A controller design method for high-speed trains to achieve robust control of lateral attitude in a multi-source disturbance environment

Abstract

The application discloses a controller design method for realizing lateral attitude robust control of a high-speed train under a multi-source disturbance environment, and is applied to the field of active control of rail transit vehicles. In view of the problem that the control performance of a high-speed train is reduced due to coexistence of multi-source disturbances (including side wind and track irregularity) and sensor measurement errors, a compound system is constructed by fusing a 17-degree-of-freedom whole-vehicle dynamics model, a disturbance model and a measurement error model based on IMU Allan variance analysis. A scalar MEE is extended to a vector / matrix form suitable for multi-state collaborative control together with a UDE framework, so as to realize joint robust control on the lateral displacement and roll angle of the vehicle body. The measurement errors in the speed and angular velocity channels are actively estimated and compensated through the MEE module, so as to effectively suppress the amplification effect of noise in the traditional UDE. Only a single parameter epsilon needs to be adjusted, so as to collaboratively optimize the disturbance suppression capability and error correction.

Description

Technical Field

[0001] This invention belongs to the field of active control of rail transit vehicles, and specifically relates to an anti-interference control technology for high-speed trains. Background Technology

[0002] The continuous increase in high-speed train operating speeds significantly enhances their sensitivity to external disturbances such as crosswinds and track irregularities. These disturbances not only reduce passenger comfort but can also, in severe cases, induce vehicle instability or even derailment, threatening operational safety. To mitigate the effects of these disturbances, current research generally employs active suspension systems to regulate the lateral dynamics of the train body.

[0003] Current mainstream anti-interference control methods include: crosswind compensation strategies based on dynamic matrix adaptation, repetitive learning control for periodic orbital excitation, and neural adaptive fault-tolerant control to handle model uncertainties. However, these methods typically assume ideal sensor measurements during the design process and do not fully consider the random measurement errors present in the inertial measurement unit (IMU) output (such as bias instability of accelerometers and gyroscopes and random walks of velocity / angle). When a low-cost IMU is used in a practical system, such errors will significantly degrade the accuracy of state feedback, leading to a sharp decline in control performance.

[0004] To improve measurement reliability, some studies have introduced Kalman filtering or multi-sensor fusion techniques (such as IMU / odometer / GNSS integrated navigation) to correct errors. However, these methods are highly dependent on auxiliary sensors and are prone to failure in scenarios such as GNSS signal rejection, wheel slippage, or extreme weather, thus limiting their robustness. In addition, although traditional pre-filtering methods can suppress high-frequency noise, the useful signal and measurement error overlap in the frequency domain, making effective separation difficult and limiting their applicability.

[0005] In summary, existing lateral control methods for high-speed trains fail to effectively decouple the relationship between "multi-source disturbance suppression" and "sensor measurement error correction," lacking a unified control framework that can simultaneously ensure stability and accuracy under conditions of strong disturbances and high noise. Therefore, a novel composite control method is urgently needed to achieve coordinated estimation and compensation of disturbances and measurement errors. Summary of the Invention

[0006] To address the significant performance degradation of existing high-speed train lateral control methods under conditions of multi-source disturbances and sensor measurement errors, this invention provides a controller design method for robust lateral attitude control of high-speed trains in multi-source disturbance environments. By constructing a composite model incorporating 17-DOF vehicle dynamics, crosswind, track irregularities, and IMU measurement errors, and extending the scalar MEE+UDE framework to a vector / matrix form, the invention achieves coordinated control of the vehicle's lateral displacement and roll angle. This method suppresses external disturbances and actively corrects measurement errors without relying on auxiliary sensors.

[0007] The technical solution adopted in this invention is: a controller design method for achieving robust lateral attitude control of high-speed trains under multi-source disturbance environments, comprising:

[0008] Constructing a 17-DOF high-speed train dynamics model;

[0009] Constructing system state vectors based on the vehicle dynamics model;

[0010] The system output is obtained from sensor measurements. Considering the measurement errors present in the sensors, the output equation is:

[0011]

[0012] in, Represents the system state vector. This indicates the measurement error present in the sensor. The system output is obtained from sensor measurements;

[0013] Based on expected output The system outputs tracking error. ;

[0014] according to Noise estimation is performed to obtain the measurement error vector. ;

[0015] according to By performing disturbance estimation, the lumped disturbance vector is obtained. ;

[0016] according to , , Design the system control law:

[0017]

[0018] in, Represents the state feedback gain matrix; Represents the control input matrix The false reversal;

[0019] based on To control high-speed trains.

[0020] The beneficial effects of this invention: To address the performance degradation problem of existing high-speed train lateral control methods under the coupling influence of multi-source disturbances and measurement errors, this invention proposes a novel extended MEE and UDE composite control framework. This framework does not rely on multi-sensor fusion or auxiliary navigation systems, and can achieve high-precision joint estimation and compensation of disturbances and errors solely through the onboard IMU. Compared with traditional UDE, repetitive learning control, and other methods, this invention has the following advantages:

[0021] (1) It can effectively suppress multi-source external disturbances such as crosswinds and track irregularities, and actively correct the random measurement errors of the speed and angular velocity channels output by the IMU, thus avoiding control failure caused by ignoring errors in traditional methods;

[0022] (2) By introducing a single adjustment parameter It can intuitively and conveniently coordinate the performance weights of disturbance suppression and error correction, simplifying the controller parameter tuning process;

[0023] (3) Based on Lyapunov stability theory and singular perturbation analysis, the global uniform boundedness of the closed-loop system is rigorously proved, ensuring the reliability and safety of the control strategy in engineering applications. Attached Figure Description

[0024] Figure 1 This invention pertains to the train model;

[0025] Figure 2 Here is the controller schematic;

[0026] Figure 3 Simulation results of this invention;

[0027] Figure 4 Simulation results of this invention;

[0028] Figure 5 Simulation results of this invention;

[0029] Figure 6 Simulation results of this invention. Detailed Implementation

[0030] To facilitate understanding of the technical content of this invention by those skilled in the art, the following description, in conjunction with the accompanying drawings, further illustrates the invention.

[0031] This embodiment focuses on the active suspension system of a high-speed train (HST). The control objective is to achieve lateral displacement of the train body under conditions of multiple external disturbances such as crosswinds and track irregularities, as well as IMU sensor measurement errors. With roll angle High-precision, robust collaborative control. The implementation process of the method of this invention includes the following steps:

[0032] S1. To achieve an accurate description of the lateral dynamics of high-speed trains, a high-speed train-disturbance composite system model is first established, incorporating overall vehicle dynamics, external disturbances, and sensor measurement errors. The model employs the following... Figure 1 The model shown is a classic 17-DOF high-speed train, including the car body, front and rear bogies, and the degrees of freedom of each wheelset. The car body has lateral displacement. lateral angle and roll angle Each bogie (front and rear) has lateral displacement. lateral angle and roll angle , This refers to the bogie serial number, in this embodiment... Each wheelset has lateral displacement. and yaw angle , For the rotation pair number, Taking the vehicle body as an example, the specific dynamic equations are as follows:

[0033] (1)

[0034] (2)

[0035] in, For vehicle body mass, Let x be the moment of inertia of the vehicle body about the x-axis; , The lateral stiffness and damping of the secondary suspension; , The vertical stiffness and damping of the two-stage suspension; , These are the vertical distances between the car body and the bogie, and between the bogie and the wheelset, respectively. The lateral velocity of the vehicle body; The yaw rate of the vehicle body; and The lateral velocities of the front and rear bogies of the car body are respectively. and These are the yaw rates of the front and rear bogies of the car body, respectively. It is half the lateral spacing of the second-stage suspension; , This refers to the active control force acting on the front and rear bogies; and These are crosswind aerodynamic force and rolling torque, respectively. For bogie mass, Let be the lateral acceleration of the i-th bogie. It is half the center pin spacing of the bogie. This is the vertical distance from the center of gravity of the steering frame to the second-stage suspension. The lateral stiffness characteristics of the primary suspension system between the train bogie and wheelsets. This is the vertical distance from the center of gravity of the bogie frame to the center of gravity of the wheelset. The lateral damping characteristics of the primary suspension between the train bogie and wheelset. The active control force output by the actuators on the front / rear bogies. The longitudinal stiffness characteristics of the secondary suspension between the train body and the bogie. The longitudinal damping characteristics of the secondary suspension between the train body and the bogie. The longitudinal stiffness characteristics of the primary suspension system between the train bogie and wheelsets. The yaw angle subscript (2i) for the wheelset is based on the wheelset numbering rule of the bogie, where i=1,2 corresponds to the front and rear bogies. Half the distance between the primary suspension mounting points on the bogie. For the longitudinal damping characteristics of the primary suspension between the train bogie and wheelset, For the wheelbase of the wheelset, The lateral displacement of the wheelset is represented by the subscript (2i−1), which is based on the odd numbering rule of the wheelset based on the bogie, where i=1,2 correspond to the front and rear bogies, respectively. For the lateral speed of odd-numbered wheelsets of high-speed trains, The vertical distance from the center of gravity of the steering frame to the second-stage suspension. Half the center pin spacing of the bogie For the quality of high-speed train wheelsets, For the lateral acceleration of high-speed train wheelsets, For the yaw angle of high-speed train bogies, For the wheel-rail creep coefficient of the second axle and second column, For train running speed, For lateral irregularity excitation of the track, For the wheel rolling circle radius of the high-speed train wheelset, For the track tilt angle, For the bogie's yaw angle, For longitudinal creep coefficient, For equivalent creep coefficient, For wheelset half gauge, This is the lateral tilt angle of the track.

[0036] Similarly, the dynamic equations for the bogie and wheelset can be derived.

[0037] Bogie dynamic equations:

[0038]

[0039]

[0040]

[0041] Wheelset dynamics equations:

[0042]

[0043]

[0044] To focus on lateral and roll control of the vehicle body, the lateral displacement of the vehicle body is extracted from the above equations. With roll angle Its derivatives constitute the system state vector:

[0045] (3)

[0046] in, This is the transpose of the matrix.

[0047] The control input vector is:

[0048] (4)

[0049] in, Let the active driving force acting on the front and rear bogies be represented. Then the system state equation can be written as:

[0050] (5)

[0051] in, yes The first derivative, and In this embodiment, the nominal system matrix is... It should be understood as a real number matrix, for example It is a 4x2 real number matrix. This is the lumped perturbation vector. Given an m x n matrix, where each element is a real number, the nominal system matrix is... and The specific form is determined by the vehicle's suspension parameters:

[0052] Wherein, the subscript "0" indicates the nominal parameter, For the lateral stiffness and damping of the secondary suspension. For the longitudinal stiffness and damping of the second-stage suspension. For the vehicle body mass and rolling inertia, The vertical distance from the suspension point. This refers to the lateral spacing of the suspension.

[0053] Lumped perturbation vector Includes model uncertainty and external input disturbances ,Right now:

[0054] (6)

[0055] in, Time represents an independent variable of the system.

[0056] S2. The system output is obtained from sensor measurements. Considering the measurement error of the IMU, the output equation is:

[0057] (7)

[0058] in, The sensor measures the output vector. This is the measurement error vector. Here, we assume lateral displacement. and roll angle The measurement error is negligible, while the lateral velocity and roll rate Measurement error and It is then significantly affected by the random error of the IMU.

[0059] Measurement error and Modeling was performed using Allan ANOVA based on IMU, primarily involving two types of random errors: bias instability and velocity / angle random walk. Bias instability was described using a first-order Gaussian-Markov process, and random walk was described using white noise. Specific model parameters (such as correlation time) were also considered. noise spectral density (etc.) can be obtained from the IMU datasheet or through experimental identification.

[0060] Measurement speed error The expression is:

[0061]

[0062] in, The accelerometer bias instability error follows a first-order Gaussian-Markov process. This represents the velocity random walk error;

[0063] The expression is:

[0064]

[0065] in, , It is a unit intensity Gaussian white noise;

[0066] The expression is: Its power spectral density is .

[0067] For the IMU used in this invention, the calibration parameters are as follows in the HST application scenario:

[0068]

[0069] Measurement angular velocity error The expression is:

[0070]

[0071] in, The accelerometer bias instability error follows a first-order Gaussian-Markov process. This refers to the random walk error caused by acceleration.

[0072] The expression is:

[0073]

[0074] in, , It is a unit intensity Gaussian white noise;

[0075] The expression is: Its power spectral density is For IMU parameters:

[0076]

[0077] S3, Crosswind Disturbance Model, based on quasi-steady theory, describes the lateral aerodynamic forces acting on the vehicle body. and rolling moment The model is as follows:

[0078] (10)

[0079] in, air density, For reference area, For reference height, Relative wind speed, For the effective yaw angle, is the aerodynamic coefficient. To test robustness, a standard turbulent crosswind model with an average wind speed of 20 m / s can be used.

[0080] S4. Next, we will design an extended MEE+UDE composite controller. The core of this invention lies in the scalar form of MEE (measurement error estimator). Figure 2 Noise estimator in the context of noise estimation and UDE (uncertainty and disturbance estimator). Figure 2 The disturbance estimator framework is extended to vector / matrix form to handle multi-state control problems simultaneously. The control law is designed as follows:

[0081] (11)

[0082] in, To output the tracking error, For the desired output, , The measurement error vector estimated by MEE. This is the lumped perturbation vector estimated by UDE. The state feedback gain matrix is ​​designed using the Linear Quadratic Regulator (LQR) method to ensure... For Hurwitz, That is, the nominal controller output ; To control the input matrix The false rebellion.

[0083] S5. After the controller design is completed, the MEE (Measurement Error Estimator) is designed. The core task of the measurement error estimator is to extract the output signal from the sensor that is contaminated by noise. In the process, the speed measurement error is reconstructed in real time. Measurement error of angular velocity , denoted as: This provides "cleaned" state feedback for subsequent disturbance estimation and control compensation. Its design follows a transformation process from a time-domain unrealizable form to a frequency-domain realizable form, ultimately culminating in a specific filtering algorithm. First, a pre-filter is introduced to process the noise-contaminated sensor measurement signal. Perform low-pass filtering, that is: in, Indicates sensor output The Laplace transform result, This is the transfer function of the preprocessing filter (usually a low-pass filter). Then, the MEE estimates and compensates for the measurement error's filtering residual, reducing the MEE's bandwidth design requirements. The noisy measurement signal within the filtered bandwidth is... According to the system model, the noisy output equation is: Decompose the state and error vectors:

[0084] (12)

[0085] in, Corresponding to precisely measurable displacement and angle. The corresponding velocity and angular velocity are affected by noise pollution. From this, the frequency domain relationship can be obtained:

[0086] (13)

[0087] because (Under zero initial conditions), the actual measurement error can be expressed as:

[0088] (14)

[0089] Equation (14) contains pure differential operators Direct implementation would amplify high-frequency noise and is physically impossible. To address this issue, a stable and strictly regular filter is introduced. The estimator can be constructed as follows:

[0090] (15)

[0091] filter Two key conditions must be met:

[0092] Regularity: Its relative order is at least 1 to ensure It is a true rational expression.

[0093] Stability: All poles are located in the left half of the complex plane, ensuring the stability of the estimator itself.

[0094] To achieve simplicity and reliability, this invention employs a first-order low-pass filter structure:

[0095] (16)

[0096] in, The time constant of MEE is a key adjustable parameter that determines the estimation performance. It is a second-order identity matrix.

[0097] Estimation principle analysis: Substituting equation (16) into (15), we get:

[0098] (17)

[0099] This structure can be viewed as: from noisy velocity / angular velocity signals In the middle, subtract the differential signal obtained by approximation through a first-order filter. Thus, the error estimate is extracted. .

[0100] Finally, the complete measurement error estimation vector used for controller feedforward compensation is:

[0101] (18)

[0102] The estimated value This will be directly used to correct the state feedback (and in the control law). (corresponding to the item), and embedded as a key input into the subsequent UDE design to block the propagation path of measurement noise in the disturbance estimation channel.

[0103] S6. Then, design the UDE (Uncertainty and Disturbance Estimator), which aims to estimate the lumped disturbances acting on the system. Including model uncertainty and external disturbances Its design needs to address the inherent problem of classical UDEs being sensitive to measurement errors.

[0104] Perform a Laplace transform on state equation (5):

[0105] (19)

[0106] in, , , They represent The frequency domain expression, This represents the initial state of x.

[0107] The perturbation expression can be obtained by rearranging:

[0108] (20)

[0109] in, Represents the identity matrix with the same order as A;

[0110] Classical UDE, through frequency domain pairing Apply a low-pass filter To obtain the estimated value:

[0111] (twenty one)

[0112] The initial condition term It is usually omitted or considered zero. The true state... Substitute into equation (21):

[0113] (twenty two)

[0114] In equation (22), the measurement error matrix Its function, its high-frequency gain is approximately as follows: When using a high cutoff frequency (small )of When attempting to accurately estimate disturbances, this high-frequency component significantly amplifies noise in the measurement error, leading to problems in disturbance estimation. Distortion leads to a deterioration in control performance.

[0115] To overcome the above-mentioned shortcomings, this invention proposes to improve the output of MEE. As a real-time compensation term for measurement error, it is embedded into the estimation channel of the UDE. Specifically, in equation (22), using... Replace the originally unknowable The improved UDE estimation law is obtained:

[0116] (twenty three)

[0117] This can be considered a state feedback signal that has undergone preliminary "purification" by MEE. By compensating for measurement errors in advance, it effectively cuts off the error transmission... The amplified path. A first-order low-pass filter is also chosen to ensure stability and simplicity:

[0118] (twenty four)

[0119] Represents a fourth-order identity matrix;

[0120] control law Substituting and rearranging, we can obtain the solution. The closed-loop implementation form:

[0121] (25)

[0122] When the system is under ideal measurement conditions, i.e. At that time, the estimated output of MEE At this point, the UDE estimation law (25) of the present invention degenerates into:

[0123] (26)

[0124] This is precisely the standard form under the classic UDE control framework. This indicates that the improved design of this invention is a generalized extension of the classic UDE under non-ideal measurement conditions, possessing good backward compatibility. Since train-mounted sensors cannot directly measure all system states, and UDE design requires full-state information for accurate disturbance estimation, a Luenberger State Observer (LSO) is introduced to reconstruct unmeasurable states and provide state estimates for the UDE. The dynamic equation of the LSO is:

[0125] (27)

[0126] in This is the state estimation vector. Let be the observer gain matrix, and design such that ( Hurwitz guarantees that the estimation error converges.

[0127] LSO and UDE form a two-way coupling structure: on the one hand, LSO provides state estimation for UDE. , so that in the UDE expression (26) Items can be based on On the one hand, the perturbation estimated by UDE is realized; on the other hand, the perturbation is estimated by UDE. Feedback is sent to the LSO, and a compensation term is added to the observer equation. This is to actively counteract the impact of disturbances on state estimation and improve the estimation accuracy of LSO under disturbed environments.

[0128] S7. To demonstrate the effectiveness of the proposed control method, a rigorous theoretical analysis of the closed-loop system is conducted based on Lyapunov stability theory and singular perturbation analysis. The state tracking error is defined as... ,in The desired trajectory is defined as follows: The measurement error estimation error is defined as... The disturbance estimation error is .

[0129] Combining the dynamic equations of the controlled object Output equation and the control law proposed in this invention After derivation, the dynamic equation for the closed-loop tracking error can be obtained as follows:

[0130] (28)

[0131] Meanwhile, based on the frequency domain design expressions of MEE and UDE (Equations (18) and (26)), the estimation error can be obtained through the inverse Laplace transform. and The dynamic equation:

[0132] (29)

[0133] (30)

[0134] Both are coupled with equation (28). Among them, The coefficient matrix, For external terms, and The expression is:

[0135] To reveal the filter time constant To fundamentally impact system performance and simplify the analysis, a singular perturbation parameter is introduced. And establish the following parameter mapping relationship:

[0136] (31)

[0137] in This is the normalization constant. The physical meaning of this mapping lies in the fact that it uses a single parameter... Simultaneously adjust the bandwidth of MEE and UDE.

[0138] Furthermore, auxiliary variables representing rapidly changing dynamics are constructed. By using the mapping relationship of formula (31) to scale and reorganize the variables of the coupled error dynamic system, it can be accurately transformed into the following standard singular perturbation form:

[0139] (32)

[0140] (33)

[0141] Based on the above singular perturbation model, Lyapunov stability theory and singular perturbation theory are used to prove its stability.

[0142] Proof of stability of boundary layer systems (rapid dynamics), boundary layer systems correspond to rapidly changing states. The dynamics, through time scale transformation and order get:

[0143] (34)

[0144] This equation describes when When the estimation error behaves in the "boundary layer" (i.e., the initial transient phase), it is called a boundary layer system.

[0145] It needs to be proven that the origin of the boundary layer system (34) is globally asymptotically stable. To this end, the following Lyapunov function candidate is constructed:

[0146] (35)

[0147] in, It is a symmetric positive definite matrix with the following block diagonal form:

[0148] here These are parameters to be determined. To ensure... For a positive definiteness, it must satisfy the following conditions: And from the bottom right corner Positive definiteness requirement of block matrix .

[0149] The parameters to select are as follows:

[0150] (36)

[0151] It is easy to verify that when Sometimes, and ,therefore It is positive.

[0152] calculate The derivative along the trajectory of the boundary layer system (34):

[0153] (36)

[0154] matrix and Substituting the specific form, we can calculate the following:

[0155] (38)

[0156] because The top-left block matrix is ​​negative definite; because The middle block matrix is ​​negative definite; for the lower right block, substituting the selected parameters yields:

[0157] (39)

[0158] Therefore, the entire matrix (25) is semi-negative definite. Further analysis shows that on non-zero trajectories, Strictly negative. And, when hour, The radial unbounded condition is satisfied. Therefore, according to Lyapunov's second method, the origin of the boundary layer system (34) is globally asymptotically stable.

[0159] Reduced-order system (slowly variable dynamics), qualitative analysis. The reduced-order system is achieved by setting... And solve for the quasi-steady state of the fast exit state. (Right now Substituting these into formula (28) together, we get:

[0160] (40)

[0161] Because this process eliminates rapid dynamics The system's order is reduced, hence it is called a reduced-order system. This reflects the dominant behavior of the system over long timescales. Because the design has already ensured... For Hurwitz, the system is input-state stable. When external disturbances occur... When bounded, state bounded; if If the origin is exponentially stable, then the origin is exponentially stable.

[0162] According to singular perturbation theory, since both the boundary layer system and the reduced-order system are exponentially stable and satisfy the necessary smoothness condition, there exists a critical value. This makes it possible for all The origin of a complete singularly perturbated system (32)(33) is exponentially stable. Combining this with the assumption that the external input signal is bounded, it can be deduced that all signals in the closed-loop system are globally uniformly bounded: the true state of the system... All estimated signals and tracking and estimation errors All are globally consistent and bounded.

[0163] Furthermore, through quantitative analysis of error dynamics, the final boundaries and parameters of estimation error and tracking error can be established. Explicit relations. It has been proven that there exist positive constants. This makes it possible to select satisfy:

[0164] (41)

[0165] This guarantees the final bounded performance described in the theorem: the estimation error converges exponentially to an adjustable boundary. That is, there exists a... Related boundary functions This makes it possible for all ,satisfy:

[0166] (42)

[0167] And when adjusting parameters Sometimes, , .

[0168] The ultimate boundedness of tracking error: The state tracking error converges exponentially to an adjustable boundary. That is, there exists a... Related boundary functions This makes it possible for all ,satisfy:

[0169] (43)

[0170] And when adjusting parameters Sometimes, .

[0171] Equation (31) clearly reveals the adjustment parameters For control accuracy ( () has a direct regulatory effect.

[0172] The above conclusions theoretically guarantee that the proposed extended MEE+UDE controller can achieve stable and accurate tracking of the lateral attitude of high-speed trains under the coexistence of multi-source disturbances and measurement errors. Furthermore, simulation comparisons are conducted between the extended MEE+UDE controller of this invention and excellent high-speed train lateral controllers in the field, such as... Figure 3 , Figure 4 , Figure 5 , Figure 6 As shown, the method of the present invention exhibits superior performance in indicators such as lateral displacement and lateral acceleration of the high-speed train body.

[0173] Those skilled in the art will recognize that the embodiments described herein are for the purpose of helping to understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Various modifications and variations can be made to the invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the scope of the claims of the invention.

Claims

1. A controller design method for achieving robust lateral attitude control of high-speed trains under multi-source disturbance environments, characterized in that, include: Constructing a 17-DOF high-speed train dynamics model; Constructing system state vectors based on the vehicle dynamics model; The system output is obtained from sensor measurements. Considering the measurement errors present in the sensors, the output equation is: ； in, Represents the system state vector. This indicates the measurement error present in the sensor. The system output is obtained from sensor measurements; Based on expected output The system outputs tracking error. ; according to Noise estimation is performed to obtain the measurement error vector. ; according to By performing disturbance estimation, the lumped disturbance vector is obtained. ; according to , , Design the system control law: ； in, Represents the state feedback gain matrix; Represents the control input matrix The false reversal; It is used as a control input vector to control the high-speed train.

2. The controller design method for robust lateral attitude control of high-speed trains under multi-source disturbance environment according to claim 1, characterized in that, System state vector Represented as: ； in, Indicates the lateral displacement of the vehicle body. Indicates the roll angle. express The derivative, express The derivative, This is the transpose of the matrix.

3. A controller design method for robust lateral attitude control of high-speed trains under multi-source disturbance environment according to claim 2, characterized in that, According to Noise estimation is performed to obtain the measurement error vector. Specifically, this is achieved by constructing a measurement error estimator, the construction process of which is as follows: right Performing a Laplace transform, we obtain ; Will The decomposition yields: ； in, Corresponding to precisely measurable displacement and angle, Corresponding to the speed and angular velocity affected by noise pollution, , This is the transpose of the matrix. The speed measurement error to be reconstructed by the measurement error estimator The measurement error of the angular velocity to be reconstructed by the measurement error estimator; The frequency domain relationship is obtained: ； Under zero initial conditions, due to The actual measurement error can then be expressed as: ； Introduce a stable and strictly regular filter The following measurement error estimator is constructed: ； in, express The estimation results, This represents a pure differential operator.

4. The controller design method for robust lateral attitude control of high-speed trains under multi-source disturbance environment according to claim 3, characterized in that, Using a first-order low-pass filter structure, the expression is: ； in, For the time constant of the measurement error estimator, , It is a second-order identity matrix.

5. A controller design method for robust lateral attitude control of high-speed trains under multi-source disturbance environment according to claim 4, characterized in that, According to By performing disturbance estimation, the lumped disturbance vector is obtained. Specifically, this is achieved by constructing a noise estimator, the construction process of which is as follows: Based on system state vector With system control input vector The system state equations can be expressed as: ； in, and For the nominal system matrix, This is the lumped disturbance vector; right Performing a Laplace transform, we get: ； The resulting perturbation expression is: ； in, Represents the identity matrix with the same order as A; Classical noise estimators work by adjusting the frequency domain... Apply a low-pass filter To obtain the estimated value: ； Will Substitute ,get: ； The output of the measurement error estimator As a real-time compensation term for measurement error, it is embedded into the estimation channel of the noise estimator, resulting in: 。 6. A controller design method for robust lateral attitude control of high-speed trains under multi-source disturbance environment according to claim 5, characterized in that, Using a first-order low-pass filter structure, the expression is: ； in, The time constant of the noise estimator. , This represents a fourth-order identity matrix.

7. A controller design method for robust lateral attitude control of high-speed trains under multi-source disturbance environment according to claim 6, characterized in that, Introducing a singular perturbation parameter , And establish the following parameter mapping relationship: ； By a single parameter Simultaneously adjust the bandwidth of the measurement error estimator and the noise estimator.

Images