Active fault-tolerant control method and system for steer-by-wire of heavy-duty vehicle

Abstract

The invention provides an active fault-tolerant control method and system for steer-by-wire of a heavy-duty vehicle, designs a rolling time domain estimator aiming at execution performance limitation of a steer-by-wire system of an automatic driving heavy-duty vehicle, can explicitly process system constraints, avoids adverse effects caused by oscillation and overshoot in an estimation convergence process, and improves the control accuracy of the steer-by-wire of the heavy-duty vehicle. The smoothness of an estimated curve and the optimality of an estimated value are ensured; meanwhile, the invention further provides an active compensation fault-tolerant control framework, a comprehensive control law can be generated through steering execution performance limit estimation in a fixed time domain, compensation and suppression of performance limit and model mismatch are achieved, and the method has the advantages of being low in conservative degree and high in robustness.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle control, and particularly to an active fault-tolerant control method and system for steer-by-wire of heavy-duty vehicles. Background Art

[0002] As a core operation equipment in the fields of energy exploitation, mineral transportation, etc., the autonomous driving technology of heavy-duty vehicles is the key breakthrough point for realizing safety protection in high-risk working conditions and improving operation efficiency. As the core actuator for vehicle lateral control, the functional safety performance of the steer-by-wire system directly determines the control accuracy of the vehicle's lateral-yaw-roll composite dynamics. However, in engineering practice, the steer-by-wire system of autonomous heavy-duty vehicles may have performance limitations in execution, mainly manifested as the deviation between the wheel angle and the desired command, ultimately resulting in significant lateral path tracking errors. This phenomenon may stem from the measurement error of the angular displacement sensor system in the steer-by-wire system, or may be caused by the clearance between mechanical components such as gears or racks. In addition, this deviation may also result from the error in the vehicle assembly process, or even sudden factors such as compressive deformation, mechanical wear, or sudden change in the tire pressure of a single side caused by external force collision in a harsh working environment. In heavy-duty working conditions, it will cause a reduction in control accuracy and operation efficiency at least, and may even induce vehicle instability or rollover accidents at worst, significantly increasing the safety risk coefficient of the autonomous driving system.

[0003] How to accurately evaluate the performance limitations of the steer-by-wire system of autonomous heavy-duty vehicles and trace the source of deviation factors for correction or compensation and suppression is the key to reducing driving risks and improving operation efficiency. The commonly used method in engineering is to perform four-wheel alignment on the vehicle to eliminate part of the steering deviation. However, there are two inherent defects in the offline calibration mode for a large number of heavy-duty vehicles: one is that it is necessary to interrupt the operation process for manual detection, which is difficult to meet the requirements of continuous production; the other is that it cannot perceive the real-time performance limitations under dynamic working conditions, restricting the large-scale deployment and application of the autonomous driving system. To solve the above limitations, current research mainly proposes two types of online solutions from the control algorithm level: passive robust fault-tolerant control strategies and active compensation fault-tolerant control strategies.

[0004] The passive robust fault-tolerant control strategy generally first obtains the boundary of the steering system performance limitation through empirical knowledge, and then designs a robust fault-tolerant controller using this boundary information to suppress the adverse effects of performance limitation injection on the closed-loop control system. However, setting the boundary of the steer-by-wire performance limitation through empirical knowledge, if set too tightly, may not cover all potential values of the performance limitation, increasing the risk of instability of the closed-loop system; if set too loosely, the designed controller will inevitably be conservative, which is not conducive to improving the control accuracy. Therefore, the above-mentioned robust fault-tolerant control strategy needs to reasonably assume the performance limitation, and it is difficult to balance strong robustness and low conservatism, and is not very suitable for working conditions with complex and changeable driving environments.

[0005] Active compensation fault-tolerant control strategies usually include an estimator or an observer for implementing and updating the performance limits of the steer-by-wire system, and then designing a controller based on the estimated values of these limits. For example, in reference [1] Fekih A, Devariste D. A fault-tolerant steering control design for automatic path tracking in autonomous vehicles [C] / / 2013 American control conference. IEEE, 2013:5146-5151., an adaptive diagnostic observer was designed to obtain an estimate of the performance limits of the steering system, and then it was combined with state feedback control to achieve active compensation control; in reference [2] Zhang L, Wang Z, Ding X, et al. Fault-tolerant control for intelligent electrified vehicles against front wheel steering angle sensor faults during trajectory tracking [J]. IEEE Access, 2021, 9:65174-65186., the extended Kalman filter was used to estimate the true vehicle steering angle, and then combined with a model predictive controller to achieve active fault-tolerant control.

[0006] The above-mentioned active compensation control strategy can, to a certain extent, achieve real-time estimation and compensation of the actuator performance limits of the steer-by-wire system, but there are the following limitations: 1. Estimators such as Kalman filters are very common in active compensation control strategies, but their optimality depends on linear models and Gaussian noise assumptions. This assumption is generally difficult to meet in actual vehicle control systems, so the optimality of the estimated values cannot be guaranteed; 2. Some observers can also achieve the estimation of steering performance limits, but there may be overshoot and oscillation phenomena during the convergence process of the estimation error. Using these estimated values to design an active compensation controller may have an adverse impact on the system stability. Summary of the Invention

[0007] (1) Technical problems to be solved

[0008] Based on this, the present invention provides a method and a system for active fault-tolerant control of the steer-by-wire system of heavy-duty vehicles to solve the problems that the passive robust fault-tolerant control strategy in the background technology is difficult to achieve the unity of low conservatism and strong robustness, and the existing active compensation control strategy cannot guarantee the optimality of the estimated values and is difficult to handle system constraints.

[0009] (2) Technical Solution

[0010] To achieve the above object, on the one hand, the present invention provides an active fault-tolerant control method for steer-by-wire of heavy-duty vehicles, including:

[0011] S1: Establish a discretized vehicle kinematic model;

[0012] S2: Design an estimator based on a rolling horizon according to the execution performance limit of the model;

[0013] S3: Obtain a control compensation amount according to the optimal steering estimation value of the estimator and the historical final desired steering control amount;

[0014] S4: Design a model predictive controller based on path tracking;

[0015] S5: Obtain the final desired steering control amount according to the optimal desired control amount of the model predictive controller and the control compensation amount;

[0016] S6: Implement active compensation fault-tolerant control for the vehicle according to the final desired steering control amount;

[0017] In S2, the vehicle pose state measured by the integrated inertial navigation at time k is denoted as and is represented by N E to represent the estimation time domain; is sent into a memory with a fixed capacity of N E +1, and the pose state at time k-N E -1 is removed At this time and the states at [k-N E , k-1] together form the historical vehicle pose state sequence

[0018] The final desired steering control amount obtained at time k-1 is denoted as u f (k-1); u f (k-1) is sent into a memory with a fixed capacity of N E , and the final desired steering control amount u E -1 at time k-N f (k-N E -1) is removed; at this time, u f (k-1) and the control amounts at [k-N E , k-2] together form the historical final desired steering control amount sequence

[0019] For the execution performance limit of the steer-by-wire system, the following rolling horizon estimation problem is constructed:

[0020]

[0021] Among them, the cost function J NMHE The first term penalizes the deviation between the estimated state and the measured state; the second term penalizes the deviation between the estimated steering angle and the desired steering angle; the third term is the arrival cost; S1, S2, and W are weight matrices of appropriate dimensions; the first constraint represents the vehicle nominal kinematic model; the second constraint is the vehicle actuator saturation constraint, indicating that the estimated true steering angle should be within the limit range [σ min , σ max that can be achieved by the steer-by-wire system; the third constraint indicates that the estimated value of the performance limit of the steer-by-wire system should be within a reasonable range [u c,min , u c,max , where u c,min and u c,max represent the minimum and maximum values of the control compensation amount.

[0022] Furthermore, in S3, solve the receding horizon estimation problem to obtain the optimal solution sequence and use it as the optimal steering angle estimation value sequence within a fixed time domain length That is:

[0023]

[0024] Combined with the historical final desired steering control amount sequence in the memory The solution formula for the control compensation amount u c can be obtained:

[0025]

[0026] Furthermore, in S4, traverse the reference path point sequence generated by the path planner according to the measured vehicle pose state to obtain the reference path point closest to the vehicle centroid coordinates (X c , Y c ), denoted as the reference matching point P0; then, drive along the reference path at the current speed v for a time step T to obtain the next projection point P1; by repeating this process, a set of reference matching points P are generated within the prediction control time domain N

[0027] According to the vehicle centroid position (X c , Y c ) and the reference matching points, the lateral error e y and the heading angle error e ψ can be calculated:

[0028]

[0029] where (X r , Y r ) are the coordinates of the reference matching point, and ψ r is the heading angle of the tangent line of the reference matching point. Assuming the control error e = [e y , e ψ T , the above equation is simplified to e(k) = g(ζ(k), ζ r (k));

[0030] Establish the following model predictive controller based on path tracking:

[0031]

[0032] s.t. ζ(k|k) = ζ(k),

[0033] ζ(k + i + 1|k) = f(ζ(k + i|k), u(k + i|k)),

[0034]

[0035] u(k - 1|k) = u * (k - 1|k - 1),

[0036]

[0037] where N P represents the prediction control time domain, represents the terminal constraint set; the first term in the cost function J NMPC is the trajectory tracking error cost, the second term is the steering angle increment change cost, and the third term is the terminal cost; Q, R, and P are weight matrices of appropriate dimensions; this model predictive controller includes the following eight constraint equations: the first constraint is the state initialization constraint, indicating that the state at the current moment will be updated before each optimization solution; the second constraint is the vehicle kinematic constraint; the third constraint is the path tracking error calculation constraint; the fourth and fifth constraints are the calculation constraints of the steering angle control increment; the sixth constraint is the control compensation limit constraint, ensuring that the control amount acting on the vehicle steering angle does not violate the input saturation constraint, where δ min , δ max represent the minimum and maximum steering angles allowed by the actuator, u c,min and u c,max represent the minimum and maximum values of the control compensation amount; the seventh constraint is the steering angle increment limit constraint, where Δδ min and Δδ max represent the maximum allowable change in the steering angle within a time step T; the eighth constraint is the terminal constraint.

[0038] ​Further, in S5, solve the model predictive controller problem, and take the first element in the obtained optimal control sequence as the optimal desired control quantity u of the k-th moment d (k):

[0039] u d (k) = u*(k|k)

[0040] Calculate the control compensation quantity u c (k), and then calculate the final desired steering control quantity u f (k) according to the following formula:

[0041] u f (k) = u d (k) - u c (k).

[0042] Further, in S1, establish the discretized vehicle kinematic model as shown below:

[0043]

[0044] Among them, represents the discrete time point; (X, Y) represents the horizontal and longitudinal coordinates of the position of the center of the rear axle of the vehicle in the geodetic coordinate system; ψ represents the heading angle of the vehicle; v represents the longitudinal speed of the vehicle; T represents the discrete time step; L represents the wheelbase; δ f represents the front wheel steering angle; Abbreviate the above formula to the following form:

[0045]

[0046] Among them, ζ represents the vehicle state, that is, ζ = [X, Y, ψ] T ; u represents the control quantity, that is, the front wheel steering angle δ f ; represents the set of positive integers from 0 to infinity.

[0047] Further, denote the distance between the installation position of the integrated inertial navigation and the center of the rear axle of the vehicle as L', and establish the relationship between the position of the integrated inertial navigation and the center of the rear axle of the vehicle:

[0048]

[0049] Among them, (X s , Y s ) represents the position coordinates of the integrated inertial navigation;

[0050] Denote the distance between the position of the vehicle center of mass and the installation position of the integrated inertial navigation as L'', and establish the relationship between the position of the vehicle center of mass and the installation position of the integrated inertial navigation as follows:

[0051]

[0052] Among them, (X c , Y c ) is the coordinate of the vehicle's center of mass position.

[0053] On the other hand, the present invention provides an active fault-tolerant control system for steer-by-wire of heavy-duty vehicles, including a vehicle system, a perception module, a planning module, and a control module connected in sequence;

[0054] The vehicle system includes an actuator for making a response according to the final desired steering control amount u f output by the control module, thereby changing the vehicle state;

[0055] The perception module includes an integrated inertial navigation system for obtaining the vehicle pose state in real time The vehicle pose state includes the vehicle longitudinal speed v, the position coordinates (X s , Y s ) of the integrated inertial navigation system, and the heading angle ψ of the vehicle;

[0056] The planning module includes a path planner for generating a sequence of reference path points [ζ according to the vehicle pose state output by the perception module r and transmitting [ζ r to the control module;

[0057] The control module includes a delay unit, an estimator, a controller, and a compensation amount calculation unit;

[0058] Among them, the controller includes a model predictive controller for receiving the sequence of reference path points [ζ r output by the planning module and the vehicle pose state output by the perception module and generating a desired control amount u d ; the delay unit is used to temporarily store the previous final desired steering control amount u f and transmit it to the storage in the estimator for use;

[0059] The estimator includes a storage and a moving horizon estimator. The storage is used to receive the real-time vehicle pose state output by the perception module and the historical final desired steering control amount u f output by the control module, and output a sequence of historical final desired steering control amounts [u f with a fixed length and a sequence of historical vehicle pose states The moving horizon estimator outputs a sequence of real vehicle steering estimates within a period of time according to the output information of the storage

[0060] The compensation amount calculation unit is used to calculate according to the sequence of historical final desired steering control amounts [u fand the sequence of true vehicle steering estimation values Calculate the control compensation amount u c ; Then combine u c and the desired control amount u d Output the final desired steering control amount u at the current moment f .

[0061] (III) Beneficial effects

[0062] As can be seen from the above technical solutions, an active fault-tolerant control method and system for steer-by-wire of heavy-duty vehicles proposed by the present invention have the following beneficial effects:

[0063] 1. The rolling horizon estimator provided by the present invention can explicitly handle system constraints, ensure that the estimated values of the performance limitations of the steer-by-wire system are always within a reasonable range, and avoid adverse effects on the closed-loop control system; moreover, without any strict assumptions, it can ensure the optimality of the estimated values within a certain time domain, and is more accurate and smooth compared with existing estimation methods.

[0064] 2. Using the estimated values to obtain the actuator performance limitations within a fixed time domain and combining them with the model predictive controller can not only offset the system deviation caused by the steering performance limitations, but also suppress the adverse effects of model mismatch on path tracking control, and significantly improve the path tracking accuracy and driving safety. Description of the drawings

[0065] The features and advantages of the present invention will be more clearly understood by referring to the accompanying drawings. The drawings are schematic and should not be construed as imposing any limitations on the present invention. In the drawings:

[0066] Figure 1 is a schematic diagram of the selection of vehicle reference matching points of the present invention;

[0067] Figure 2 is a schematic diagram of the architecture of the active fault-tolerant control system for steer-by-wire of heavy-duty vehicles of the present invention. Detailed implementation manners

[0068] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0069] The present invention provides an active fault-tolerant control method for steer-by-wire of heavy-duty vehicles, including:

[0070] S1: Establish a discretized kinematic model of the vehicle;

[0071] S2: Design an estimator based on the rolling time domain according to the execution performance limit of the model;

[0072] S3: Obtain a control compensation amount according to the optimal steering estimation value of the estimator and the historical final desired steering control amount;

[0073] S4: Design a model predictive controller based on path tracking;

[0074] S5: Obtain the final desired steering control amount according to the optimal desired control amount of the model predictive controller and the control compensation amount;

[0075] S6: Implement active compensation fault tolerance control for the vehicle according to the final desired steering control amount.

[0076] Specifically, in S1, establish a discretized vehicle kinematic model as follows:

[0077]

[0078] Among them, represents the discrete time point; (X, Y) represents the horizontal and longitudinal coordinates of the center position of the vehicle's rear axle in the earth coordinate system; ψ represents the vehicle's heading angle; v represents the vehicle's longitudinal speed; T represents the discrete time step; L represents the wheelbase; δ f represents the front wheel steering angle. Abbreviate formula (1) to the following form:

[0079]

[0080] Among them, ζ represents the vehicle state, that is, ζ = [X, Y, ψ] T ; u represents the control amount, that is, the vehicle front wheel steering angle δ f ; represents the set of positive integers from 0 to infinity (the following symbol in the form of represents the set of positive integers from x to y).

[0081] Denote the distance between the installation position of the integrated inertial navigation and the center of the vehicle's rear axle as L', and establish the relationship between the position of the integrated inertial navigation and the center position of the vehicle's rear axle:

[0082]

[0083] Among them, (X s , Y s ) represents the position coordinates of the integrated inertial navigation.

[0084] Denote the distance between the vehicle's center of mass position and the installation position of the integrated inertial navigation as L'', and establish the relationship between the vehicle's center of mass position and the installation position of the integrated inertial navigation as follows:

[0085]

[0086] Among them, (X c , Y c ) is the coordinate of the vehicle's center of mass position.

[0087] In S2, the vehicle pose state measured by the integrated inertial navigation at time k is denoted as and is represented by N E to denote the estimation time domain. Send into a memory with a fixed capacity of N E +1, and eliminate the pose state at time k - N E -1 At this time together with the states at [k - N E , k - 1] form the historical vehicle pose state sequence

[0088]

[0089] Similarly, the finally desired steering control amount obtained at time k - 1 is denoted as u f (k - 1). Send u f (k - 1) into a memory with a fixed capacity of N E , and eliminate the finally desired steering control amount u E at time k - N f (k - N E -1). At this time, u f (k - 1) together with the control amounts at [k - N E , k - 2] form the historical finally desired steering control amount sequence

[0090] For the execution performance limitations of the steer-by-wire system, the following receding horizon estimation optimization problem is constructed:

[0091]

[0092] Among them, the cost function J NMHEThe first term penalizes the deviation between the estimated state and the measured state; the second term penalizes the deviation between the estimated steering angle and the desired steering angle; the third term is the arrival cost, which reflects the internal connection between the beginning of the estimated time domain at the current moment and the estimated value at the previous moment, and is closely related to the stability of the estimator. S1, S2, and W are weight matrices of appropriate dimensions, where S1 reflects the degree of trust in the state measurement value. The larger its value, the more trust is placed in the measurement result; S2 represents the degree of trust in the performance of the steer-by-wire actuator. The larger its value, the lower the degree of limitation of its steering performance is considered; W reflects the degree of trust in the estimated value at the previous moment. The larger its value, the more trust is placed in the previous estimated information.

[0093] Equation (5) contains three main constraint equations: the first constraint represents the vehicle nominal kinematic model; the second constraint is the vehicle actuator saturation constraint, indicating that the true steering angle estimated should be within the limit range [σ min , σ max that the steer-by-wire actuator can reach; the third constraint indicates that the estimated value of the performance limitation of the steer-by-wire actuator should be within a reasonable range [u c,min , u c,max , where u c,min and u c,max represent the minimum and maximum values of the control compensation amount, reflecting the aggressiveness of the compensation intervention. The second and third constraints are the keys for the present invention to avoid oscillation and overshoot of the steering performance limitation estimation and ensure smooth update of the estimated value, and the proposed receding horizon estimation mechanism is the key to ensuring the optimality of the estimated value.

[0094] In S3, solving the receding horizon estimation problem, that is, Equation (5), can obtain the optimal solution sequence and use it as the optimal steering estimation value sequence within a fixed time domain length That is:

[0095]

[0096] Combined with the historical final desired steering control amount sequence in the memory The solution formula for the control compensation amount u c can be obtained:

[0097]

[0098] The averaging operation in the above formula can evaluate the average impact of the lumped uncertainty caused by the performance limitation of the steer-by-wire execution and model mismatch within the fixed time domain on the control system. This solution formula can also eliminate the estimated value spikes at certain moments, reduce the sensitivity of the control system to noise, and ensure smooth update of the control compensation amount.

[0099] In S4, traverse the reference path point sequence (The reference path point sequence is generated by the path planner according to the measured vehicle pose state ), and the reference path point closest to the vehicle centroid coordinates (X c , Y c ) is obtained and denoted as the reference matching point P0, as shown in Figure 1 . In practical engineering applications, when the reference path points are dense enough, the reference matching point can be approximated as the orthogonal projection point of the vehicle centroid onto the reference path. Then, drive along the reference path at the current speed v for a time step T to obtain the next projection point P1. By repeating this process, a set of reference matching points can be generated within the prediction control horizon N P .

[0100] According to the vehicle centroid position and the reference matching point, the lateral error e y and the heading angle error e ψ can be calculated as follows:

[0101]

[0102] where (X r , Y r ) are the coordinates of the reference matching point, and ψ r is the heading angle of the tangent line of the reference matching point. Assuming the control error e = [e y , e ψ T , the formula (8) is simplified to e(k) = g(ζ(k), ζ r (k)).

[0103] The following model predictive controller based on path tracking is established:

[0104]

[0105] where, N P represents the prediction control horizon, represents the terminal constraint set. The cost function J NMPC ​The first term is the trajectory tracking error cost, the second term is the steering angle increment change cost, and the third term is the terminal cost. Q, R, and P are weight matrices of appropriate dimensions. Q reflects the degree of emphasis on trajectory tracking accuracy, and R reflects the degree of emphasis on control smoothness. In practical engineering applications, the Q and R weight matrices can be adjusted to balance control accuracy and smoothness. This model predictive controller includes the following eight constraint equations: The first constraint is the state initialization constraint, indicating that the state at the current moment will be updated before each optimization solution; the second constraint is the vehicle kinematic constraint; the third constraint is the path tracking error calculation constraint; the fourth and fifth constraints are the calculation constraints for the steering angle control increment; the sixth constraint is the control compensation limit constraint, ensuring that the control amount applied to the vehicle steering angle does not violate the input saturation constraint, where δ min and δ max represent the minimum and maximum steering angles allowed by the actuator, and u c,min and u c,max represent the minimum and maximum values of the control compensation amount; the seventh constraint is the steering angle increment limit constraint, where Δδ min and Δδ max represent the maximum allowable change in the steering angle within a time step T; the eighth constraint is the terminal constraint, ensuring the stability of the model predictive controller.

[0106] In S5, solve formula (9), and take the first element in the obtained optimal control sequence as the optimal desired control amount u d (k) at time k:

[0107] u d (k) = u*(k|k) (10)

[0108] Calculate the control compensation amount u c (k) at time k according to formula (7), and then calculate the final desired steering control amount u f (k) according to the following formula:

[0109] u f (k) = u d (k) - u c (k) (11)

[0110] In S6, transfer the final desired steering control amount u f (k) to the actuator in the vehicle system to complete a control cycle (the time interval between two adjacent sampling points is called a control cycle).

[0111] The present invention also provides a heavy vehicle steer-by-wire active fault-tolerant control system, as Figure 2 shown, including a vehicle system, a sensing module, a planning module, and a control module connected in sequence.

[0112] Among them, the vehicle system includes an actuator, which is used to change the vehicle state by making a response according to the final desired steering control quantity u output by the control module. f , so that the actuator makes a response, thereby changing the vehicle state.

[0113] The perception module includes an integrated inertial navigation system, which is used to obtain the vehicle pose state in real time. The vehicle pose state includes the vehicle longitudinal speed v, the position coordinates (X s , Y s ) of the integrated inertial navigation system, and the heading angle ψ of the vehicle.

[0114] The planning module includes a path planner, which is used to generate a sequence of reference path points [ζ according to the vehicle pose state output by the perception module, and transmit [ζ r to the control module. r

[0115] The control module includes a delay unit, an estimator, a controller, and a compensation quantity calculation unit.

[0116] Among them, the controller includes a model predictive controller, which is used to receive the sequence of reference path points [ζ r output by the planning module and the vehicle pose state output by the perception module and generate a desired control quantity u d ; the delay unit is used to temporarily store the previous final desired steering control quantity u f , and transmit it to the memory in the estimator for use;

[0117] The estimator includes a memory and a moving horizon estimator. The memory is used to receive the real-time vehicle pose state output by the perception module and the historical final desired steering control quantity u output by the control module f , and output a sequence of historical final desired steering control quantities [u f with a fixed length and a sequence of historical vehicle pose states The moving horizon estimator outputs a sequence of real vehicle steering estimation values within a period of time according to the output information of the memory.

[0118] The compensation quantity calculation unit is used to calculate the control compensation quantity u f according to the sequence of historical final desired steering control quantities [u and the sequence of real vehicle steering estimation values; then combine u c and the desired control quantity u c to output the final desired steering control quantity u d at the current moment. f .

[0119] The present invention is an active fault-tolerant control strategy, which can estimate the performance limits of the steering system within a fixed time domain, reduce control conservatism, and without assuming the performance limits, achieving the unity of low conservatism and strong robustness. At the same time, based on the rolling horizon estimation strategy, the present invention can comprehensively consider the influence of the steering performance limits on the system over a period of time, without relying on linear models and Gaussian noise assumptions, to achieve the optimal and accurate estimation of the true steering angle within the fixed time domain; and can explicitly handle system constraints, strictly guarantee the transient performance boundary of the estimator convergence, and enhance the safety of the control system.

[0120] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention.

Claims

1. An active fault-tolerant control method for heavy-duty vehicle steer-by-wire, characterized in that: include: S1: Establish a discretized vehicle kinematic model; S2: Design an estimator based on the rolling horizon according to the execution performance limitations of the model; S3: obtaining a control compensation amount according to the optimal steering estimation value of the estimator and the historical final expected steering control amount; S4: Design a model predictive controller based on path tracking; S5: Obtaining a final desired steering control amount according to the optimal desired control amount of the model prediction controller and the control compensation amount; S6: Actively compensating for fault-tolerant control of the vehicle according to the final desired steering control amount; In S2, the vehicle posture state measured by the combined inertial navigation at time k is recorded as And use N E represents the estimated time domain; Enter a fixed capacity of N E +1 storage, remove kN E -1 moment posture status at this time With [kN E ,k-1] together constitute the historical vehicle posture state sequence The final desired steering control amount obtained at time k-1 is recorded as u f (k-1); u f (k-1) is fed into a fixed capacity N E Storage of kN is eliminated E The final desired steering control amount u at time -1 f (kN E -1); at this time u f (k-1) and [kN E ,k-2] together form the historical final expected steering control sequence Considering the performance limitations of the steer-by-wire system, the following receding horizon estimation problem is constructed: Among them, the cost function J NMHE The first term penalizes the deviation between the estimated state and the measured state; the second term penalizes the deviation between the estimated steering angle and the expected steering angle; the third term is the arrival cost; S1, S2 and W are weight matrices of appropriate dimensions; the first constraint represents the vehicle nominal kinematic model; the second constraint is the vehicle actuator saturation constraint, which means that the estimated true steering angle should be within the limit range that the wire steer can reach [σ min ,σ max ]; the third constraint indicates that the estimated value of the performance limit of the wire-controlled steer should be within a reasonable range [u c,min ,u c,max ], where u c,min and u c,max Indicates the minimum and maximum values ​​of the control compensation amount.

2. The method according to claim 1, characterized in that In S3, solve the rolling horizon estimation problem and obtain the optimal solution sequence and use it as the optimal steering estimate sequence within a fixed time domain length Right now: Combined with the historical final desired steering control sequence in the storage The control compensation u can be obtained c The solution formula is:

3. The method according to claim 2, characterized in that In S4, the traversal is performed by the path planner based on the measured vehicle posture state Generated reference path point sequence Get the distance from the vehicle's center of mass coordinates (X c ,Y c ) is the nearest reference path point, recorded as the reference matching point P0; then, it drives along the reference path at the current speed v for a time step T to obtain the next projection point P1; by repeating this process, in the predictive control domain N P Generate a set of reference matching points According to the vehicle center of mass position (X c ,Y c ) and the reference matching point, the lateral error e can be calculated y and heading angle error e ψ : Where (X r ,Y r ) is the coordinate of the reference matching point, ψ r is the heading angle of the tangent line of the reference matching point. Assume that the control error e=[e y ,e ψ ] T , simplifying the above formula to e(k) = g(ζ(k),ζ r (k)); Build the following path-following based model predictive controller: Among them, N P represents the predictive control time domain, represents the terminal constraint set; the cost function J NMPC The first term is the trajectory tracking error cost, the second term is the steering angle increment change cost, and the third term is the terminal cost; Q, R, and P are weight matrices of appropriate dimensions; the model predictive controller contains the following eight constraint equations: the first constraint is the state initialization constraint, which means that the current state will be updated before each optimization solution; the second constraint is the vehicle kinematic constraint; the third constraint is the path tracking error calculation constraint; the fourth and fifth constraints are the calculation constraints of the steering angle control increment; the sixth constraint is the control compensation limit constraint, which ensures that the steering angle control amount acting on the vehicle does not violate the input saturation constraint, where δ min , δ max Indicates the minimum and maximum steering angles allowed by the actuator, u c,min and u c,max represents the minimum and maximum values ​​of the control compensation; the seventh constraint is the steering angle increment limit constraint, where Δδ min and Δδ max It represents the maximum change of steering angle allowed within a time step T; the eighth constraint is the terminal constraint.

4. The method according to claim 3, characterized in that In S5, the model predictive controller problem is solved, and the first element in the obtained optimal control sequence is taken as the optimal expected control quantity u at time k d (k): you d (k)=u*(k|k) Calculate the control compensation u at time k c (k), and then calculate the final desired steering control amount u according to the following formula: f (k): in f (k)=u d (k)-u c (k).

5. The method according to claim 1, characterized in that In S1, the discretized vehicle kinematic model is established as follows: in, represents a discrete time point; (X, Y) represents the horizontal and vertical coordinates of the center position of the rear axle of the vehicle in the geodetic coordinate system; ψ represents the heading angle of the vehicle; v represents the longitudinal speed of the vehicle; T represents the discrete time step; L represents the wheelbase; δ f represents the front wheel turning angle; the above formula is simplified into the following form: Among them, ζ represents the vehicle state, that is, ζ=[X,Y,ψ] T ; u represents the control amount, that is, the vehicle front wheel steering angle δ f ; Represents the set of positive integers from 0 to infinity.

6. The method according to claim 5, characterized in that The distance between the installation position of the combined inertial navigation system and the center of the rear axle of the vehicle is recorded as L', and the relationship between the position of the combined inertial navigation system and the center of the rear axle of the vehicle is established: Among them, (X s ,Y s ) represents the position coordinates of the combined inertial navigation; The distance between the vehicle's center of mass and the combined inertial navigation installation position is recorded as L", and the relationship between the vehicle's center of mass and the combined inertial navigation installation position is established as follows: Among them, (X c ,Y c ) is the coordinate of the vehicle center of mass.

7. An active fault-tolerant control system for heavy-duty vehicle steer-by-wire, characterized in that: Running the method according to any one of claims 1 to 6, comprising a vehicle system, a perception module, a planning module and a control module connected in sequence; The vehicle system includes an actuator for controlling the final desired steering control amount u according to the output of the control module. f , so that the actuator responds, thereby changing the vehicle state; The perception module includes a combined inertial navigation system to obtain the vehicle's position and posture in real time. The vehicle posture state includes the vehicle longitudinal velocity v, the position coordinates of the combined inertial navigation (X s ,Y s ), the vehicle’s heading angle ψ; The planning module includes a path planner, which is used to calculate the vehicle posture state according to the output of the perception module. Generate a reference path point sequence [ζ r ], and [ζ r ] is passed to the control module; The control module includes a delay device, an estimator, a controller and a compensation amount solving unit; The controller includes a model prediction controller, which is used to receive the reference path point sequence [ζ r ] and the vehicle posture state output by the perception module And generate the desired control quantity u d ; The delay device is used to temporarily store the last desired steering control value u f , and passed to the storage in the estimator for use; The estimator includes a storage and a rolling time domain estimator. The storage is used to receive the real-time vehicle posture state output by the perception module. The historical final desired steering control amount u output by the control module f , and output a fixed-length historical final desired steering control sequence [u f ] and historical vehicle pose state sequence The rolling horizon estimator outputs a sequence of vehicle true steering estimation values ​​within a time domain based on the output information of the storage. The compensation amount calculation unit is used to calculate the final expected steering control amount sequence [u f ] and the vehicle's true steering estimate sequence Calculate the control compensation u c ; Then combine u c and the desired control quantity u d Output the final desired steering control amount u at the current moment f .

Images