A method and system for redundancy control of multi-actuator braking in distributed electric drive vehicles
By using multi-actuator collaborative control, the longitudinal braking force and lateral stability problems of distributed electric drive vehicles under brake actuator failure are solved. The system realizes the reconstruction of braking force and recovery of lateral stability under fault conditions, and is suitable for joint simulation verification of single-wheel and dual-wheel failures.
Patent Information
- Application Number
- CN202611555099.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-09
- Publication Date
- 2026-11-17
AI Technical Summary
In existing distributed electric drive vehicles, brake actuator failure makes it difficult to simultaneously meet longitudinal braking force and lateral stability requirements, especially during emergency braking or on split-road surfaces. The dynamic load transfer and the coupling of longitudinal and lateral forces of the tires make it difficult to balance braking performance and yaw stability.
By using multi-actuator coordinated control, including electromechanical brakes, hub motors, and rear wheel steering actuators, the longitudinal braking force of the vehicle is restored and lateral stability is maintained by using cross-domain stability margin, deceleration deviation, and constraint correction feedback to jointly adjust the recovery coefficient.
It realizes the restoration of longitudinal braking force and maintenance of lateral stability under the condition of brake actuator failure through braking force reconstruction and closed-loop recovery. It is applicable to joint simulation verification of single-wheel and dual-wheel failures, and improves the braking performance and yaw response of the vehicle.
Smart Images

Figure CN123104676A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of braking and yaw stability control technology for distributed electric drive vehicles, and specifically to a multi-actuator braking redundancy control method and system for distributed electric drive vehicles. Background Technology
[0002] The development of distributed drive, steer-by-wire, and brake-by-wire has transformed vehicle stability control from single-actuator intervention to multi-actuator coordination involving steering, drive, and braking. Existing technologies, from the perspective of yaw stability in distributed drive vehicles, review the technological evolution of active steering, direct yaw moment control, and integrated control. Hybrid MPC is employed to coordinate active front-wheel steering and differential braking; vehicle yaw dynamics and wheel slip control are unified into a predictive model, and integrated stability-traction control is verified through real-vehicle testing; centralized MPC simultaneously handles lateral stability, traction / braking, and rollover constraints; and further, priority is set for front-to-rear axle torque distribution and differential braking based on vehicle conditions. For distributed drive vehicles, electronic stability control is achieved through optimized distribution of drive / regenerative braking torque, and torque vector control and electronic stability control are coordinated using function allocation. These studies demonstrate that multi-actuator coordination can alleviate the impact of single differential braking on longitudinal speed and tire adhesion margin, but its effectiveness depends on actuator health and available tire force.
[0003] Electromechanical brakes (EMBs) establish frictional braking force through a motor-driven mechanical transmission mechanism, enabling independent control of the wheel ends. In-wheel motors can also provide redundant braking force via electric braking; these are different actuators. An EMB malfunction reduces the frictional braking capability of the corresponding wheel end and may cause asymmetry in braking force between the left and right sides. During emergency braking or on split-plane surfaces, the coupling of dynamic load transfer and longitudinal and lateral tire forces makes it difficult to simultaneously achieve optimal braking performance and yaw stability.
[0004] For brake-by-wire failure, existing technologies employ sequential quadratic programming to reconstruct the remaining three-wheel braking force and use steer-by-wire to suppress yaw caused by single-wheel failure. They utilize Lagrange multipliers and KKT conditions to allocate additional braking force at available wheel ends, balancing deceleration tracking and suppression of undesirable yaw moments. A two-layer strategy of "brake force redistribution—sliding mode yaw control" is constructed. For cornering conditions, single-wheel and dual-wheel failures are handled separately using brake force reconstruction and path tracking. Single-wheel braking failure is combined with heavy vehicle rollover prevention control and hardware-in-the-loop verification is conducted. For EMB actuator failure in commercial vehicles, coordinated vehicle stability control is studied. Existing work has demonstrated the effectiveness of fault isolation, available wheel end reconstruction, and steering-braking coordination, but the verification objects are mostly concentrated on single-wheel failure, specific corners, or single stability risks. For complete failure extending from a single wheel to two wheels, especially when split-wheel adhesion, measurement noise, and control delay coexist, a unified control chain and systematic comparative verification are still lacking between longitudinal braking capacity, yaw compensation requirements, and actuator constraints.
[0005] Model predictive control can consider state and actuator constraints within a finite prediction time domain, but its performance is affected by the consistency between the predictive model, measurement signals, and actual execution responses. Pipeline model predictive control handles bounded uncertainties through nominal optimization, error feedback, and constraint tightening. For brake actuator failures, rear-wheel steering can provide yaw compensation, but the stability margin of steering control still needs to be linked to braking capacity constraints, and the constraint correction results should be fed back to subsequent recovery commands to avoid continuously proposing unrealizable braking force increments. Summary of the Invention
[0006] This invention addresses brake actuator failure scenarios by employing multi-actuator collaboration to reconstruct and restore braking force in a closed-loop manner, thereby restoring longitudinal braking force and maintaining lateral stability. The multi-actuator includes an electromechanical brake (EMB), wheel hub motors, and a rear-wheel steering actuator, where the wheel hub motors provide electro-braking torque. The recovery coefficient is adjusted through cross-domain stability margin, deceleration deviation, and constraint correction feedback, ensuring that the additional braking force adapts to the remaining actuator capacity and the vehicle's lateral movement.
[0007] To achieve the above objectives, the present invention provides a multi-actuator braking redundancy control method for distributed electric drive vehicles, comprising the following steps: S1. Vehicle status information acquisition and fault status coding: acquire vehicle motion status, wheel-end brake actuator fault status and road adhesion parameters, and determine the availability status of each wheel's electromechanical brake and wheel hub motor respectively. S2. Dynamic wheel load acquisition and wheel-by-wheel available braking capacity calculation: Acquire the dynamic vertical load and tire lateral force of each wheel, calculate the remaining longitudinal adhesion capacity of the tire based on the adhesion parameters, and determine the upper limit of the available torque of the corresponding actuator based on the available status of the electromechanical brake and hub motor of each wheel, so as to obtain the upper limit of the available braking force of each wheel. S3. Braking force benchmark allocation based on dynamic wheel load: The target deceleration is converted into total braking force demand, the initial braking force is allocated according to the dynamic vertical load of each wheel, the initial braking force is limited according to the upper limit of the available braking force of each wheel, and the redistribution and braking asymmetry adjustment are carried out according to the remaining capacity of each wheel to form a four-wheel benchmark braking force. S4. Braking yaw moment calculation and prediction model construction: The additional braking yaw moment is calculated based on the wheel-end braking force and wheel track. It and the front wheel steering angle are used as the exogenous inputs of the rear wheel steering prediction model, and the control targets of the center of gravity sideslip angle and yaw rate are determined. S5. Rear wheel steering based on Tube-MPC algorithm: The rear wheel steering prediction model is decomposed into a nominal system and an error system. Taking the control targets of the center of gravity sideslip angle and yaw rate as references, the optimization target of the prediction state tracking error is constructed, and the nominal rear wheel steering angle sequence is solved. The first nominal control quantity is superimposed with the error feedback correction quantity to obtain the candidate rear wheel steering angle command. Then, the steering angle and steering rate constraints are applied to form the constrained rear wheel steering angle command. S6. Cross-domain robust stability margin calculation: The stability margin is formed based on the utilization rate of the error relative to the robust positive invariant set, the rear wheel steering angle occupancy, the center of gravity sideslip angle, the yaw rate and the tire adhesion utilization rate, and the activation conditions for brake force recovery are determined in combination with the effectiveness of rear wheel steering control. S7. Generation of Braking Force Increment Command: When the conditions for restarting braking force are met, the recovery coefficient is determined based on the stability margin, deceleration deviation, feedback of the constraint correction of the previous cycle and the start-up process. Based on the four-wheel reference braking force, the braking force is allocated and superimposed according to the remaining braking capacity of each wheel actuator. The increase of the braking force after the constraint correction of the previous cycle is used to limit the growth of the increment in this cycle, thus forming candidate four-wheel braking force commands. S8. Safety constraint verification and correction feedback: The candidate four-wheel braking force command is subjected to fault actuator shielding, attachment and actuator capacity verification and analysis correction to obtain the corrected four-wheel braking force command; with the sum of the four-wheel reference braking forces in the same control cycle as a reference, the braking force increment command and the brake force increment after constraint correction are calculated respectively to form feedback coefficients and save them to the next control cycle. S9. Control command output and retraction: Convert and distribute the corrected four-wheel braking force command into braking torque commands for the electromechanical brake and wheel hub motor, and output the constrained rear wheel steering angle command; reduce or withdraw the additional braking force when the recovery condition is not met; switch to the regular rear wheel steering command when the rear wheel steering control is invalid; output the constrained rear wheel steering angle command to the rear wheel steering actuator; the faulty actuator always maintains the zero torque command.
[0008] A distributed electric drive vehicle multi-actuator braking redundancy control system includes a controller and an information acquisition and fault coding module, a wheel-by-wheel available braking capacity calculation module, a reference braking force distribution module, a rear wheel steering duct model predictive control module, a robust stability margin assessment module, a braking force recovery module, a safety constraint verification module, a constraint correction feedback module, and a command output and rollback module.
[0009] Therefore, the present invention employs the above-mentioned distributed electric drive vehicle multi-actuator braking redundancy control method and system, which has the following beneficial effects: (1) The actuator failure and the available braking capacity at the wheel end are handled separately. Zero torque shielding is implemented for the failed electromechanical brake, while the still available hub motor braking capacity is taken into account, and the allocation boundary is determined based on tire adhesion and actuator capacity.
[0010] (2) The rear wheel steering stability margin is used to adjust the braking force recovery. The utilization rate of the robust positively invariant set (RPI), the occupancy of the active rear-wheel steering (RWS) angle, and the lateral running state of the vehicle are combined to form a margin that can be input into the braking recovery process.
[0011] (3) Form a constraint correction feedback closed loop. Based on the incremental relationship between the candidate braking force and the constraint-corrected braking force relative to the benchmark value of the same period, adjust the recovery coefficient and allowable growth amount for the next period to limit the repeated proposal of braking increments that exceed the available capacity.
[0012] (4) The recovery command is subject to the fault actuator shield and capacity boundary. Through analysis and verification and command correction, the wheel end torque and rear wheel angle are limited within the set constraint range; the actual tire adhesion utilization rate and error pipeline coverage are still evaluated in conjunction with the vehicle response.
[0013] (5) The same control structure can be used for single-wheel and same-side dual-wheel electromechanical brake failures. The joint simulation of uniform high-adhesion split road surface is used to evaluate the overall control effect. The specific degree of improvement and constraint satisfaction depend on the fault combination, road surface and measurement disturbance. The applicable boundaries are shown in the example.
[0014] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0015] Figure 1 This is a flowchart illustrating an embodiment of a distributed electric drive vehicle multi-actuator braking redundancy control method and system according to the present invention. Figure 2 This is a schematic diagram of the braking force distribution and constraint correction feedback closed loop of a distributed electric drive vehicle multi-actuator braking redundancy control method and system embodiment of the present invention; Figure 3 This is a schematic diagram of the rear wheel steering pipeline model predictive control and robust stability margin formation in an embodiment of a distributed electric drive vehicle multi-actuator braking redundancy control method and system of the present invention. Figure 4 This is a schematic diagram of the system architecture for multi-actuator collaborative braking force reconstruction and closed-loop recovery in a distributed electric drive vehicle multi-actuator braking redundancy control method and system embodiment of the present invention. Figure 5 In an embodiment of the distributed electric drive vehicle multi-actuator braking redundancy control method and system of the present invention, the left front wheel electromechanical brake fails, S1, S2, without added measurement noise and random sensor delay, the left column (a), (c), (e) are uniform high attachment, and the right column (b), (d), (f) are split road surfaces; Figure 6 In the embodiment of the distributed electric drive vehicle multi-actuator braking redundancy control method and system of the present invention, the left front wheel electromechanical brake fails, S3, S4, Gaussian measurement noise and random sensor delay are superimposed, the left column (a), (c), (e) are uniform high-adjustment, and the right column (b), (d), (f) are split road surfaces; Figure 7 In an embodiment of the distributed electric drive vehicle multi-actuator braking redundancy control method and system of the present invention, the left front and left rear electromechanical brakes fail simultaneously, D1 and D2, without added measurement noise and random sensor delay, the left column (a), (c), and (e) are uniform high attachment, and the right column (b), (d), and (f) are split road surfaces; Figure 8 In an embodiment of the distributed electric drive vehicle multi-actuator braking redundancy control method and system of the present invention, the left front and left rear electromechanical brakes fail simultaneously, D3 and D4, superimposed with Gaussian measurement noise and random sensor delay, the left column (a), (c) and (e) are uniform high-adjustment, and the right column (b), (d) and (f) are split road surfaces; in, Figures 5-8 In the figure, (a) and (b) represent deceleration; (c) and (d) represent the vehicle's trajectory; and (e) and (f) represent yaw rate. Detailed Implementation
[0016] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments. Figure 1 As shown, this embodiment is implemented in the following order: information acquisition and fault coding, dynamic wheel load acquisition and available braking capacity calculation, baseline braking force distribution, braking yaw moment calculation and prediction information construction, rear wheel steering control, robust stability margin assessment, braking force recovery, safety constraint verification and feedback, and command output and rollback; constraint correction feedback is used for recovery adjustment in the next control cycle.
[0017] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0018] Example Please see Figures 1-8 This invention provides a method and system for multi-actuator braking redundancy control of distributed electric drive vehicles. The braking redundancy control specifically involves coordinating the remaining available electromechanical brakes, hub motors and rear wheel steering actuators under the condition of brake actuator failure, restoring and compensating longitudinal braking capacity through braking force reconstruction and closed-loop recovery, and using rear wheel steering to provide yaw compensation, while taking into account the lateral stability of the vehicle.
[0019] In this embodiment, within each control cycle following brake actuator failure, the capacity is first calculated based on wheel load, tire force, adhesion parameters, and actuator status to form a baseline braking force. Then, the rear wheel steering pipeline model predictive controller generates constrained steering angle and stability margin information. Subsequently, based on stability margin, deceleration deviation, and constraint correction feedback from the previous cycle, an incremental braking force command is generated. After final safety constraint verification, the wheel-end torque is output, and the corrected increment is fed back to the next cycle. Specifically, the baseline braking force distribution utilizes remaining available braking capacity to reconstruct the braking force, rear wheel steering provides yaw compensation, and the additional braking force achieves closed-loop recovery through constraint correction feedback.
[0020] In this embodiment, the constraint correction feedback comes from the controller's output command, rather than a direct measurement of the actual tire force. Both the candidate total braking force and the corrected total braking force are compared with the baseline total braking force for the same cycle; a negative correction increment indicates that the safety constraint causes the total command to be lower than the baseline value, and is not included in the feedback ratio as recovered braking force.
[0021] The verification conditions used were left front wheel fault (S1-S4) and left side dual wheel fault (D1-D4), covering uniform high-adhesion, split road surface, no disturbance, and Gaussian measurement noise superimposed with 30-70 ms random sensor delay, to evaluate the braking performance and yaw response of the control method under different fault and disturbance conditions.
[0022] Wheel-end index set is Let these represent the front left, front right, rear left, and rear right, respectively. This can be represented using the wheel end set as... The available braking force of each wheel is greater than zero; when the electromechanical brake fails but the hub motor remains available, the wheel can still have a non-zero braking capacity. In this embodiment, only when the electromechanical brake completely fails, the hub motor remains available.
[0023] Table 1. Meaning of each symbol parameter
[0024] This manual uses non-negative braking force amplitude in braking force distribution and capacity calculation; the actual longitudinal force of the tires uses a signed quantity in the vehicle coordinate system, and the two should not be mixed. The superscripts 0, cand, and out represent the same-cycle baseline, candidate before constraint correction, and output after constraint correction, respectively; cmd is used to represent incremental commands. The subscript Σ indicates summation of all four wheels, and k represents the control cycle. The braking torque of the hub motor is converted to a negative sign at the output interface.
[0025] The following explains the specific implementation of each step.
[0026] S1, Vehicle Information Acquisition and Fault Status Encoding. The vehicle's motion state and externally provided actuator fault information are read and converted into wheel-by-wheel effective coefficients. In the co-simulation, faults are applied through fault coding and physical torque masking; the fault detection algorithm itself is not part of the design scope of this embodiment.
[0027] The effective coefficients of electromechanical brakes and hub motors are defined as follows: and For complete failure, the value is zero; for normal operation, the value is one. The output torque of a faulty electromechanical brake is set to zero; the corresponding hub motor can still provide braking force within its capacity. ; in, and These are the braking torque amplitudes of the electromechanical brake and the hub motor, respectively. and These are the corresponding maximum allowable amplitudes; the fault wheel in this embodiment =0, each round =1. A zero-torque command from a certain actuator does not equate to the longitudinal contact force of the tire being constantly zero.
[0028] S2, Dynamic wheel load acquisition and wheel-by-wheel available braking capacity calculation. Dynamic vertical load of each wheel. Tire lateral force The slip ratio is output from the CarSim vehicle model, sampled, and processed before being input to the controller; the adhesion parameters are given according to the test conditions. The controller uses these inputs to perform capacity calculations, without setting up an independent wheel load state observer or an online adhesion estimator.
[0029] "Available braking capacity at the wheel end" in this manual refers to the upper limit of the distributable longitudinal braking force, in N. Calculate the actuator capacity separately. With tire remaining longitudinal capacity Then take the smaller value. One implementation that includes lateral force protection and slip derating is as follows: ; in, The effective rolling radius; μ is the adhesion parameter used by the controller; For adhesion safety factor; To calculate the lower limit of protection; The lateral force protection factor is less than one. This is the sliding depreciation factor. For dynamic vertical loads, This is the lateral force of the tire. It can be generated by the wheel end axle. A value greater than 0 indicates that all actuators on that wheel are functioning correctly, without requiring all actuators on that wheel to be in good working order. Protection parameters can be calibrated according to the vehicle and tire specifications.
[0030] S3 is a braking force benchmark allocation based on dynamic wheel load. Initial values are first formed according to the wheel-by-wheel load ratio, then capacity trimming and remaining capacity compensation are implemented. The rule layer also adjusts the degree of braking force asymmetry based on the rear wheel steering setup and vehicle risk, forming a benchmark for subsequent recovery.
[0031] Total demand is expressed as the magnitude of the non-negative braking force: ; in, m For the overall vehicle quality, This is the deceleration amplitude used for allocation; it can be bounded and corrected by combining the driver's target deceleration amplitude with the measured deceleration deviation. This embodiment uses deceleration feedback compensation and does not additionally include the same driving resistance in the total demand.
[0032] The initial allocation of dynamic wheel load weights and capacity after capacity trimming for each wheel is as follows: ; ; in, This represents the dynamic wheel load weight for the i-th round; if you need to describe the axle load weight... , Two coaxial wheels Add them together. Assigning values to the load proportions. This is the capacity reduction value. For its remaining capacity, Let represent the total remaining capacity, and D represent the total braking force demand that has not yet been met. i , j All are wheel-end indexes.
[0033] when When the capacity is greater than 0, additional allocation will be made according to the remaining capacity ratio, and the total amount of additional allocation will be limited to [a certain limit]. within; when When the value is 0, the capacity reduction value is maintained. This process is applicable to situations where a single wheel has insufficient capacity or multiple actuators fail, and it does not depend on any one axis; two usable wheel ends must be retained.
[0034] ; ; in, To emphasize the distribution of total braking force, The left-right pairing results formed by the rule layer are conservative. This is a mixing coefficient adjusted based on steering build-up, yaw, sideslip, and path deviation. The conservative pairing results prioritize symmetrical braking force within the allowable capacity range of the two wheels on the same axle. After mixing and subsequent yaw smoothing, a capacity check is performed again to obtain the baseline braking force for all four wheels. .
[0035] The rule layer corrects the difference in braking force between the left and right sides based on available capacity and smooths changes in braking yaw moment. Additional braking yaw moment and rule-based rear wheel steering work together to suppress asymmetric response after a fault.
[0036] like Figure 2 As shown, the baseline allocation and subsequent braking force recovery together form a braking control closed loop; the constraint correction feedback is formed based on the baseline total braking force in the same cycle, according to the increment of total braking force before and after candidate command verification, and is used for recovery adjustment in the next control cycle. The specific relationship is as follows.
[0037] The baseline allocation first considers both longitudinal demand and lateral stability after a failure; based on this, the recovery phase increases the unidirectional braking force according to the stability margin and the remaining capacity of the actuators. The additional amount should not be allocated solely based on the proportion of the upper limit of each wheel's capacity, but should be determined based on the remaining capacity relative to the baseline value.
[0038] ; in, As the baseline total braking force, This represents the remaining braking capacity of the actuator relative to this benchmark. The recovery candidate quantity is first constrained by the actuator capacity, and then corrected by the final verification in conjunction with tire adhesion and slip conditions.
[0039] When the stability margin is insufficient or the rear wheel steering has not yet been effectively established, the recovery coefficient is reduced; when the stability margin and remaining capacity allow, the total braking force is gradually increased. The difference between the left and right braking forces is adjusted together according to the yaw stability requirements and the available capacity.
[0040] S4, Construction of Braking Yaw Moment Calculation and Prediction Information. Taking forward yaw along the vehicle's x-axis and leftward / left-turn yaw along the y-axis as positive, the yaw moment is calculated after converting the tire forces to the vehicle's coordinate system. Under the small turning angle approximation, the non-negative braking force amplitude... correspond ≈- Therefore, the braking yaw moment is approximately: ; This refers to the lateral coordinates at the wheel end. This article uses the vehicle yaw moment symbol consistently; the braking force difference symbol in the execution interface is converted according to this convention.
[0041] Predicted state Where β is the sideslip angle of the center of mass and r is the yaw rate; control input u = This is the common steering angle for the rear wheels. Front wheel steering angle. and the braking yaw torque of the previous cycle This constitutes an exogenous input. This embodiment is for linear fault braking. and These represent the reference values for the sideslip angle and yaw rate, respectively, both set to zero.
[0042] S5, Tube model predictive control (Tube-MPC) solution for rear wheel steering. The rear wheel steering prediction model uses the center of gravity sideslip angle and yaw rate as states, the rear wheel steering angle as the control input, and the front wheel steering angle and the additional yaw moment from the previous braking cycle as exogenous inputs to predict the vehicle's lateral and yaw motions. The nominal response and its deviation are described by a nominal system and an error system.
[0043] like Figure 3As shown, the nominal system predicts the vehicle response, and the error system describes the deviation of the actual state from the nominal state; the product of the auxiliary feedback gain K and the error e, Ke, is the rear wheel steering angle error feedback correction. Candidate commands are formed by superimposing the nominal input with this correction, and then the execution layer processes the steering angle and speed constraints.
[0044] Linearization errors, nonlinear tire effects, scheduling errors, and measurement disturbances can cause the actual trajectory to deviate from the nominal trajectory. Pipeline design is based on the declared scheduling domain and bounded set of disturbances; Gaussian noise and random sensor delay tests do not automatically satisfy this bounded envelope assumption.
[0045] With a fixed scheduling amount within the control period k, the relationship between the discrete system, the nominal system, and the error feedback is expressed as follows: ; in, Vehicle status. In nominal state, The difference between the two; As an ideal synthesis input, K represents the nominal rear wheel steering angle, and K represents the auxiliary feedback gain. , These are the discrete state matrix and the control input matrix, respectively. Given exogenous terms, For the exogenous input matrix, It consists of the front wheel steering angle and the braking yaw moment of the previous cycle; For unmodeled errors and disturbances, satisfy the set used in the design. The above error recursion formula applies when the actual applied input equals the ideal synthesized input.
[0046] Define the Minkowski sum, Pontryagin difference, and set linear mapping for error set construction and constraint tightening: ; Where A and B are sets, a and b are elements of the corresponding sets, L is the linear mapping matrix, ⊕ is the Minkowski sum, and ⊖ is the Pontryagin difference. A, B, and the system matrix in set operations. , Distinguish by context.
[0047] Offline selection of common auxiliary feedback gain K, and ensuring that the error set Z including the origin satisfies the following for the selected scheduling vertex: ; Where j is the scheduling vertex number, The total number of vertices. , This is the discrete model matrix of the vertex. Let Z be the set of corresponding perturbations. Z is called the common robust positive invariant set of the error system. Its polyhedral inequality form is used for online calculation of error utilization.
[0048] Z is represented online using a normalized half-space, with the right-hand side of each half-space normalized to one: ; in, It is a polyhedral half-space matrix. For its l-th row, This represents the utilization rate of the error relative to the set Z. The offline design parameters given in this embodiment are: Z has 82 vertices, and its outer envelope is: ; in, This is the error in the centroid sideslip angle. This represents the yaw rate error. The maximum closed-loop spectral radius of the scheduling vertex for this set of design parameters is approximately 0.94044; the spectral radius is calculated and verified using the system matrix and feedback gain, and is not used as an independently set control input.
[0049] The disturbance was When the covering and set inclusion relationships hold and the actual input conforms to the error system assumptions. ∈Z can be deduced ∈Z, the actual state is located at Within the conduit formed by ⊕Z. This conclusion pertains to state error, not to requiring the actual state to lie directly within Z.
[0050] Let the actual state and the set of input constraints be X and U, respectively. The nominal constraints are tightened by set difference as follows: ; in, , These represent the tightened nominal state and input constraints, respectively. The actual state boundaries of this design example are |β|≤0.22 rad, |r|≤0.50 rad / s, and the nominal boundaries are approximately | |≤0.211409 rad、| |≤0.478626 rad / s. The maximum steering angle of the active rear wheel steering (RWS) is 20°, the maximum feedback within the set Z is about 2.07588°, and the nominal maximum steering angle is about 17.92412°; the 80° / s rate limit is implemented by the execution layer.
[0051] The nominal prediction uses an N-step sequence of rear wheel angles, and the objective function includes the predicted state, the nominal input, and the changes in adjacent inputs: ; in, For the nominal state at step j predicted by period k, The nominal rear wheel steering angle; The final step uses terminal weighting to determine the state weights. , These are the rotation amplitude and the weight of the change, respectively; Take the nominal input from the previous cycle. The reference state for linear braking is zero.
[0052] In this embodiment, the sampling period is 0.01 s, the prediction sequence length is N=10, and the basic state weight is Q=diag(100,18). Using the nominal turning sequence as the optimization variable, the stacked prediction relationship is: ; Among them, vector z is formed by stacking the nominal states of each step in the prediction time domain, and vector v is formed by stacking the nominal rear wheel rotation angles of each step; matrices F and G are obtained by recursion along the time domain from the discrete model, and vector d is the cumulative effect of known exogenous terms.
[0053] ; in, The iteration step size, It is an index that incorporates penalties for violating state and terminal constraints into the objective function. It is its gradient with respect to the corner sequence; clip indicates element-wise upper and lower bounds. , The sequence boundary is formed to tighten the input boundary.
[0054] The online solution first initializes with an unconstrained quadratic form, then performs a finite number of projected gradient iterations. After iteration, the predicted state, terminal conditions, and numerical finiteness are checked; the corresponding candidate control is activated only when the tolerance condition is met.
[0055] Angle change penalty is used to suppress rapid changes, and the final angle rate constraint is imposed separately by the execution layer. The output of a finite number of iterations is used as a candidate solution after constraint feasibility verification.
[0056] The warm-start sequence is achieved by left-shifting the previous sequence and repeating the last term; when the mode or fault changes, the controller logic is reset. Terminal weights, terminal sets, and feasibility tolerance are used together for candidate solution verification.
[0057] The execution layer performs rotation and rate corrections on the ideal synthesized input: ; in, For candidate common turns selected by the control source, This is the corrected output. This is the upper limit of the angle amplitude. This is the upper limit of the rate. The sampling period is specified. When the control source switches, instruction changes are also restricted according to the set execution layer rules.
[0058] The ideal candidate turning angle is synthesized from nominal control and error feedback as follows: ; in, K( is the first term of the current nominal sequence) - This is the feedback correction amount for the rear wheel steering angle error. If the execution layer modifies the candidate input, this correction amount should be taken into account the applicability conditions of the error system.
[0059] make The error recursion relationship is as follows: This represents the deviation between the actual input and the ideal input caused by amplitude limiting, rate limiting, or a combination of permissions. ; The aforementioned pipeline maintenance relationship is based on the premise that the total error is covered by the design disturbance set and the actual input meets the corresponding constraints. In this embodiment, Gaussian noise and a random sensor delay of 30-70 ms are applied through the measurement feedback channel.
[0060] S5 also includes brake yaw feedforward and fault-related rear wheel steering control adjustment, as detailed below: (a) Rear wheel steering feedforward.
[0061] To compensate for the yaw moment caused by asymmetrical braking, the rear wheel steering angle is fed forward using the yaw moment from the previous braking cycle: ; in, The vehicle's yaw direction is used as the standard. The distance from the center of gravity to the rear axle. The equivalent lateral stiffness of the rear axle; in the linear small steering angle model, the effect of the rear wheel steering angle on the yaw rate is approximately: Therefore, the aforementioned feedforward is used to counteract the asymmetric yaw effect of braking.
[0062] The nominal optimizer solves for the rotation correction near the feedforward. ,Right now = + This feedforward is the channel through which braking yaw information enters the rear wheel steering controller; longitudinal recovery is regulated by margin and constraint correction feedback.
[0063] (ii) Rear wheel steering control adjustment. For a single rear wheel electromechanical brake failure, the control control is adjusted using a combination of the center of gravity sideslip angle risk and yaw risk: ; in, , For normalized lateral and yaw risks, ξ represents portfolio risk. This represents a risk-adjusted state with hysteresis and rate-of-change limitations. This represents the upper limit of permissions for this fault mode, where 'a' is the pipeline control participation coefficient. and These are candidate turns for rule and pipeline control, respectively.
[0064] The fact that the rear wheel steering command is still executed even when the permission coefficient is zero does not mean that the rear wheel angle is zero. Front wheel and multi-wheel faults use full pipeline permissions. Permission adjustments do not change offline K, W, and Z; their input correction effects should be handled according to the aforementioned error relationships.
[0065] S6, Cross-domain robust stability margin calculation. The recovery phase uses the effectiveness of the rear wheel steering channel control, error utilization rate, steering angle, and execution layer protection status from the previous cycle, and combines them with the current center of gravity sideslip angle, yaw rate, and tire utilization rate to form recovery enable and margin information.
[0066] S61, define a margin function that decreases as the amount of risk increases: ; in, , These are the thresholds for initiating recovery of contraction and for complete contraction, respectively. Thresholds are set for error utilization, steering angle occupancy, yaw rate amplitude, center of gravity sideslip angle amplitude, and tire utilization, resulting in five margin coefficients.
[0067] S62, the cross-domain stability margin is the minimum value of all margin coefficients: ; in, , , , , These correspond to the robust positive invariant set utilization rate, rear wheel steering angle, yaw rate, center of gravity sideslip angle, and tire adhesion utilization rate, respectively; each coefficient is between zero and one. The steering rate protection participates in the recovery enable judgment through the rear wheel steering execution layer state.
[0068] S7, Braking force increment command generation. Mapping deceleration deviation to a coefficient. Mapping the constraint correction feedback state to coefficients The startup process is mapped to coefficients. When the recovery enabling condition is met, the target recovery coefficient is the minimum value of each coefficient multiplied by the maximum allowable recovery ratio. ; in, To achieve the maximum recovery rate, , The rate of increase and decrease of the recovery coefficient is defined as follows: the deceleration deviation is the non-negative part of the difference between the target and measured deceleration amplitudes. When recovery enable is not enabled, the target coefficient is set to zero and reduced according to the exit rules; a reset is performed when the control mode exits or a fault switch occurs.
[0069] The initial additional amount is calculated based on the remaining capacity of the actuator, and an upper limit for growth is applied when the instruction in the previous cycle reaches the feedback threshold: ; in, This is an additional amount to be allocated. To allow for the growth rate, Enable a feedback threshold. (Order) = When it is greater than zero, the amount added to the i-th round is... Otherwise, the additional amount is zero. The candidate braking force for each round is the baseline value plus the additional amount; if it is less than the minimum additional threshold, recovery is not implemented.
[0070] S8, Safety Constraint Verification and Correction Feedback. The final braking verification recalculates tire and actuator capacities, applies hard limiting to candidate torques and redistributes them among wheel ends with available margins, and performs analytical corrections based on low-speed yaw requirements; rear-wheel steering is subject to angle and speed limits applied by an independent actuator layer. Constraints are implemented in the order of fault masking, capacity verification, and command correction.
[0071] definition =max( - ,0), = - These are the braking force increment command and the constraint-corrected braking force increment, respectively. When the increment command reaches the feedback threshold, ; below the threshold =1. (The rest of the text appears to be incomplete and requires further context.) The feedback state is updated by increasing and decreasing the speed limit, and the minimum feedback coefficient is retained to allow for subsequent increases in braking force; this state and the paired increments are used in the next cycle. It can be negative, but its absolute value is not considered as the restored braking force.
[0072] S9, Control Command Output and Reversal. The total braking torque at the rear wheel end is distributed within the allowable range of the electromechanical brake and hub motor, outputting the non-negative torque of the electromechanical brake and the negative braking torque of the hub motor, and outputting the constrained rear wheel steering angle. A faulty electromechanical brake always maintains zero command; when pipeline control fails, speed falls below the recovery threshold, or execution layer protection is activated, the recovery coefficient decreases according to rules; the recovery state is reset when exiting control mode or switching to a fault, and rear wheel steering uses a rule-based reversal command.
[0073] Table 2. A set of implementation parameters and constraint settings (can be recalibrated according to vehicle and control design).
[0074] The offline parameter examples above are taken from the project parameter file; in specific implementations, the controller parameters corresponding to the test should be saved. The example recovery feedback threshold is 50 N, the minimum additional threshold is 20 N, the minimum feedback coefficient is 0.10, and the allowable braking force growth rate is 10000 N / s; the start-up waiting time and full opening time are 0.45 s and 0.75 s, respectively. Parameter values are not the sole limitation of this method and can be calibrated according to the vehicle and control design.
[0075] S10, Co-simulation and Result Analysis. CarSim and MATLAB / Simulink were used, with an initial vehicle speed of 80 km / h and a sampling period of 0.01 s. S1-S4 represent the complete failure of the left front wheel electromechanical brake, and D1-D4 represent the complete failure of both left front and left rear wheels on the same side, while the wheel hub motors remained usable. For a uniform high-adhesion road surface, μ=0.85 and braking demand was 0.6g; for a split road surface, after 10 m longitudinally, μ=0.20 on the left and μ=0.85 on the right, with a braking demand of 0.4g, the controller used an equivalent adhesion parameter of 0.525.
[0076] S101, Operating Conditions and Evaluation Methods. S1, S2, D1, and D2 do not apply measurement noise or random delay, comparing non-fault-tolerant control, conventional MPC, and Tube-MPC. S3, S4, D3, and D4 add Gaussian measurement noise to the centroid sideslip angle, yaw rate, and longitudinal acceleration, with standard deviations of approximately 0.01627°, 0.09605° / s, and 0.01g respectively, and apply synchronous random sensor delays of three to seven sampling periods, comparing conventional MPC and Tube-MPC. Tube group resumes operation with on, while conventional MPC group resumes operation with off. The non-fault-tolerant group does not provide fault codes to the controller but applies the same EMB failure to the vehicle side.
[0077] The stopping criterion is that the vehicle speed first drops to 0.5 m / s or below after 0.25 s of simulation time. The stopping distance is calculated by accumulating the Euclidean distance between adjacent vehicle plane positions before that moment; if the criterion is not met, the distance traveled within the simulation window is recorded. Average deceleration and yaw rate statistics are output from the vehicle simulation model and calculated within a window of 0.5 s to 4 s (the stopping moment) with a vehicle speed not lower than 5 m / s. Figures 5-8 As shown, each section contains six sub-graphs. The left columns (a), (c), and (e) correspond to the uniform high-adhesion working condition, while the right columns (b), (d), and (f) correspond to the open road working condition. The first row gives the deceleration, the second row gives the vehicle path, and the third row gives the yaw rate. Figure 5 and Figure 7 The results shown are simulation results without additional measurement noise and random time delay; Figure 6 and Figure 8 As shown, the deceleration and yaw rate curves include applied Gaussian measurement noise and random sensor delays, and the paths are taken from the vehicle simulation model output. Each curve is plotted up to the corresponding stopping criterion trigger time, and those that do not stop are plotted up to the recording endpoint. Figure 5 and Figure 7 The path subplot has a separate zoom-in box, displaying the regular MPC and Tube-MPC trajectories in the same coordinate units, while the main plot retains all comparisons. It should be noted that... Figure 6 and Figure 8 The measured deceleration and yaw rate in the simulation are signals superimposed with measurement noise and random sensor delays, and do not represent actual vehicle signal acquisition.
[0078] S102, Left front wheel failure without disturbance conditions S1 and S2, see Figure 5 S1 corresponds to the left column, and S2 corresponds to the right column. In S1, the stopping distances for no-fault-tolerant control, conventional MPC, and Tube-MPC are 62.92 m, 51.42 m, and 49.71 m, respectively, with maximum lateral offsets of 6.449 m, 0.227 m, and 0.130 m, respectively. The average deceleration of Tube-MPC is 0.529 g, approximately 0.023 g higher than that of conventional MPC, indicating that the proposed method can improve stopping distance and path-keeping performance while performing fault braking.
[0079] In S2, the stopping distances for the conventional MPC and Tube-MPC are 86.52 m and 86.50 m, respectively, with similar braking performance; the maximum lateral offsets are 0.143 m and 0.117 m, respectively.
[0080] S103, conditions S3 and S4, where the left front wheel failure is superimposed with Gaussian measurement noise and random sensor delay, are as follows: Figure 6As shown, S3 corresponds to the left column and S4 corresponds to the right column. In S3, Tube-MPC, compared with conventional MPC, reduces the root mean square yaw rate from 11.044 to 2.666° / s, shortens the stopping distance from 73.29 m to 52.12 m, and increases the average deceleration from 0.325g to 0.499g, demonstrating the effectiveness of the method in suppressing yaw oscillations and improving braking performance under this disturbance condition.
[0081] In S4, Tube-MPC reduces the maximum lateral offset from 5.175 m to 0.170 m, the root mean square yaw rate from 5.069 m to 0.668° / s, and the stopping distance from 109.65 m to 88.77 m compared to conventional MPC, thus improving path holding and yaw response in open-road fault braking.
[0082] S104, same-side dual-wheel failure with no disturbance conditions D1, D2, such as Figure 7 As shown, D1 corresponds to the left column and D2 corresponds to the right column. In D1, the non-tolerance group has a parking distance of 101.41 m and a maximum lateral offset of approximately 30.269 m; the parking distances for conventional MPC and Tube-MPC are 60.18 m and 57.10 m, respectively, with maximum lateral offsets of 0.501 m and 0.160 m, respectively. The average deceleration of Tube-MPC is approximately 0.461 g, indicating that the method can still achieve braking even when both wheels on the same side fail, and can shorten the parking distance and reduce path offset.
[0083] In D2, the stopping distances for the conventional MPC and Tube-MPC were 87.53 m and 87.43 m, respectively, with similar braking performance. The non-fault-tolerant group recorded no stopping at the endpoint within 10 s, and its cumulative planar trajectory length of 141.92 m within this simulation window was recorded.
[0084] S105, conditions D3 and D4, where a dual-wheel fault on the same side is superimposed with Gaussian measurement noise and random sensor delay, are as follows: Figure 8 As shown, D3 corresponds to the left column and D4 corresponds to the right column. In D3, Tube-MPC reduces the root mean square yaw rate from 11.684 to 2.250° / s, the maximum lateral offset from 1.267 to 0.552 m, and the parking distance from 76.98 m to 59.54 m compared to the conventional MPC.
[0085] In D4, Tube-MPC reduces the root mean square yaw rate from 5.523 to 1.431° / s, the maximum lateral offset from 5.199 to 1.012 m, and the stopping distance from 108.76 m to 95.02 m compared to the conventional MPC.
[0086] Based on the simulation results above, under conditions without additional measurement noise and random time delay, the proposed method maintains braking performance similar to conventional MPC, and further shortens the stopping distance and reduces path deviation under S1 and D1 conditions. Under four sets of conditions with superimposed Gaussian measurement noise and 30–70 ms random sensor delay, the root mean square yaw rate is relatively reduced by 74.1%–86.8%, and the stopping distance is shortened by 13.74–21.17 m, demonstrating the proposed method's ability to suppress yaw oscillations and improve braking performance under the combined effects of faults and measurement disturbances.
[0087] Simulation records of the recovery braking force show that the peak values of the recovery coefficients of S1 and D1 are approximately 0.784 and 0.259, respectively, and the average values of the braking force increment after constraint correction are approximately 218.16 and 292.65 N, respectively. This indicates that the recovery mechanism can generate a braking force increment according to the operating conditions and adjust the subsequent recovery command through constraint correction feedback.
[0088] The above results are used to evaluate the overall performance of the complete control scheme under the listed simulation conditions.
[0089] like Figure 4 As shown, the present invention also provides a multi-actuator cooperative braking force reconfiguration and closed-loop recovery system, the input, output and module relationships of which are as follows.
[0090] The controller inputs include target deceleration, vehicle speed, center of gravity sideslip angle, yaw rate, longitudinal acceleration, tire vertical load and longitudinal lateral force, slip information, adhesion parameters, actuator fault status and torque limit, front wheel steering angle, and rear wheel steering effectiveness, constraint occupancy and braking force correction information from the previous cycle.
[0091] The controller outputs include torque commands for the four-wheel electromechanical brakes, braking torque commands for the four-wheel hub motors, and active rear wheel steering angle commands.
[0092] The system consists of an information acquisition and fault coding module, a wheel-by-wheel available braking capacity calculation module, a baseline braking force distribution module, a rear wheel steering pipeline model predictive control module, a robust stability margin assessment module, a braking force recovery module, a safety constraint verification module, a constraint correction feedback module, and a command output and rollback module.
[0093] The information acquisition and fault coding module forms a unified input and actuator effective coefficient; the capacity calculation module outputs the upper limit for each wheel; the benchmark allocation module generates the four-wheel benchmark braking force; the rear wheel steering module outputs the candidate steering angle and its effectiveness; the margin assessment module outputs the cross-domain stability margin; the recovery module generates additional braking force; the safety verification module corrects the candidate commands; the feedback module calculates the incremental relationship before and after correction and saves it to the next cycle; the output and rollback module completes the actuator command conversion and control switching.
[0094] Therefore, the present invention adopts the above-mentioned multi-actuator braking redundancy control method and system for distributed electric drive vehicles. By adjusting the braking force recovery through the rear wheel steering robust stability margin, deceleration deviation and constraint correction feedback, the additional braking command changes with the available actuator capacity and the lateral state of the vehicle. The incremental is defined using the same period reference, and the next period command is corrected with the feedback state and growth constraint to form a closed-loop braking force recovery.
[0095] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. A method for redundant braking control of multi-actuator vehicles in a distributed electric drive vehicle, characterized in that, In the event of brake actuator failure, the following steps are implemented in coordination by the electromechanical brake, hub motor, and rear wheel steering actuator: S1. Vehicle status information acquisition and fault status coding: acquire vehicle motion status, wheel-end brake actuator fault status and road adhesion parameters, and determine the availability status of each wheel's electromechanical brake and wheel hub motor respectively. S2. Dynamic wheel load acquisition and wheel-by-wheel available braking capacity calculation: Acquire the dynamic vertical load and tire lateral force of each wheel, calculate the remaining longitudinal adhesion capacity of the tire based on the adhesion parameters, and determine the upper limit of the available torque of the corresponding actuator based on the available status of the electromechanical brake and hub motor of each wheel, so as to obtain the upper limit of the available braking force of each wheel. S3. Braking force benchmark allocation based on dynamic wheel load: The target deceleration is converted into total braking force demand, the initial braking force is allocated according to the dynamic vertical load of each wheel, the initial braking force is limited according to the upper limit of the available braking force of each wheel, and the redistribution and braking asymmetry adjustment are carried out according to the remaining capacity of each wheel to form a four-wheel benchmark braking force. S4. Braking yaw moment calculation and prediction model construction: The additional braking yaw moment is calculated based on the wheel-end braking force and wheel track. It and the front wheel steering angle are used as the exogenous inputs of the rear wheel steering prediction model, and the control targets of the center of gravity sideslip angle and yaw rate are determined. S5. Rear wheel steering based on Tube-MPC algorithm: The rear wheel steering prediction model is decomposed into a nominal system and an error system. Taking the control targets of the center of gravity sideslip angle and yaw rate as references, the optimization target of the prediction state tracking error is constructed, and the nominal rear wheel steering angle sequence is solved. The first nominal control quantity is superimposed with the error feedback correction quantity to obtain the candidate rear wheel steering angle command. Then, the steering angle and steering rate constraints are applied to form the constrained rear wheel steering angle command. S6. Cross-domain robust stability margin calculation: The stability margin is formed based on the utilization rate of the error relative to the robust positive invariant set, the rear wheel steering angle occupancy, the center of gravity sideslip angle, the yaw rate and the tire adhesion utilization rate, and the activation conditions for brake force recovery are determined in combination with the effectiveness of rear wheel steering control. S7. Generation of Braking Force Increment Command: When the conditions for restarting braking force are met, the recovery coefficient is determined based on the stability margin, deceleration deviation, feedback of the constraint correction of the previous cycle and the start-up process. Based on the four-wheel reference braking force, the braking force is allocated and superimposed according to the remaining braking capacity of each wheel actuator. The increase of the braking force after the constraint correction of the previous cycle is used to limit the growth of the increment in this cycle, thus forming candidate four-wheel braking force commands. S8. Safety constraint verification and correction feedback: The candidate four-wheel braking force command is subjected to fault actuator shielding, attachment and actuator capacity verification and analysis correction to obtain the corrected four-wheel braking force command; with the sum of the four-wheel reference braking forces in the same control cycle as a reference, the braking force increment command and the brake force increment after constraint correction are calculated respectively to form feedback coefficients and save them to the next control cycle. S9. Control command output and retraction: Convert and distribute the corrected four-wheel braking force command into braking torque commands for the electromechanical brake and wheel hub motor, and output the constrained rear wheel steering angle command; reduce or withdraw the additional braking force when the recovery condition is not met; switch to the regular rear wheel steering command when the rear wheel steering control is invalid; output the constrained rear wheel steering angle command to the rear wheel steering actuator; the faulty actuator always maintains the zero torque command.
2. The method according to claim 1, characterized in that, S1 sets the effective coefficient of the electromechanical brake and the effective coefficient of the hub motor respectively; when the electromechanical brake of a certain wheel completely fails, the upper limit of the torque and the output command of the electromechanical brake are set to zero, while the hub motor braking capability of the wheel is still usable; when both types of actuators are unavailable, the active braking torque command of the wheel is set to zero.
3. The method according to claim 1, characterized in that, S2 includes: S21. Obtain the dynamic vertical load, tire lateral force, and adhesion parameters of each wheel as input for capacity calculation; S22. Determine the adhesion boundary based on the adhesion parameters and dynamic vertical load, deduct the lateral force occupation, and determine the remaining longitudinal adhesion capacity in combination with the slip state; S23. Calculate the upper limit of the longitudinal braking force allowed by the actuator based on the effective coefficient, maximum braking torque and effective rolling radius of the electromechanical brake and hub motor. S24. Take the smaller value between the remaining longitudinal adhesion capacity of the tire and the upper limit of the longitudinal braking force allowed by the actuator as the upper limit of the available braking force of the wheel.
4. The method according to claim 1, characterized in that, S3 includes: S31. Determine the total braking force requirement based on the vehicle mass and target deceleration; S32. Distribute the initial braking force according to the proportion of the dynamic vertical load of each wheel to the total wheel load, and limit the initial value of each wheel to the upper limit of the corresponding available braking force; S33. When the total braking force demand is not yet met, the unmet portion shall be redistributed according to the proportion of the remaining capacity of each wheel, and the total amount of additional braking force shall not exceed the sum of the remaining capacity; when there is no remaining capacity, additional braking force shall be stopped. S34. Based on the fault location, rear wheel steering establishment status, and vehicle lateral movement status, adjust the longitudinal braking capacity and the degree of left-right braking asymmetry to form a reference braking force that meets the actuator and tire capacity constraints.
5. The method according to claim 1, characterized in that, S5 includes: S51. A linear two-degree-of-freedom vehicle model is adopted, with the sideslip angle and yaw rate of the center of gravity constituting the state, and the common rotation angle of the rear wheels as the control input. S52. Determine the model scheduling quantity in each sampling period, and construct known exogenous terms from the front wheel angle and the braking yaw moment of the previous period; S53. Offline design of the auxiliary feedback gain for vertex sharing, verification of closed-loop stability conditions, and calculation of the common robust positive invariant set of the error system; S54. Tighten the nominal state and input constraints according to the error set; S55. Construct an objective function that includes the predicted state deviation, the rear wheel steering angle and its change, solve the nominal control sequence in the finite prediction time domain and verify the feasibility of the constraints. S56. Add the first term of the nominal sequence to the product of the auxiliary feedback gain and the state error to form the candidate command for the rear wheel steering angle; S57. Implement angle and angle rate limits on candidate commands and provide control effectiveness and constraint occupancy information to the braking force recovery stage; S58. When the optimization calculation does not meet the feasibility requirements or the control effectiveness conditions are not met, use the constrained rule rear wheel steering command. S59. When a fault state or control mode switch occurs, reset the relevant nominal state and control sequence; when the offline disturbance envelope and invariance condition are met, use error feedback to limit the state error to the robust positive invariant set.
6. The method according to claim 1, characterized in that, S6 includes: S61. The robust positive invariant set utilization rate, rear wheel steering angle occupancy, center of gravity sideslip angle, yaw rate and tire adhesion utilization rate are respectively mapped to margin coefficients that decrease with increasing risk. S62. Take the minimum value of each margin coefficient as the cross-domain robust stability margin, and determine the recovery enable state in combination with the effectiveness of rear wheel steering control, vehicle speed and fault state.
7. The method according to claim 1, characterized in that, S5 also includes rear wheel steering feedforward and fault-related permission adjustments: S51a. Calculate the rear wheel steering angle feedforward based on the additional braking yaw moment, rear axle lateral stiffness, and distance from the center of gravity to the rear axle. The nominal optimizer then solves for the correction near this feedforward. S51b: When a single rear wheel electromechanical brake fails, a permission adjustment amount is generated based on the normalized center of gravity sideslip angle risk and yaw rate risk, and hysteresis and rate of change limits are implemented. S51c: Weight the transition between the rule-based rear wheel steering command and the pipeline model predictive control candidate command using the authority coefficient, and then implement the actuator constraint; The authority coefficient represents the degree of participation of pipeline model predictive control candidate commands in the final candidate turning angle; when the coefficient is zero, the rule-based rear wheel steering command is used, and when the coefficient is one, the pipeline model predictive control candidate command is used; front wheel failure and multiple wheel failures can maintain complete pipeline control authority.
8. The method according to claim 1, characterized in that, S7 includes: S71. Based on the difference between the target deceleration and the measured deceleration, a deceleration deviation coefficient is formed, and the ratio of the braking force increment after the constraint correction in the previous cycle to the corresponding increment command is converted into a feedback coefficient. S72. Determine the target recovery coefficient based on the cross-domain robust stability margin, deceleration deviation coefficient, feedback coefficient, and initiation coefficient, and impose limits on the rise and fall rates of the recovery coefficient. S73. Determine the remaining capacity of each wheel based on the baseline braking force and actuator capacity, limit the total recovery to the range of total demand gap and remaining capacity allowable, and then form candidate increments according to the proportion of remaining capacity. S74. When the incremental command of the previous cycle reaches the preset feedback threshold, the non-negative part of the incremental after the constraint correction of the current cycle plus the allowable growth amount constitutes the upper limit of the incremental in the current cycle; when the fault state or control mode changes, the reset and feedback state are restored.
9. The method according to claim 1, characterized in that, S8 uses the total braking force of the same period as a reference to calculate the braking force increment before and after constraint correction; The feedback ratio uses the non-negative part of the increment after constraint correction and is limited to between zero and one. Independent processing branches are set for incremental instructions that are less than the preset threshold. S9 reduces the recovery coefficient when the recovery conditions are not met, and withdraws the recovery increment according to the reset rules when the control mode is exited or the fault state is switched.
10. A distributed electric drive vehicle multi-actuator braking redundancy control system, used to implement the method according to any one of claims 1-9, characterized in that: It includes a controller and, executed by the controller, an information acquisition and fault coding module, a wheel-by-wheel available braking capacity calculation module, a baseline braking force distribution module, a rear wheel steering pipeline model predictive control module, a robust stability margin assessment module, a braking force recovery module, a safety constraint verification module, a constraint correction feedback module, and a command output and rollback module; the constraint correction feedback module feeds back the braking force increment after safety constraint verification to the braking force recovery module, which is used to adjust the braking force increment command for the next control cycle.
