Vehicle robust control method and device based on online estimation of self-vehicle state noise

Abstract

The application discloses a kind of vehicle robust control method and device based on online estimation self-vehicle state noise, belong to the field of automatic driving, the method includes: using real-time self-covariance least square method to estimate the noise in the process of intelligent vehicle operation, dynamically update vehicle state estimation error;According to the pre-set signal interval of vehicle state estimation error, determine the estimation error set of vehicle state uncertainty range;Using the estimation error set, construct linear programming problem, calculate minimum robust positive invariance set;The minimum robust positive invariance set calculated is applied to the model predictive control algorithm based on pipe, controls vehicle.The method is based on real-time self-covariance least square method dynamically update vehicle state estimation error, according to the distribution definition error boundary of vehicle state estimation error, describe the uncertainty range of vehicle state, constrain error system in minimum robust positive invariance set, avoid the safety accident caused by external disturbance or model uncertainty.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of autonomous driving, and in particular to a vehicle robust control method and device based on online estimation of self-vehicle state noise. BACKGROUND

[0002] Autonomous driving technology can improve road traffic safety and also has profound significance in promoting intelligent transportation systems and promoting the innovative development of related industries. In the driving process of intelligent vehicles, the complexity of the driving environment and the uncertainty of the road and sensors require that the vehicle control system must have high robustness to cope with changing driving environments and ensure the safety of the vehicle in any situation. Vehicle state estimation is the basis for realizing robust control of intelligent vehicles. Based on accurate vehicle state estimation information, the control system can make effective decisions.

[0003] Currently, many vehicle state estimation methods are implemented based on Kalman filters. These methods recursively update the vehicle state by utilizing sensor information and system dynamic model predictions. However, Kalman filters rely on the setting of prior noise covariance matrices, which are often difficult to obtain or accurately model in practical applications. When noise characteristics are unknown or change, traditional Kalman filtering methods may face problems of decreased estimation accuracy or filter divergence. SUMMARY

[0004] The main purpose of the present application is to avoid safety accidents caused by external disturbances or model uncertainties.

[0005] The present application provides a vehicle robust control method based on online estimation of self-vehicle state noise, comprising: estimating the noise in the running process of the intelligent vehicle using real-time self-covariance least squares method, and dynamically updating the vehicle state estimation error; determining the estimation error set of the vehicle state uncertainty range according to the pre-set confidence interval of the vehicle state estimation error; using the estimation error set to construct a linear programming problem and calculate the minimum robust positive invariant set; applying the calculated minimum robust positive invariant set to the tube-based model predictive control algorithm to control the vehicle.

[0006] In the vehicle robust control method based on online estimation of self-vehicle state noise of the present application, the noise in the running process of the intelligent vehicle is estimated using real-time self-covariance least squares method, comprising:

[0007] Converting the state space equation of the vehicle system into a state space equation based on innovation;

[0008] For each time window, if the current window is the first window, the estimation error covariance matrix, the process noise covariance matrix and the observation noise covariance matrix initialized according to experience information are assigned to the current window, otherwise the process noise covariance matrix and the observation noise covariance matrix at the initial time of the current window are determined according to the estimation values of the last window;

[0009] In each time window, the Kalman filter is used to update the vehicle state and the estimation error covariance matrix at each time until the end of the window, the innovation self-covariance sequence is calculated based on the new information obtained from the two state space equations in the current window, and a least square problem is constructed to estimate the process noise covariance matrix and the observation noise covariance matrix.

[0010] In the vehicle robust control method based on online estimation of self-vehicle state noise, the state space equation comprises:

[0011]

[0012] The state space equation based on innovation comprises:

[0013]

[0014] wherein k represents the kth time, x is the state vector of the system, u is the control input of the system, y is the observation of the state space system, is the estimation error, is the estimation value of the state quantity at the kth time, is the innovation, A, B and C are the coefficient matrices of the system, w and v are the process noise and the observation noise respectively, L is the steady-state Kalman gain, I is the unit matrix,

[0015] In the vehicle robust control method based on online estimation of self-vehicle state noise, the preset confidence interval comprises 99.7%.

[0016] In the vehicle robust control method based on online estimation of self-vehicle state noise, the innovation self-covariance sequence calculated based on the innovation obtained from the two state space equations in the current window comprises:

[0017] The innovation self-covariance sequence Ξ(N) of length N in the nth window is calculated according to the following formula n :

[0018]

[0019] wherein α i is a forgetting factor, and the calculation formula comprises:

[0020]

[0021] wherein m is a parameter, different values are selected according to different characteristics of the system, the larger m is, the smaller the early innovation weight in the window is.

[0022] In the vehicle robust control method based on online estimation of self-vehicle state noise, the construction of the least square problem, the estimation of the process noise covariance matrix and the observation noise covariance matrix comprises the following steps: The covariance matrix of the process noise and the observation noise is calculated, wherein:

[0023]

[0024] wherein, is A LS in the n th window. W n is V n is the estimated value of the process noise covariance matrix and the observation noise covariance matrix in the n th window, S n-1 is a weight matrix, when n = 1, S n-1 is a unit matrix, when n > 1, S n-1 The calculation formula of S

[0025]

[0026] Each parameter matrix is as follows:

[0027]

[0028] wherein the process noise covariance matrix W = E (ww T ), the observation noise covariance matrix V = E (vv T ).

[0029] In the vehicle robust control method based on online estimation of self-vehicle state noise, the estimation error set is used to construct a linear programming problem, and the minimum robust positive invariance set is calculated, comprising the following steps:

[0030] The control rate is used to divide the actual vehicle system into a nominal system and an error system, and the state space equation of the error system is obtained:

[0031] e k+1 = (A + BK) e k + w k

[0032] wherein, x k and The state quantity of the system includes the lateral velocity v y , the yaw rate , the tracking error e d and the heading error

[0033] F ∞ is calculated according to the calculation formula of the minimum robust positive invariant set:

[0034]

[0035] Wherein, F ∞ is the minimum robust positive invariant set, Θ is the step estimation error set, ⊕ is the Minkowski sum between sets, A K =A+BK, K is the state feedback gain matrix.

[0036] In the vehicle robust control method based on online estimation of self-vehicle state noise, the minimum robust positive invariant set calculated is applied to the tube-based model predictive control algorithm to control the vehicle, comprising:

[0037] The tube-based model predictive control algorithm divides the actual vehicle system into a nominal system and an error system, and converts the control of the actual system into the control of the nominal system.

[0038]

[0039] Wherein, Q and R are weight matrices, and M is the prediction time domain of the model predictive control algorithm.

[0040] When solving the nominal control quantity, the state constraints and control constraints of the nominal system are strengthened according to the constraints of the actual system, and the constraints are in the form of:

[0041]

[0042] After solving the nominal control quantity , the control quantity needed to be applied to the actual vehicle system is calculated by .

[0043] The application also provides a vehicle robust control device based on online estimation of self-vehicle state noise, comprising: a self-vehicle state noise online estimation module, which estimates noise in the process of intelligent vehicle operation by using real-time self-covariance least squares method, dynamically updates vehicle state estimation error, and determines an estimation error set of vehicle state uncertainty range according to a preset confidence interval of the vehicle state estimation error; a minimum robust positive invariance set calculation module, which is used for constructing a linear programming problem by using the estimation error set and calculating a minimum robust positive invariance set; and a vehicle robust control module, which is used for applying the calculated minimum robust positive invariance set to a tube-based model predictive control algorithm to control the vehicle.

[0044] The application has the beneficial effects that: for the robust control problem of intelligent vehicles in complex driving environments, the method of online estimation of vehicle state noise is adopted, the vehicle state estimation error is dynamically updated based on real-time self-covariance least squares method; the error boundary is defined and the uncertainty range of the vehicle state is described according to the distribution of the vehicle state estimation error, such as a 99.7% confidence interval; the error boundary information is combined, and the error system is constrained in the minimum robust positive invariance set, so as to avoid safety accidents caused by external disturbances or model uncertainty. BRIEF DESCRIPTION OF DRAWINGS

[0045] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0046] Figure 1 is a flowchart of the vehicle robust control method based on online estimation of self-vehicle state noise of the embodiment of the present application;

[0047] Figure 2 is a flowchart of online estimation of vehicle state noise of the embodiment of the present application;

[0048] Figure 3 is a structural schematic diagram of the vehicle robust control device based on online estimation of self-vehicle state noise of the embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of the present application more clear, the following will further describe the present application in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0050] In complex driving environments, intelligent vehicles need to deal with multiple uncertain factors, such as the uncertainty of vehicle dynamics model, the change of sensor noise and external disturbance. When facing these challenges, the existing vehicle control method is often difficult to keep the system safe under disturbance conditions at all times, especially when the vehicle state estimation error is large or the external disturbance is strong, the system is prone to instability or control precision decline. The traditional control method is based on prior information state estimation, which lacks enough flexibility to respond to real-time changes in the environment and system state, resulting in unsatisfactory control effect in complex driving environment.

[0051] The present application proposes a robust control method and system based on online estimation of self-vehicle state noise, which uses real-time self-covariance least squares to estimate the state estimation error of the vehicle online, and dynamically adjusts the control strategy by calculating the 99.7% confidence interval of the error to respond to the change of the vehicle state estimation error in real time. In addition, the present application also combines the method based on support function to efficiently calculate the minimum robust positive invariant set, and applies it to the tube-based model predictive control algorithm, by dividing the actual system into nominal system and error system, the control of the actual system is transformed into the control of the nominal system, at the same time, the error system is constrained within the safe range, so as to avoid safety accidents caused by external disturbance or model uncertainty.

[0052] The following will be combined Figures 1 to 3 The vehicle robust control method and device based on online estimation of self-vehicle state noise of the present application are described, Figure 1 The flowchart of the vehicle robust control method based on online estimation of self-vehicle state noise of the embodiment of the present application is shown in Figure 1 The present application provides a vehicle robust control method based on online estimation of self-vehicle state noise, comprising:

[0053] S1, using real-time self-covariance least squares to estimate the noise in the running process of intelligent vehicle, and dynamically updating the vehicle state estimation error;

[0054] S2, according to the pre-set confidence interval of the vehicle state estimation error, determining the estimation error set of the vehicle state uncertainty range; in some embodiments, the pre-set confidence interval includes 99.7%;

[0055] S3, using the estimation error set to construct a linear programming problem, and calculating the minimum robust positive invariant set;

[0056] S4, applying the calculated minimum robust positive invariant set to the tube-based model predictive control algorithm to control the vehicle.

[0057] At each moment, according to the estimation error covariance matrix obtained in step S1, the confidence interval of the estimation error, such as the 99.7% confidence interval, is calculated, and the error set is defined according to the confidence interval, and the boundary of the error set is set as the boundary of the confidence interval. The error set fully considers the distribution characteristics of the error, covers most of the state estimation errors, and thus ensures that the vehicle can still maintain safe driving under disturbance conditions. Even in extreme cases, the actual state estimation error may exceed the range of the error set, but through the rolling update characteristics of the model predictive control algorithm, the system can dynamically adjust and continuously optimize the control strategy, and still ensure the driving safety of the vehicle.

[0058] The vehicle robust control method based on online estimation of self-vehicle state noise provided by the application aims at the robust control problem of intelligent vehicles in complex driving environments, adopts an online estimation method of vehicle state noise, dynamically updates the vehicle state estimation error based on real-time self-covariance least squares method; according to the distribution of the vehicle state estimation error, the confidence interval such as 99.7% is calculated, so as to define the error boundary and describe the uncertainty range of the vehicle state; combined with the error boundary information, the error system is constrained in the minimum robust positive invariant set, to avoid safety accidents caused by external disturbances or model uncertainties.

[0059] In some embodiments, the noise during the operation of the intelligent vehicle is estimated by using the real-time self-covariance least squares method, including:

[0060] S11, converting the state space equation of the vehicle system into a state space equation based on innovation;

[0061] The state space equation includes:

[0062]

[0063] The state space equation based on innovation includes:

[0064]

[0065] Wherein, k represents the kth moment, x is the state vector of the system, u is the control input of the system, y is the observation of the state space system, is the estimation error, is the estimation value of the state quantity at the kth moment, is the innovation, A, B, and C are the coefficient matrices of the system, w and v are the process noise and observation noise respectively, L is the steady-state Kalman gain, I is the unit matrix,

[0066] The innovation refers to the error between the best prediction value and the measured value according to the existing information.

[0067] S12, for each time window, if the current window is the first window, the estimation error covariance matrix, the process noise covariance matrix and the observation noise covariance matrix initialized according to the empirical information are assigned to the current window, otherwise, the process noise covariance matrix and the observation noise covariance matrix at the initial time of the current window are determined according to the estimation values of the last window;

[0068] As shown in Figure 2 , the initial estimation error covariance P0, the process noise covariance matrix W0 and the observation noise covariance matrix V0 are given according to the empirical information. Wherein the process noise covariance matrix W = E (ww T ), the observation noise covariance matrix V = E (vv T ).

[0069] Determine whether the current window is the first window, if it is the first window, the parameters initialized in S12 are directly assigned to the current window, otherwise, the process noise covariance and the observation noise covariance matrix at the initial time of the current window are updated using the estimation values of the last window, that is, W n = W n-1 , V n = V n-1 .

[0070] S13, in each time window, the Kalman filter is used to update the vehicle state x k and the estimation error covariance matrix at each time until the end of the window, the innovation based on the two state space equations obtained in the current window is calculated to obtain the autocovariance sequence, and a least squares problem is constructed to estimate the process noise covariance matrix and the observation noise covariance matrix.

[0071] At each time, the Kalman filter is used to update the vehicle state x k and the estimation error covariance matrix P k according to the two state space equations of S12. The estimation error covariance matrix will be used to calculate the smallest robust positive invariant set.

[0072] Determine whether the current window is over, if it is not over, continue S13, otherwise, calculate the autocovariance sequence according to the innovation observed in the current window. Determine whether the entire estimation process is over, if it is not over, apply the estimation results of the current window to the next window.

[0073] In some embodiments, the innovation based on the two state space equations obtained in the current window is calculated to obtain the autocovariance sequence, including: calculating the autocovariance sequence Ξ (N) n of length N in the nth window according to the following formula:

[0074]

[0075] where α i is a forgetting factor, and the calculation formula includes:

[0076]

[0077] where m is a parameter, and different values are selected according to different characteristics of the system, and the larger m is, the smaller the weight of the early innovation in the window is.

[0078] In some embodiments, the constructing the least square problem, estimating the process noise covariance matrix and the observation noise covariance matrix includes calculating the covariance matrix of the process noise and the observation noise according to

[0079]

[0080] where, is A LS in the nth window. W n is V n is the estimated value of the process noise covariance matrix and the observation noise covariance matrix in the nth window, S n-1 is a weight matrix, and when n = 1, S n-1 is a unit matrix, and when n > 1, S n-1 The calculation formula of S

[0081]

[0082] Each parameter matrix is:

[0083]

[0084] where the process noise covariance matrix W = E(ww T ), and the observation noise covariance matrix V = E(vv T ).

[0085] In some embodiments, the constructing the linear programming problem, calculating the minimum robust positive invariant set using the set of estimation errors includes:

[0086] The actual vehicle system is divided into a nominal system and an error system by using the control rate , and the state space equation of the error system is obtained:

[0087] e k+1 = (A + BK)e k + w k

[0088] where, x​k and respectively represent the state variables of the actual system and the nominal system, the state variables of the system include lateral velocity v y , yaw rate tracking error e d and heading error

[0089] F ∞ is calculated according to the calculation formula of the minimum robust positive invariant set:

[0090]

[0091] wherein F ∞ is the minimum robust positive invariant set, Θ is the state estimation error set obtained in step S2, ⊕ is the Minkowski sum between sets, A K =A+BK, K is a state feedback gain matrix, and K can be calculated by using the LQR (linear quadratic regulator) method.

[0092] Since the calculation method of the minimum robust positive invariant set is an infinite iteration process, in actual application, an approximate precision can be set, and the approximate solution of the minimum robust positive invariant set is calculated according to the approximate precision. According to the approximate precision, γ and β values meeting the following conditions are selected:

[0093]

[0094] wherein γ is a positive integer, and β ∈ [0, 1).

[0095] An auxiliary calculation set C is constructed to facilitate subsequent calculation of the approximate solution of the minimum robust positive invariant set:

[0096]

[0097] wherein H x and H u are matrices for describing the state constraint X={x|H x x≤b x} of the actual system and the control constraint U={x|H u x≤b u} respectively, represents the support function of Θ to .

[0098] The approximate solution Z of the minimum robust positive invariant set is obtained through recursive calculation, and the calculation formula is:

[0099]

[0100] wherein H p,0 =Hc , b p,0 = b c .

[0101] In some embodiments, the applying the calculated minimum robust positively invariant set to the tube-based model predictive control algorithm to control the vehicle comprises:

[0102] The tube-based model predictive control algorithm divides the actual vehicle system into a nominal system and an error system, and converts the control of the actual system into the control of the nominal system, and when the nominal system is controlled, the objective function comprises:

[0103]

[0104] Wherein, Q and R are weight matrices, and M is a prediction horizon of the model predictive control algorithm.

[0105] When the nominal control quantity is solved, the state constraints and control constraints of the nominal system are strengthened according to the constraints of the actual system, and the constraints are in the form of:

[0106]

[0107] After the nominal control quantity is solved , the control quantity needed to be applied to the actual vehicle system is calculated by .

[0108] Referring to Figure 3 , the application further provides a vehicle robust control device based on online estimation of self-vehicle state noise, comprising: a self-vehicle state noise online estimation module, which estimates the noise in the running process of the intelligent vehicle by using real-time self-covariance least squares method, dynamically updates the vehicle state estimation error, and determines an estimation error set of the vehicle state uncertainty range according to a preset confidence interval of the vehicle state estimation error; a minimum robust positively invariant set calculation module, which is used for constructing a linear programming problem by using the estimation error set, and calculating a minimum robust positively invariant set; and a vehicle robust control module, which is used for applying the calculated minimum robust positively invariant set to a tube-based model predictive control algorithm to control the vehicle.

[0109] The vehicle robust control device based on online estimation of self-vehicle state noise provided by the embodiment of the application has the same implementation principle, technical effects and the like as the aforementioned vehicle robust control method based on online estimation of self-vehicle state noise, and for the sake of brevity, the part of the vehicle robust control device based on online estimation of self-vehicle state noise that is not mentioned in the embodiment can refer to the corresponding content in the aforementioned vehicle robust control method based on online estimation of self-vehicle state noise.

[0110] The application further provides a computer readable storage medium, such as a flash memory, a hard disk, a multimedia card, a card memory (for example, an SD or DX memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read only memory (ROM), an electrically erasable programmable read only memory (EEPROM), a programmable read only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, a server, an App application store, and the like, which stores a computer program, and the program is executed by a processor to realize corresponding functions. The computer readable storage medium of the embodiment is executed by the processor to realize the vehicle robust control method based on online estimation of the self-vehicle state noise.

[0111] It should be noted that, according to the needs of implementation, each step / component described in the application can be split into more steps / components, or two or more steps / components or part of the operation of the steps / components can be combined into a new step / component, so as to realize the purpose of the application.

[0112] The size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiment of the application.

[0113] It should be understood that, for those skilled in the art, improvements or changes can be made according to the above description, and all these improvements and changes should belong to the protection scope of the appended claims of the application.

Claims

1. A vehicle robust control method based on online estimation of vehicle state noise, characterized in that: include: The real-time autocovariance least squares method is used to estimate the noise during the operation of intelligent vehicles and dynamically update the vehicle state estimation error; Determining a set of estimated errors of a vehicle state uncertainty range based on a preset confidence interval of the vehicle state estimated error; Using the estimated error set, constructing a linear programming problem and calculating a minimum robust positive invariant set; The calculated minimum robust positive invariant set is applied to the tube-based model predictive control algorithm to control the vehicle; The method of estimating noise during operation of an intelligent vehicle using a real-time autocovariance least squares method includes: Convert the state space equation of the vehicle system into a state space equation based on new information; For each time window, if the current window is the first window, the estimated error covariance matrix, process noise covariance matrix, and observation noise covariance matrix initialized based on the empirical information are assigned to the current window. Otherwise, the process noise covariance matrix and observation noise covariance matrix at the initial moment of the current window are determined based on the estimated values ​​of the previous window. In each time window, the vehicle state and the estimated error covariance matrix are updated at each moment using the Kalman filter until the end of the window. The autocovariance sequence is calculated based on the new information obtained from the two state-space equations in the current window, and a least squares problem is constructed to estimate the process noise covariance matrix and the observation noise covariance matrix. The method of constructing a linear programming problem using the estimated error set and calculating the minimum robust positive invariant set includes: Utilization control rate , the actual vehicle system is divided into the nominal system and the error system, and the state space equation of the error system is obtained: in, , and Represent the state quantities of the actual system and the nominal system respectively. The state quantities of the system include the lateral velocity , yaw angular velocity , tracking error and heading error ; Calculated according to the calculation formula of the minimum robust positive invariant set : in, is the minimum robust positive invariant set, is the estimated error set, is the Minkowski sum between sets, , is the state feedback gain matrix.

2. The vehicle robust control method based on online estimation of vehicle state noise according to claim 1, characterized in that: The state space equations include: The state space equation based on the new information includes: in, represents the kth moment, is the state vector of the system, is the control input of the system, is the system observation quantity in the state space, is the estimation error, is the estimated value of the state quantity at the kth moment, For the new breath, is the coefficient matrix of the system, and are process noise and observation noise, respectively. , is the steady-state Kalman gain, , is the identity matrix, .

3. The vehicle robust control method based on online estimation of vehicle state noise according to claim 1, characterized in that: The preset confidence interval includes 99.7%.

4. The vehicle robust control method based on online estimation of vehicle state noise according to claim 2, characterized in that: The method of calculating the autocovariance sequence based on the innovation information obtained from the current window based on the two state space equations includes: Calculate the following formula The length of the window is The autocovariance sequence of : in, is the forgetting factor, and the calculation formula includes: ； in, is a parameter, and different values ​​are selected according to the characteristics of different systems. The larger it is, the smaller the weight of new information in the early stage of the window.

5. The vehicle robust control method based on online estimation of vehicle state noise according to claim 4, characterized in that: The least squares problem is constructed to estimate the process noise covariance matrix and the observation noise covariance matrix, including Compute the covariance matrix of process noise and observation noise, where: in, For the In the window , , and For the The estimated values ​​of the process noise covariance matrix and the observation noise covariance matrix in the window, is the weight matrix, when hour, is the identity matrix, when hour, The calculation formula is: The parameter matrices are: Among them, the process noise covariance matrix Observation noise covariance matrix .

6. The vehicle robust control method based on online estimation of vehicle state noise according to claim 1, characterized in that: The method of applying the calculated minimum robust positive invariant set to a tube-based model predictive control algorithm to control the vehicle includes: The model predictive control algorithm based on the tube divides the actual vehicle system into a nominal system and an error system, and transforms the control of the actual system into the control of the nominal system. When controlling the nominal system, the objective function includes: in, and are all weight matrices, It is the prediction time domain of the model predictive control algorithm; When solving the nominal control quantity, strengthen the state constraints of the nominal system according to the constraints of the actual system X and control constraints U , the constraint is expressed as: Where Z is the approximate solution of the minimum robust positive invariant set, and the nominal control quantity is solved Afterwards, use Calculate the control amount that needs to be applied to the actual vehicle system.

7. A vehicle robust control device based on online estimation of vehicle state noise, characterized in that: include: The ego vehicle state noise online estimation module uses the real-time autocovariance least squares method to estimate the noise during the operation of the intelligent vehicle, dynamically updates the vehicle state estimation error, and determines the estimated error set of the vehicle state uncertainty range based on the preset confidence interval of the vehicle state estimation error; A minimum robust positive invariant set calculation module is used to construct a linear programming problem using the estimated error set and calculate the minimum robust positive invariant set; A vehicle robust control module is used to apply the calculated minimum robust positive invariant set to a tube-based model predictive control algorithm to control the vehicle; Among them, the process of estimating the noise during the operation of the intelligent vehicle by the online estimation module of the vehicle state noise using the real-time autocovariance least squares method specifically includes: converting the state space equation of the vehicle system into a state space equation based on new information; for each time window, if the current window is the first window, the estimation error covariance matrix, process noise covariance matrix and observation noise covariance matrix initialized according to the empirical information are assigned to the current window; otherwise, the process noise covariance matrix and observation noise covariance matrix at the initial moment of the current window are determined according to the estimated values ​​of the previous window; in each time window, the vehicle state and the estimation error covariance matrix are updated at each moment using the Kalman filter until the end of the window, and the autocovariance sequence is calculated according to the new information obtained from the two state space equations in the current window, and a least squares problem is constructed to estimate the process noise covariance matrix and the observation noise covariance matrix; Among them, the minimum robust positive invariant set calculation module uses the estimated error set to construct a linear programming problem, and the calculation of the minimum robust positive invariant set specifically includes: using the control rate , the actual vehicle system is divided into the nominal system and the error system, and the state space equation of the error system is obtained: in, , and Represent the state quantities of the actual system and the nominal system respectively. The state quantities of the system include the lateral velocity , yaw angular velocity , tracking error and heading error ; Calculated according to the calculation formula of the minimum robust positive invariant set : in, is the minimum robust positive invariant set, is the estimated error set, is the Minkowski sum between sets, , is the state feedback gain matrix.

8. A computer storage medium, characterized in that A computer program executable by a processor is stored therein, and the computer program executes the vehicle robust control method based on online estimation of vehicle state noise as described in any one of claims 1 to 6.

Images