicle motion state models to describe the vehicle driving motion state according to the vehicle state parameters; interactively update the probability and weight of the vehicle motion state model to obtain the multi-vehicle model state estimation results according to the vehicle driving operation state, and perform weighted averaging processing on them to obtain vehicle state estimation and covariance estimation; search for the system performance extreme point through system output measurement operation and control input design operation.

[0013] The present invention adopts an extreme value search control method, adjusting the controller parameters by optimizing performance indicators to gradually achieve adaptive regulation of the system. This method can find the optimal performance parameters of the vehicle by searching for the extreme value of the objective function. Even if the vehicle parameters change during driving, this method can still find new optimal performance parameters, thereby achieving the effect of optimizing the vehicle's shock absorption performance in real time. The present invention has low information requirements for the controlled object, which means that excessive vehicle information parameters are not required. Therefore, it can solve the problem of traditional control methods' dependence on the number and accuracy of sensors, reduce vehicle costs, and avoid the complex data processing involved in multi-sensor data fusion, reducing the system's computational burden.

[0014] In a more specific technical solution, in S2, the following logic is used to express the eight-degree-of-freedom vehicle model in the discrete space model:

[0015]

[0016] i=fl,fr,rl,rr

[0017] Where β(k) represents the vehicle sideslip angle at time k; γ(k) represents the vehicle yaw rate at time k; κ i (k) represents the tire slip rate of the four tires at time k; T s represents a fixed time step; F yf Indicates the lateral force on the vehicle's front tires; F yr Indicates the lateral force of the vehicle's rear tires; δ f (k) represents the active front steering angle; m represents the mass of the vehicle; v represents the vehicle speed; l f Indicates the distance from the vehicle's center of gravity to the front axle; l r Indicates the distance from the vehicle's center of gravity to the rear axle; I z Represents the vehicle's yaw moment of inertia; M z represents the vehicle yaw moment; r represents the tire radius; J represents the moment of inertia of each tire; C kiIndicates the longitudinal stiffness of the four tires; T i (k) represents the motor torque of each tire at time k.

[0018] In a more specific technical solution, in S2, the state estimation prediction equation is expressed using the following logic:

[0019] X j (k+1)=F j ·X j (k)+ω j (k),j=1,2,3,4

[0020] Using the following logic, the covariance estimation prediction equation is expressed as:

[0021]

[0022] Using the following logic, express the measurement equation:

[0023] Z j (k)=H j ·X j (k)+ν j (k),j=1,2,3,4

[0024] Where, X j (k+1) is the state vector of the vehicle under a certain model at time k+1, β j (k), γ j (k),κ j,fl (k),κ j,fr (k),κ j,rl (k), k j,rr (k) represents the vehicle sideslip angle, vehicle yaw rate, tire slip rate of the left front wheel, tire slip rate of the right front wheel, tire slip rate of the left rear wheel and tire slip rate of the right rear wheel at the time k under the jth model; F j is the state transition matrix; ω j (k) is the process noise with zero mean; P j (k) represents the covariance of vehicle state information; Q j (k) is the process noise ω j The covariance matrix of (k); Z j (k) is the observation data at time k Z j1 (k), Z j2 (k), Z j3 (k), Z j4 (k), Z j5 (k), Z j6(k) represents the vehicle sideslip angle, vehicle yaw rate, tire slip rate of the left front wheel, tire slip rate of the right front wheel, tire slip rate of the left rear wheel and tire slip rate of the right rear wheel of the vehicle measured at time k respectively; H j is the observation matrix of the jth model; v j (k) is the Gaussian observation noise.

[0025] In more specific technical solutions, S3 includes:

[0026] S31. Obtain vehicle state parameters of the vehicle at the previous moment, input the vehicle state parameters into different vehicle motion state models for input interaction, and obtain interaction values;

[0027] S32, inputting the interaction value into the Kalman filter corresponding to each vehicle motion state model, performing parallel filtering processing, and obtaining a filtering estimation result;

[0028] S33. Update the probability of each vehicle motion state model according to the maximum likelihood function of each filter and the Markov state transfer matrix to obtain a vehicle model probability update value;

[0029] S34, performing output interaction operations on the filter estimation results corresponding to each Kalman filter predictor and the vehicle model probability update value of each vehicle motion state model to obtain a predicted value of the vehicle state information at the next moment;

[0030] S35, using the predicted value of the vehicle state information at the next moment as the input value of the stabilizing steering stabilizer, wherein the current vehicle state, the optimal covariance estimate, and the vehicle model probability update value output by the Kalman filter predictor are calculated to obtain a vehicle state estimate and a vehicle covariance estimate;

[0031] S36, selecting a controller input value correlation index as an objective function to represent the current vehicle shock absorption information;

[0032] S37, using extreme value search control to periodically adjust input values ​​based on current vehicle shock absorption information to search for an extreme value of the objective function;

[0033] S38. When the objective function reaches an extreme value, the optimal vehicle state information parameters are obtained based on the extreme value point searched at this time.

[0034] The present invention designs a control method combining interactive multi-model (IMM) and extreme value search control (ESC), which can improve the shock absorption performance of the vehicle during driving by optimizing the performance parameters of the vehicle.

[0035] In a more specific technical solution, in S31, the vehicle motion state model is used to process the vehicle state parameters to obtain the current moment observation data;

[0036] The state information is input into the four vehicle motion state models respectively, and the input model interaction is performed. According to the covariance state estimation of different models and the latest measurement value z of the initial model k , the vehicle model is reinitialized using the Markov transformation matrix between different models;

[0037] in, and m,j=1,2,3,4 are the initial state estimates and covariance matrices of each sub-model of the filter input at time k-1. The following logic is used to reinitialize the j-th vehicle motion state model at the k-th time input:

[0038]

[0039] Note that q j (k-1) is the latest probability of the j-th model at time k-1, and the results are as follows:

[0040]

[0041] Here, is the mixing probability, and its calculation formula is

[0042] Obtaining the initial values ​​of the vehicle motion model after interaction, including: the initial values ​​of vehicle state estimation and covariance estimation;

[0043] The interaction value is obtained based on the current observation data and the initial value of the vehicle motion model after the interaction.

[0044] During driving, a vehicle may traverse a variety of different roads. Because a traditional single road model cannot cover all road conditions, the present invention utilizes an IMM, which includes four road models with different road surface grades: A, B, C, and D. By interactively integrating these four road models, the vehicle can adapt to more complex road conditions. Extreme value search control is then used to find the extreme value of the objective function, obtaining the optimal vehicle performance parameters at the current moment. These parameters are then input into the vehicle's stable steering controller to achieve the best possible shock absorption effect. Due to the real-time optimization characteristics of extreme value search control, it can always obtain the optimal vehicle performance parameters regardless of changes in road conditions. Therefore, this method can effectively cope with variable road conditions, such as bumps and uneven surfaces, while ensuring vehicle stability during driving.

[0045] In a more specific technical solution, in S32, the following logic is used to calculate the Kalman filter variable residual based on the current observation data and state estimation data of the vehicle motion state model:

[0046]

[0047] Where, Represents the variable residual in Kalman filtering;

[0048] Using the following logic, we construct an observation matrix based on the current observation data, and use the covariance estimate and the observation matrix to calculate the Kalman filter mean square error:

[0049]

[0050] Where S j (k) represents the mean square error in Kalman filtering, and R represents the measurement noise;

[0051] According to the following logic, the Kalman gain is obtained using the covariance estimate, the observation matrix, and the Kalman filter mean square error:

[0052]

[0053] Where K j (k) represents the Kalman gain;

[0054] The following logic is used to combine the state estimation prediction equation, Kalman gain, and Kalman filter variable residual to obtain the updated value of the vehicle state estimate at the current moment:

[0055]

[0056] The following logic is used to combine the prediction covariance matrix, Kalman gain, and mean square error to obtain the covariance update value:

[0057]

[0058] In a more specific technical solution, in S33, for the j-th vehicle motion state model, at the k-th time, a likelihood function matching operation is performed:

[0059]

[0060] Obtain the maximum likelihood function, use the following logic to express the Markov transfer matrix, and calculate the probability of each vehicle motion state model at the current moment:

[0061]

[0062] Where p mjis the transition probability from the mth model to the jth model;

[0063] Using the following logic, we can find the j-th vehicle motion state model and the vehicle model probability update value q at the k-th moment: j (k):

[0064]

[0065] Where c is the normalization factor.

[0066] In a more specific technical solution, in S35, the vehicle state estimation is expressed using the following logic:

[0067]

[0068] The vehicle covariance estimate is expressed using the following logic:

[0069]

[0070] Perform the next recursion, use the output value of the current time as the interactive input value, and continue to perform new recursive output operations.

[0071] In a more specific technical solution, in S36, an objective function is used to set an input value and an ideal value, and a mean square error is calculated based on the input value and the ideal value to serve as the objective function;

[0072] When the controller input value associated indicators have constraints, the constrained optimization method is used to integrate all constraints into the objective function.

[0073] In a more specific technical solution, the interactive multi-model extreme value search control system for vehicle shock absorption includes:

[0074] The vehicle status acquisition module is used to obtain vehicle operating status data to obtain vehicle status parameters;

[0075] A system model establishment module is used to establish a system model based on a discrete space model, and to set a vehicle motion state model in the system model for at least two road surface unevenness excitation scenarios, wherein the vehicle motion state model includes: a state estimation prediction equation, a covariance estimation prediction equation, a measurement equation, and an eight-degree-of-freedom vehicle model;

[0076] The multi-vehicle model state estimation module is used to combine the interactive multi-model IMM and the extreme value search control model ESC to construct and utilize the interactive multi-model extreme value search control IMM-ESC algorithm as a control strategy. Among them, the interactive multi-model extreme value search control (IMM-ESC) algorithm adopts no less than two vehicle motion state models to describe the vehicle driving motion state according to the vehicle state parameters; the probability and weight of the vehicle motion state model are interactively updated to obtain the multi-vehicle model state estimation results according to the vehicle driving operation state, and weighted average processing is performed based on the result to obtain vehicle state estimation and covariance estimation; through system output measurement operation and control input design operation, the system performance extreme point is searched, and the multi-vehicle model state estimation module is connected with the system model establishment module and the vehicle state acquisition module.

[0077] Compared with the prior art, the present invention has the following advantages:

[0078] The present invention adopts an extreme value search control method, adjusting the controller parameters by optimizing performance indicators to gradually achieve adaptive regulation of the system. This method can find the optimal performance parameters of the vehicle by searching for the extreme value of the objective function. Even if the vehicle parameters change during driving, this method can still find new optimal performance parameters, thereby achieving the effect of optimizing the vehicle's shock absorption performance in real time. The present invention has low information requirements for the controlled object, which means that excessive vehicle information parameters are not required. Therefore, it can solve the problem of traditional control methods' dependence on the number and accuracy of sensors, reduce vehicle costs, and avoid the complex data processing involved in multi-sensor data fusion, reducing the system's computational burden.

[0079] The present invention designs a control method combining interactive multi-model (IMM) and extreme value search control (ESC), which can improve the shock absorption performance of the vehicle during driving by optimizing the performance parameters of the vehicle.

[0080] During driving, a vehicle may traverse a variety of different roads. Because a traditional single road model cannot cover all road conditions, the present invention utilizes an IMM, which includes four road models with different road surface grades: A, B, C, and D. By interactively integrating these four road models, the vehicle can adapt to more complex road conditions. Extreme value search control is then used to find the extreme value of the objective function, obtaining the optimal vehicle performance parameters at the current moment. These parameters are then input into the vehicle's stable steering controller to achieve the best possible shock absorption effect. Due to the real-time optimization characteristics of extreme value search control, it can always obtain the optimal vehicle performance parameters regardless of changes in road conditions. Therefore, this method can effectively cope with variable road conditions, such as bumps and uneven surfaces, while ensuring vehicle stability during driving.

[0081] The present invention solves the technical problems existing in the prior art, such as high dependence on the number and accuracy of sensors, heavy system calculation burden caused by processing complex data of multi-sensor data fusion, which restricts the shock absorption effect of the vehicle, and high vehicle cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] Figure 1 Schematic diagram of the basic steps of the interactive multi-model extreme value search control method for vehicle shock absorption according to embodiment 1 of the present invention;

[0083] Figure 2 Schematic diagram of data flow processing of the IMM-ESC algorithm according to Example 1 of the present invention;

[0084] Figure 3 Schematic diagram of the specific steps of designing and applying the interactive multi-model extreme value search control algorithm according to embodiment 1 of the present invention;

[0085] Figure 4 Schematic diagram of the structural advantages of the interactive multi-model extreme value search control method for vehicle shock absorption according to embodiment 2 of the present invention;

[0086] Figure 5 This is a functional diagram of the interactive multi-model extreme value search control method for vehicle shock absorption according to embodiment 2 of the present invention. DETAILED DESCRIPTION

[0087] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0088] Example 1

[0089] like Figure 1 As shown, the interactive multi-model extreme value search control method for vehicle shock absorption provided by the present invention includes the following basic steps:

[0090] S1. Obtain vehicle data;

[0091] In this embodiment, the lateral force on the vehicle's four tires, the vehicle's speed during travel, the distance from the vehicle's center of gravity to the front and rear axles, the vehicle's yaw moment of inertia, the vehicle's yaw moment, the tire radius, the longitudinal stiffness of each tire, and the motor torque are measured. Parameters obtained through calculation in this embodiment include, but are not limited to, the vehicle's sideslip angle, the vehicle's yaw rate, and the slip rates of the four tires.

[0092] S2. Establish system model and initialize parameters;

[0093] In this embodiment, the aforementioned system model is in the form of a discrete space model, and four vehicle models are set for four road unevenness excitations. In this embodiment, the motion model includes a state estimation equation, a covariance estimation equation, and a measurement equation, where:

[0094] The vehicle model based on eight degrees of freedom is as follows:

[0095]

[0096]

[0097] i=fl,fr,rl,rr

[0098] Among them, β(k) represents the vehicle sideslip angle at time k; γ(k) represents the vehicle yaw rate at time k; k i (k) represents the tire slip rate of the four tires at time k; T s represents a fixed time step; F yf Indicates the lateral force of the vehicle's front tires; F yr Indicates the lateral force of the vehicle's rear tires; δ f (k) represents the active front steering angle; m represents the mass of the vehicle; v represents the vehicle speed; l f Indicates the distance from the vehicle's center of gravity to the front axle; l r Indicates the distance from the vehicle's center of gravity to the rear axle; I z Represents the vehicle's yaw moment of inertia; M z represents the vehicle yaw moment; r represents the tire radius; J represents the moment of inertia of each tire; C ki Indicates the longitudinal stiffness of the four tires; T i (k) represents the motor torque of each tire at time k.

[0099] In this embodiment, the state estimation prediction equation is as follows:

[0100] X j (k+1)=F j ·X j (k)+ω j (k),j=1,2,3,4

[0101] In this embodiment, the covariance estimation prediction equation is as follows:

[0102]

[0103] In this embodiment, the measurement equation is as follows:

[0104] Z j(k)=H j ·X j (k)+ν j (k),j=1,2,3,4

[0105] Among them, X j (k+1) is the state vector of the vehicle under a certain model at time k+1, β j (k), γ j (k),κ j,fl (k),κ j,fr (k),κ j,rl (k), k j,rr (k) represents the vehicle sideslip angle, vehicle yaw rate, tire slip rate of the left front wheel, tire slip rate of the right front wheel, tire slip rate of the left rear wheel and tire slip rate of the right rear wheel at the time k under the jth model; F j is the state transition matrix; ω j (k) is the process noise with zero mean; P j (k) represents the covariance of vehicle state information; Q j (k) is the process noise ω j The covariance matrix of (k); Z j (k) is the observation data at time k Z j1 (k), Z j2 (k), Z j3 (k), Z j4 (k), Z j5 (k), Z j6 (k) represents the vehicle sideslip angle, vehicle yaw rate, tire slip rate of the left front wheel, tire slip rate of the right front wheel, tire slip rate of the left rear wheel and tire slip rate of the right rear wheel of the vehicle measured at time k respectively; H j is the observation matrix of the jth model; v j (k) is the Gaussian observation noise.

[0106] S3. Design and apply interactive multi-model extreme value search control algorithm;

[0107] like Figure 2As shown, in this embodiment, the interactive multi-model extreme value search control (IMM-ESC) algorithm is adopted as a control strategy, combining the advantages of interactive multi-model (IMM) and extreme value search control (ESC) to meet the optimization and control requirements of complex systems. The interactive multi-model extreme value search control (IMM-ESC) algorithm uses multiple models to describe the motion state of the vehicle when traveling on different roads, with each model corresponding to the motion state of the vehicle when traveling on the corresponding road. The interactive multi-model extreme value search control (IMM-ESC) algorithm uses an interactive method to update the probabilities and weights of the vehicle models to reflect the most likely motion state of the current vehicle. The interactive multi-model extreme value search control (IMM-ESC) algorithm combines the estimation results of multiple vehicle models and calculates them through a weighted average method to obtain the final vehicle state estimate and covariance estimate. The interactive multi-model extreme value search control (IMM-ESC) algorithm also searches for extreme points of system performance, including but not limited to: maximum or minimum values, by measuring system outputs and designing control inputs.

[0108] In this embodiment, the interactive multi-model extreme value search control (IMM-ESC) algorithm can automatically adjust models and strategies based on the dynamic changes of the system to adapt to system changes; even when the system parameters are uncertain or there are external interferences, the algorithm can maintain stability and effectiveness; it can optimize the system in real time and quickly respond to system changes; by fusing multiple models, it can more comprehensively describe and handle the complexity of the system. Therefore, interactive multi-model extreme value search control can be applied to scenarios such as adaptive control and optimization of multi-agent systems. Through the algorithm's adaptive and robust characteristics, the performance and reliability of the system can be improved, while reducing the reliance on precise knowledge of the system model.

[0109] S31. Obtain the vehicle's own state information at the previous moment, and input the state information into different vehicle models for input interaction, and obtain an interaction value;

[0110] In this embodiment, the vehicle state information is processed using the previously established vehicle model to infer the observation data at the current moment. The state information is input into the four vehicle motion models for input model interaction. According to the covariance state estimation of different models and the latest measurement value z of the initial model, k The vehicle model is reinitialized using the Markov transformation matrix between different models. Initial values ​​of the four vehicle motion models after interaction are then obtained, including initial values ​​for vehicle state estimation and covariance estimation. The interaction value is obtained based on the observed data and the initial values.

[0111] in, and m,j=1,2,3,4 are the initial state estimates and covariance matrices of each sub-model at the filter input at time k-1. The input of the j-th model at time k can be reinitialized as follows:

[0112]

[0113] Note that q j (k-1) is the latest probability of the j-th model at time k-1, and the results are as follows:

[0114]

[0115] Here, is the mixing probability, and its calculation formula is

[0116] S32, inputting the interaction value into the Kalman filter corresponding to each vehicle model, performing parallel filtering processing, and obtaining a filtering estimation result;

[0117] In this embodiment, the variable residual of the Kalman filter predictor is calculated using the measured value and state estimation of the vehicle model. The following formula is the calculation formula of the variable residual in the Kalman filter:

[0118]

[0119] in, Represents the variable residual in the Kalman filter.

[0120] In this embodiment, the mean square error is calculated using the covariance estimation and the observation matrix. The following formula is the calculation formula for the mean square error in the Kalman filter:

[0121]

[0122] Among them, S j (k) represents the mean square error in Kalman filtering, and R represents the measurement noise.

[0123] In this embodiment, the Kalman gain is calculated using the covariance estimation, the observation matrix, and the mean square error. The following formula is the calculation formula of the Kalman gain:

[0124]

[0125] Among them, K j (k) represents the Kalman gain.

[0126] In this embodiment, the updated value of the vehicle state estimate at the current moment is calculated by combining the prediction equation of the state estimate, the Kalman gain, and the variable residual of the Kalman filter predictor. The following formula is the calculation formula for the updated value of the vehicle state estimate:

[0127]

[0128] In this embodiment, the updated value of the covariance is calculated by combining the predicted covariance matrix, the Kalman gain, and the predicted mean square error. The following formula is the calculation formula for the updated value of the covariance estimate:

[0129]

[0130] S33, updating the probability of each vehicle model according to the likelihood function of the new information in each filter and the Markov state transfer matrix to obtain an updated value of the vehicle model probability;

[0131] In this embodiment, the maximum likelihood function is used for model selection, that is, the model that best explains the data is selected from multiple candidate models. For the j-th model, the maximum likelihood function matching at the k-th time is as follows:

[0132]

[0133] In this embodiment, after obtaining the maximum likelihood function, the Markov transition matrix is ​​used to calculate the probability of each model at the current moment. These probabilities reflect the possibility of each model describing the current system state. The Markov transition matrix is:

[0134]

[0135] where p mj is the transition probability from the mth model to the jth model.

[0136] So the latest probability q of the jth model at the kth moment j The calculation formula for (k) is as follows:

[0137]

[0138] Here c is the normalization factor, and the calculation formula is

[0139] At this point, the update of the model probability is completed through the above calculations.

[0140] S34, outputting the filter estimation results corresponding to each filter and the updated values ​​of each vehicle model probability interactively to obtain a predicted value of the vehicle state information at the next moment;

[0141] S35, using the obtained predicted value as an input value of a stable steering stabilizer;

[0142] In this embodiment, the optimal estimate of the vehicle state and covariance at the current moment output by the Kalman filter predictor is calculated with the updated vehicle model probability to obtain the final result, namely the vehicle state estimate and covariance estimate.

[0143] The state estimation of the vehicle model after probability update is as follows:

[0144]

[0145] The covariance estimate after probability update is as follows:

[0146]

[0147] At the next iteration, the output value at the current time will be used as the input value of the interactive input, and a new iteration output will be restarted. At this point, the entire recursive process of the interactive multi-model is completed.

[0148] S36, selecting a performance indicator related to the controller input value as an objective function, and reflecting the shock absorption effect of the vehicle at this time through the magnitude of the objective function;

[0149] In this embodiment, an index related to the controller input value and the vehicle's shock absorption performance is selected as the objective function, and the index includes the vehicle's sideslip angle, the vehicle's yaw rate, and the slip rates of the vehicle's four tires;

[0150] Using the selected indicators, a suitable objective function is formulated, and the optimal values ​​of the vehicle sideslip angle, vehicle yaw rate, and slip rate of the four tires of the vehicle are set. Then, the mean square error between the input value and the ideal value is used as the objective function;

[0151] When the chosen metric has constraints, constrained optimization methods can be used to integrate all constraints into the objective function, using penalty terms or modifying the objective function to include these constraints. If the constraints are violated, the penalty term will increase the value of the objective function, making the solution undesirable. According to the described method, the objective function is formulated as follows:

[0152] J=w1·f1(κ)+w2·f2(β)+w3·f3(ψ)

[0153] in:

[0154] f1(κ), f2(β), and f3(ψ) are cost functions of tire slip rate, vehicle sideslip angle, and vehicle yaw rate;

[0155] w1, w2, and w3 are the weights corresponding to each indicator, which are used to adjust their relative importance in the overall objective function.

[0156] The specific form of the cost function is as follows:

[0157] The tire slip cost function is as follows:

[0158] f1(κ)=(κ-κ opt ) 2

[0159] where κ opt is the ideal slip ratio, usually close to zero.

[0160] In this embodiment, the vehicle sideslip angle cost function is as follows:

[0161] f2(β)=(β-β opt ) 2

[0162] where β opt is the ideal slip ratio, which is usually close to zero.

[0163] In this embodiment, the vehicle yaw rate cost function is as follows:

[0164] f3(ψ)=var(ψ)

[0165] Among them, var(ψ) is the variance of the vehicle yaw rate, which can reflect the stability of the vehicle yaw rate.

[0166] In this embodiment, the constraint of the indicator is as follows:

[0167] The constraint of the vehicle sideslip angle is as follows:

[0168] |β|≤arctan(0.02μ·g)

[0169] The constraint on the vehicle's yaw rate is as follows:

[0170]

[0171] The modified objective function is as follows:

[0172]

[0173] Among them, λ1 and λ2 are penalty factors, which are used to adjust the penalty intensity when the constraint is violated.

[0174] The smaller the objective function is, the more stable the vehicle is during driving, that is, the better the shock absorption effect of the vehicle is.

[0175] S37. Based on the vehicle's shock absorption performance, use extreme value search control to periodically adjust input values ​​to search for the extreme value of the objective function. Initially, the control input is varied by a large amplitude to search for the extreme value of the performance indicator within the entire operating range. Then, the amplitude of the control input variation is gradually reduced to more precisely approximate the discovered extreme value point.

[0176] In this embodiment, the control input and performance index are clearly defined. In the present invention, the control input includes tire slip rate, vehicle sideslip angle and vehicle yaw rate, while the performance index is the above-mentioned cost function f1(κ), f2(β), f3(ψ);

[0177] In this embodiment, an excitation signal needs to be designed. It is usually a periodic signal, such as a sine wave, used to adjust the control input. This signal will be added to the normal operating point of the system. There are two excitation signals in the algorithm, which are as follows:

[0178] u1(t)=sin(ω i t+β i -φ i )

[0179] u2(t)=A i sin(ω i t+β i ), i=1,2,3,4,5,6

[0180] Among them, A i is the amplitude of the excitation signal corresponding to each search variable, ω i is the frequency of the excitation signal corresponding to each search variable, t is the time, β i is the initial phase of the excitation signal corresponding to each search variable, φ i is the phase difference of the excitation signal corresponding to each search variable;

[0181] Apply stimulus signals to the system and measure performance indicators for each cycle. Record how performance indicators change over time.

[0182] Analyze the data of performance indicators changing with the excitation signal to determine the extreme points. Find the control input value that maximizes or minimizes the performance indicator;

[0183] Based on the results of the extreme value search, the control input of the system is updated to make the system tend to the extreme value point. In the present invention, when the objective function reaches the minimum value, the steering stability of the vehicle is the best. Therefore, we need to use the extreme value search control to find the minimum value of the objective function. If the objective function decreases with the increase of the control input, the control input should be appropriately increased; conversely, if the objective function increases with the increase of the control input, the control input should be appropriately reduced.

[0184] During the iteration process, ensure that the control input and system state meet all operational constraints. If a constraint violation is detected, add a penalty term or adjust the stimulus signal;

[0185] Gradually refine the control input through multiple iterations to more accurately approach the extreme point;

[0186] When the system performance index stabilizes near the extreme value, reduce the amplitude of the excitation signal to lock the system at this point. The specific manifestations are as follows:

[0187] The objective function we developed indicates that the search process requires optimizing the extremum of multiple variables. Therefore, we employ multivariable extremum search control. This method uses different sinusoidal excitation signals and filters to obtain the gradients of each search variable. The integrated values ​​of each search variable are then synthesized and demodulated to serve as the system's control variable. Because the objective function gradients for each search variable are not identical, the separate search variables must be separated to prevent interference, requiring a design structure with independent channels.

[0188] In the above excitation signal, parameter i=1,2,…,6, when i is an odd number, the frequency ω of the excitation signal i+1 =ω i , and the initial phase of the excitation signal is defined as:

[0189]

[0190] In the entire multivariable extreme value search control system structure, there are also two filters, high-pass and low-pass. Through the combined action of the excitation signal and the filter, the gradient of the objective function J(u) can be measured. Along the negative gradient direction of the objective function J(u), the variable can gradually converge to the extreme value u. * , so that the objective function reaches its minimum value.

[0191] in, The high-pass filter is Low pass filter is -C ii (s)·Γ ui (s).

[0192] Notice:

[0193] Γ f (s) and Γ ui (s) are all strictly true functions;

[0194] For all i=1,2,…,6, the filter and -C ii (s)·Γ ui (s) are all true.

[0195] Under the action of the filter, ignoring the influence of high-order coefficients, the objective function can be expressed as:

[0196] J(u)=J(u * )+(uu * ) T·P·(uu * ) (*)

[0197] Where P = P T >0, Represents the extreme point of the objective function.

[0198] During the design process, the frequency variation of the excitation signals u1(t) and u2(t) is usually much smaller than the system frequency. By performing Taylor expansion on the objective function, we can obtain:

[0199]

[0200] The high-pass filter can eliminate the low-frequency signal J(u) in J(u+u2), so the signal can be obtained in the control loop. After coupling with the excitation signal u1(t) and the action of the low-pass filter, the output signal of the low-pass filter will only retain the first-order derivative part, that is, the output signal is The output signal at this time is the gradient of the objective function J(u).

[0201] Any vector u with the minimum real quadratic form * The objective function can be locally approximated by the formula (*). When P>0, if the function J(u) has a maximum value, C ii (s) is replaced by -C ii (s). Through gradual approximation, the objective function will eventually reach its minimum value, and at this time, each search variable will also reach its extreme value point, that is, the vehicle performance parameters will reach the optimal value.

[0202] In summary, the steps of multivariable extreme value search control can be simply summarized as first, through the interaction of the excitation signal and the filter, the objective function J(u) is estimated for each component u i The first-order partial derivative of , and then by continuously approximating its first-order partial derivative, the vector u also gradually converges to its extreme value vector u * , so that the objective function also reaches its minimum value.

[0203] S38. When the objective function reaches an extreme value, based on the vehicle state information parameters that have reached the extreme value point, the vehicle performance parameters are all optimized, and the vehicle's shock absorption performance is also improved;

[0204] In this embodiment, the vehicle state information parameter corresponding to the performance index that is stable near the extreme value is used as the latest input of the stable steering controller, so that the vehicle can achieve more stable driving;

[0205] By improving the stability of the vehicle during driving, the shock absorption performance of the vehicle during driving can be effectively improved.

[0206] The Interacting Multiple Model (IMM) algorithm is a filtering method for handling uncertain system dynamics, particularly when the target's motion model may switch between several known models. The core idea of ​​the IMM algorithm is to simultaneously use multiple filters, each corresponding to a possible target motion model, and interactively determine which model is most likely to describe the current target state. The following are the core steps of the IMM algorithm and the principles it is based on:

[0207] In the model selection process of this embodiment, a set of models is determined, each with its corresponding dynamic equations. Specifically, based on an understanding of the system's behavior, a set of models that can describe the system's possible states is selected. These models should cover all possible motion states of the system.

[0208] In the filter initialization process of this embodiment, an initial state estimate and a covariance matrix are set for each model to initialize the filter; specifically, a filter is initialized for each model, and these filters run in parallel, and each filter performs state estimation according to the dynamic equation of its corresponding model.

[0209] In the estimation mixing process of this embodiment, the estimation weights of each filter are calculated. These weights are usually based on the model probability and the filter performance, including but not limited to: likelihood. Specifically, at each time step, the estimation results of all filters are mixed according to certain weights to obtain a global estimate.

[0210] In the model probability update process of this embodiment, the probability of each model is updated according to the model transition probability and filter performance; specifically, a Markov chain model is used to describe the transition probability between models, that is, the probability that the target will transfer to another model at the next moment.

[0211] In the filter update process of this embodiment, a filtering algorithm is applied to update the state estimate of each filter in combination with new observation data. Specifically, for each model, a corresponding filtering algorithm (such as a Kalman filter) is used to update the state estimate and covariance matrix of the filter.

[0212] In the model likelihood calculation process of this embodiment, for each filter, its likelihood for the new observation data is calculated: specifically, the degree of fit of each model for the new observation data, that is, the likelihood, is calculated, which helps to evaluate the applicability of the model.

[0213] In the model weight updating process of this embodiment, the weight of each model is updated in combination with the likelihood and the model transition probability. Specifically, the weight of each model is updated according to the performance of the filter and the model transition probability to prepare for the next estimation of the mixture.

[0214] In the iterative process of this embodiment, when new observation data arrives, estimation mixing, filter update, model probability update and model weight update are repeatedly performed; specifically, the IMM algorithm is an iterative process, and the above steps are repeated at each time step to adapt to the dynamic changes of the target motion.

[0215] In this embodiment, the IMM algorithm uses multiple models to describe the possible states of the system, which can more comprehensively cover the dynamic behavior of the system. This method is more adaptable to the variable motion of the target than a single model. Because it combines the outputs of multiple filters, it has good robustness to noise and model uncertainty. Even if the prediction of a certain model deviates, the predictions of other models can provide supplementary information. The algorithm can automatically adjust the usage weights of each model according to the motion characteristics of the target, so that the filter can adaptively track the changes of the target, thereby improving the tracking accuracy. The algorithm also allows users to select and customize different models according to the needs of actual applications. This flexibility allows the IMM algorithm to be applied to a variety of different systems and scenarios. Therefore, IMM can enable vehicles to adaptively drive on the road by integrating road models of different levels, thereby improving the shock absorption effect of the vehicle during driving.

[0216] The Extreme Seeking Control (ESC) used in this embodiment is an adaptive control method that designs control inputs by measuring system outputs to optimize system performance. This algorithm does not rely on a precise mathematical model of the controlled object, but instead adjusts the control strategy through real-time feedback from the system. The following are the core steps of Extreme Seeking Control and the principles behind it:

[0217] In the system modeling process of this embodiment, the input, output and state variables of the system, as well as the relationship between them, are identified; specifically, although extreme value search control does not rely on a precise mathematical model, a basic description of the system is required to determine which variables are controllable, which are measurable, and how they affect the system output.

[0218] In the reference model establishment process of this embodiment, the reference model is established based on performance indicators, for example, minimizing error or maximizing output; specifically, the reference model is a mathematical description of the desired system performance, and the goal of extreme value search control is to make the system output track or reach this reference model.

[0219] In the objective function definition process of this embodiment, an objective function is defined, such as the sum of squared errors or a loss function; specifically, the objective function is an indicator that measures the difference between the system output and the reference model, which will be used to guide the search process.

[0220] In the extreme value search mechanism of this embodiment, an extreme value search mechanism is implemented to find the extreme value point corresponding to the objective function; by changing the control input and observing the change of the performance function, the optimal control input is searched, that is, the control strategy of finding the extreme value point that makes the performance function reach the extreme value is found.

[0221] In the feedback and iteration operation of this embodiment, system output data is collected, performance functions are calculated, and control inputs are adjusted according to an extreme value search mechanism. Specifically, real-time feedback of the system output is used to adjust the control strategy, and the optimal solution is gradually approached through an iterative process.

[0222] In the stability and convergence analysis process of this embodiment, mathematical analysis or simulation is performed to verify the stability and convergence of the algorithm under different conditions; specifically, the stability and convergence of the algorithm are ensured, that is, the system can stabilize in an optimal or near-optimal state over time.

[0223] Example 2

[0224] In this embodiment, the interactive multi-model extremum search control method for vehicle shock absorption is deployed in an actual system to achieve real-time control and optimization.

[0225] In this embodiment, the algorithm is implemented on an actual hardware or software platform to perform real-time data acquisition, processing and control.

[0226] Extremum-seeking control is an adaptive control strategy that optimizes system performance by adjusting control inputs to achieve a maximum or minimum target value. This algorithm does not require a precise mathematical model of the system, meaning it can be applied to complex systems where precise models are difficult to establish. It can adapt to changes in system parameters and external environmental disturbances, remaining effective despite system dynamics. It can adjust control inputs in real time, rapidly responding to changes in system state, thereby achieving real-time optimization. It can be flexibly applied to different types of systems and performance metrics, whether maximizing or minimizing the objective function. Furthermore, because it does not rely on model accuracy, the algorithm is inherently robust to model errors and external disturbances. In multi-objective optimization problems, extremum-seeking control can help find the optimal trade-off solution that meets multiple performance metrics. Furthermore, this algorithm can be combined with other control strategies to improve overall control performance and can be applied to nonlinear systems. With appropriate search strategies and algorithm adjustments, it can effectively handle nonlinear problems. Therefore, extremum-seeking control can improve vehicle steering stability and vibration damping.

[0227] Combining the interactive multi-model (IMM) with extreme value search control (ESC) and applying it to the vehicle shock absorption system can significantly improve the vehicle's shock absorption effect. This combination utilizes the model diversity and adaptability of the IMM algorithm and the optimization capability of extreme value search control to achieve more efficient and intelligent shock absorption control. This method enables the vehicle's shock absorption system to adaptively adjust according to road conditions and driving behavior to achieve the best shock absorption effect. Both algorithm frameworks are easy to integrate and can work together with other intelligent systems in the vehicle. The multi-model characteristics of the IMM algorithm and the adaptability of extreme value search control can also improve the system's robustness to some uncertain external disturbances. Therefore, this method can improve the vehicle's shock absorption performance and provide a more reliable solution.

[0228] like Figure 4 As shown, the present invention utilizes a modular design in its structure, allowing each control module to be independently developed and tested, facilitating maintenance and upgrades. This method utilizes an interactive multi-model algorithm to integrate multiple vehicle models, enabling the selection of the most appropriate model based on varying road conditions and vehicle states, thereby improving the system's adaptability and flexibility. This method improves the accuracy and reliability of data processing by fusing data from multiple sensors, including tire slip rate, vehicle sideslip angle, and vehicle yaw rate. Furthermore, this method utilizes extreme value search control to rapidly find the optimal values ​​for parameters in the stable steering controller, thereby improving the real-time performance of the vehicle's shock absorption control.

[0229] like Figure 5 As shown, the present invention dynamically adjusts damping parameters based on real-time data to adapt to changing driving conditions, ensuring optimal damping. By optimizing the damping control strategy, vehicle jolting is reduced on various road conditions, significantly improving ride comfort. Precisely controlling vehicle stability reduces the risk of loss of control due to uneven road surfaces, enhancing driving safety. Optimized damping control reduces unnecessary energy consumption and improves vehicle energy efficiency.

[0230] Vehicles can travel on a variety of different road types, such as highways. The interactive multi-model extreme value search method analyzes vehicle data in real time and automatically selects or adjusts the vehicle model most suitable for the current road surface. This reduces vehicle body sway and vibration during high-speed driving, ensuring optimal shock absorption on all road surfaces, thereby improving vehicle stability and ride comfort. Highways often have a roughness rating of B or C. This method utilizes the four road models fused by the IMM, which include those for Class B and C. Therefore, the matching of vehicle models is improved, resulting in better shock absorption during driving.

[0231] In addition, highways usually have good road conditions, reducing bumps and unevenness, and allow driving at higher speeds on the road, so the vehicle will be more stable during driving.

[0232] In summary, the present invention adopts an extreme value search control method to adjust the controller parameters by optimizing performance indicators to gradually achieve adaptive regulation of the system. This method can find the optimal performance parameters of the vehicle by searching for the extreme values ​​of the objective function. Even if the vehicle parameters change during driving, this method can find new optimal performance parameters, thereby achieving the effect of real-time optimization of the vehicle's shock absorption performance. The present invention has low information requirements for the controlled object, which means that excessive vehicle information parameters are not required. Therefore, it can solve the problem of traditional control methods' dependence on the number and accuracy of sensors, reduce vehicle costs, and avoid the complex data problems involved in processing multi-sensor data fusion, reducing the computational burden of the system.

[0233] The present invention designs a control method combining interactive multi-model (IMM) and extreme value search control (ESC), which can improve the shock absorption performance of the vehicle during driving by optimizing the performance parameters of the vehicle.

[0234] During driving, a vehicle may traverse a variety of different roads. Because a traditional single road model cannot cover all road conditions, the present invention utilizes an IMM, which includes four road models with different road surface grades: A, B, C, and D. By interactively integrating these four road models, the vehicle can adapt to more complex road conditions. Extreme value search control is then used to find the extreme value of the objective function, obtaining the optimal vehicle performance parameters at the current moment. These parameters are then input into the vehicle's stable steering controller to achieve the best possible shock absorption effect. Due to the real-time optimization characteristics of extreme value search control, it can always obtain the optimal vehicle performance parameters regardless of changes in road conditions. Therefore, this method can effectively cope with variable road conditions, such as bumps and uneven surfaces, while ensuring vehicle stability during driving.

[0235] The present invention solves the technical problems existing in the prior art, such as high dependence on the number and accuracy of sensors, heavy system calculation burden caused by processing complex data of multi-sensor data fusion, which restricts the shock absorption effect of the vehicle, and high vehicle cost.

[0236] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. An interactive multi-model extreme value search control method for vehicle shock absorption, characterized by: The method comprises: S1. Obtain vehicle operating status data to obtain vehicle status parameters; S2. Establishing a system model based on the discrete space model, and setting a vehicle motion state model in the system model for at least two road surface unevenness excitation scenarios, wherein the vehicle motion state model includes: a state estimation prediction equation, a covariance estimation prediction equation, a measurement equation, and an eight-degree-of-freedom vehicle model; S3. Combining the interactive multi-model IMM and the extreme value search control model ESC, construct and utilize the interactive multi-model extreme value search control IMM-ESC algorithm as a control strategy, wherein the interactive multi-model extreme value search control (IMM-ESC) algorithm adopts no less than two vehicle motion state models to describe the vehicle driving motion state according to the vehicle state parameters; interactively update the probabilities and weights of the vehicle motion state models to obtain multi-vehicle model state estimation results according to the vehicle driving operation state, and perform weighted averaging processing based on the results to obtain vehicle state estimation and covariance estimation; search for system performance extreme points through system output measurement operations and control input design operations; S3 includes: S31. Obtain vehicle state parameters of the vehicle at the previous moment, input the vehicle state parameters into different vehicle motion state models for input interaction, and obtain interaction values; S32, inputting the interaction value into the Kalman filter predictor corresponding to each vehicle motion state model, performing parallel filtering processing, and obtaining a filtering estimation result; S33. Update the probability of each vehicle motion state model according to the maximum likelihood function of each filter and the Markov state transfer matrix to obtain a vehicle model probability update value; S34, performing output interaction operations on the filter estimation results corresponding to each Kalman filter predictor and the vehicle model probability update value of each vehicle motion state model to obtain a predicted value of the vehicle state information at the next moment; S35, using the predicted value of the vehicle state information at the next moment as the input value of the stabilizing steering stabilizer, wherein the current vehicle state, the optimal covariance estimate, and the vehicle model probability update value output by the Kalman filter predictor are calculated to obtain a vehicle state estimate and a vehicle covariance estimate; S36, selecting a controller input value correlation index as an objective function to represent the current vehicle shock absorption information; S37, using extreme value search control to periodically adjust input values ​​based on current vehicle shock absorption information to search for an extreme value of the objective function; S38. When the objective function reaches an extreme value, the optimal vehicle state information parameters are obtained based on the extreme value point searched at this time.

2. The interactive multi-model extreme value search control method for vehicle shock absorption according to claim 1, characterized in that: In S2, the eight-degree-of-freedom vehicle model in the discrete space model is expressed using the following logic: i=fl,fr,rl,rr Where β(k) represents the vehicle sideslip angle at time k; γ(k) represents the vehicle yaw rate at time k; κ i (k) represents the tire slip rate of the four tires at time k; T s represents a fixed time step; F yf Indicates the lateral force of the vehicle's front tires; F yr Indicates the lateral force of the vehicle's rear tires; δ f (k) represents the active front steering angle; m represents the mass of the vehicle; v represents the vehicle speed; l f Indicates the distance from the vehicle's center of gravity to the front axle; l r Indicates the distance from the vehicle's center of gravity to the rear axle; I z Represents the vehicle's yaw moment of inertia; M z represents the vehicle yaw moment; r represents the tire radius; J represents the moment of inertia of each tire; C ki Indicates the longitudinal stiffness of the four tires; T i (k) represents the motor torque of each tire at time k.

3. The interactive multi-model extreme value search control method for vehicle shock absorption according to claim 1, characterized in that: In S2, the state estimation prediction equation is expressed using the following logic: X j (k+1)=F j ·X j (k)+ω j (k),j=1,2,3,4 The covariance estimation prediction equation is expressed using the following logic: The measurement equation is expressed using the following logic: From j (k)=H j ·X j (k)+v j (k),j=1,2,3,4 Where, X j (k+1) is the state vector of the vehicle under a certain model at time k+1, β j (k), γ j (k),κ j,fl (k),κ j,fr (k), k j,rl (k), k j,rr (k) represents the vehicle sideslip angle, vehicle yaw rate, tire slip rate of the left front wheel, tire slip rate of the right front wheel, tire slip rate of the left rear wheel and tire slip rate of the right rear wheel at the time k under the jth model; F j is the state transition matrix; ω j (k) is the process noise with zero mean; P j (k) represents the covariance of vehicle state information; Q j (k) is the process noise ω j The covariance matrix of (k); Z j (k) is the observation data at time k Z j1 (k), Z j2 (k), Z j3 (k), Z j4 (k), Z j5 (k), Z j6 (k) represents the vehicle sideslip angle, vehicle yaw rate, tire slip rate of the left front wheel, tire slip rate of the right front wheel, tire slip rate of the left rear wheel and tire slip rate of the right rear wheel of the vehicle measured at time k respectively; H j is the observation matrix of the jth model; v j (k) is the Gaussian observation noise.

4. The interactive multi-model extreme value search control method for vehicle shock absorption according to claim 1, characterized in that: In S31, the vehicle state parameters are processed using the vehicle motion state model to obtain current moment observation data; The state information is input into the four vehicle motion state models respectively, and the input model interaction is performed. According to the covariance state estimation of different models and the latest measurement value z of the initial model, k , the vehicle model is reinitialized using the Markov transformation matrix between different models; in, and m, j = 1, 2, 3, 4 are the initial state estimates and covariance matrices of each sub-model of the filter input at time k-1; the reinitialization operation is performed on the j-th vehicle motion state model at the k-th time input using the following logic: Note that q j (k-1) is the latest probability of the j-th model at time k-1, and the results are as follows: Here, is the mixing probability, and its calculation formula is Obtaining the initial values ​​of the vehicle motion model after interaction, including: the initial values ​​of vehicle state estimation and covariance estimation; The interaction value is obtained by processing the observation data at the current moment and the initial value of the vehicle motion model after the interaction.

5. The interactive multi-model extreme value search control method for vehicle shock absorption according to claim 1, characterized in that: In S32, the following logic is used to calculate the Kalman filter variable residual based on the current observation data and state estimation data of the vehicle motion state model: Where, Represents the variable residual in Kalman filtering; The following logic is used to construct an observation matrix based on the current observation data, and the Kalman filter mean square error is calculated using the covariance estimate and the observation matrix: Where S j (k) represents the mean square error in Kalman filtering, and R represents the measurement noise; The Kalman gain is calculated using the covariance estimate, the measurement matrix, and the Kalman filter mean square error according to the following logic: Where K j (k) represents the Kalman gain; The following logic is used to combine the state estimation prediction equation, the Kalman gain, and the Kalman filter variable residual to obtain the updated value of the vehicle state estimate at the current moment: The following logic is used to combine the prediction covariance matrix, the Kalman gain, and the mean square error to obtain the covariance update value:

6. The interactive multi-model extreme value search control method for vehicle shock absorption according to claim 1, characterized in that: In S33, for the j-th vehicle motion state model, at the k-th time, a likelihood function matching operation is performed: The maximum likelihood function is obtained, and the Markov transfer matrix is ​​expressed using the following logic to obtain the probability of each vehicle motion state model at the current moment: Where p mj is the transition probability from the mth model to the jth model; Using the following logic, the vehicle model probability update value q of the j-th vehicle motion state model at the k-th time is obtained: j (k): Where c is the normalization factor.

7. The interactive multi-model extreme value search control method for vehicle shock absorption according to claim 1, characterized in that: In S35, the vehicle state estimation is expressed using the following logic: The vehicle covariance estimate is expressed using the following logic: In the next iteration, the output value at the current time is used as the interactive input value to continue the new iterative output operation.

8. The interactive multi-model extreme value search control method for vehicle shock absorption according to claim 1, characterized in that: In S36, using the objective function, setting input values ​​and ideal values, and calculating the mean square error based on the input values ​​and the ideal values ​​to serve as the objective function; When the controller input value associated index has constraints, all constraints are integrated into the objective function using a constrained optimization method.

9. An interactive multi-model extreme value search control system for vehicle shock absorption, configured to execute the interactive multi-model extreme value search control method for vehicle shock absorption according to any one of claims 1 to 8, characterized in that: The system comprises: The vehicle status acquisition module is used to obtain vehicle operating status data to obtain vehicle status parameters; a system model establishment module, configured to establish a system model based on a discrete space model, and to respectively set a vehicle motion state model in the system model for at least two road surface unevenness excitation scenarios, wherein the vehicle motion state model includes: a state estimation prediction equation, a covariance estimation prediction equation, a measurement equation, and an eight-degree-of-freedom vehicle model; A multi-vehicle model state estimation module is used to combine the interactive multi-model IMM and the extreme value search control model ESC to construct and utilize the interactive multi-model extreme value search control IMM-ESC algorithm as a control strategy, wherein the interactive multi-model extreme value search control (IMM-ESC) algorithm adopts no less than two of the vehicle motion state models to describe the vehicle driving motion state according to the vehicle state parameters; the probability and weight of the vehicle motion state model are interactively updated to obtain the multi-vehicle model state estimation result according to the vehicle driving operation state, and weighted average processing is performed on the result to obtain vehicle state estimation and covariance estimation; through system output measurement operation and control input design operation, the system performance extreme point is searched, and the multi-vehicle model state estimation module is connected to the system model establishment module and the vehicle state acquisition module.

Images