Real-time torque optimal distribution method and device and storage medium
By constructing solvers and predictors, and combining multi-objective optimization and neural networks, the appropriate initial additional torque is automatically predicted, solving the problems of slow calculation speed and inaccurate iteration direction in existing technologies, and achieving efficient torque optimization allocation.
Patent Information
- Application Number
- CN202511402442.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-28
- Publication Date
- 2026-01-09
AI Technical Summary
In existing technologies, the iterative algorithm for solving the optimal additional torque for four wheels of a vehicle using multi-objective optimization requires significant computing power and it is difficult to determine a suitable initial additional torque, which affects the calculation speed and iteration direction.
A solver and predictor are constructed, and a multi-objective optimization torque vector allocation algorithm and neural network are used. The loss function is trained by gradient backpropagation to automatically predict the appropriate initial additional torque, ensuring that the numerical value and iteration direction are close to the optimum.
It significantly improves computing speed, reduces computing power, ensures the accuracy of the initial additional torque and the correctness of the iteration direction, and avoids the uncontrollability of neural networks.
Smart Images
Figure CN121302875A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of intelligent torque distribution in automobiles, specifically to a real-time torque optimization distribution method, device, and storage medium. Background Technology
[0002] Multi-objective torque vector allocation algorithms based on MPC (Model Predictive Control) are widely used in research on longitudinal and lateral cooperative stability control of distributed drive vehicles. The algorithm receives vehicle motion state information and road adhesion information from onboard sensors and vehicle state estimation outputs, forming a multi-objective optimization problem to solve for the additional torque of the four wheels. Through iterative solving of the multi-objective optimization, the optimal additional torque for the four wheels of the distributed drive vehicle is finally output to adjust the vehicle's motion attitude under different operating conditions and ensure stable driving. However, the iterative solution of this multi-objective optimization requires significant computing power, which is usually insufficient for real-world vehicle controllers.
[0003] To accelerate the calculation process, it is necessary to set four appropriate initial additional torques for iteration during the solution process; however, determining the appropriate initial additional torque is quite difficult. The appropriate initial additional torque must not only be close to the optimal additional torque in terms of numerical value, but also close to the optimal additional torque in terms of iteration direction.
[0004] Neural networks have a strong ability to fit nonlinear functions. Through early training, neural networks can fit complex functions well and perform fast calculations during subsequent inference. However, neural networks also have problems such as poor interpretability and incomplete controllability. Summary of the Invention
[0005] To address the shortcomings of existing technologies, the technical problem solved by this application is: how to provide an initial additional torque that is close to the optimal additional torque in both numerical value and iteration direction when calculating the optimal additional torque for the four wheels of a vehicle, thereby significantly improving the calculation speed and reducing the computing power.
[0006] To achieve the above objectives, in a first aspect, embodiments of this application provide a real-time torque optimization distribution method, which includes the following steps: Construct a solver to find the optimal additional torque for the four wheels of a vehicle. The solver takes vehicle state parameters as input and outputs the optimal additional torque. Construct a predictor to determine the initial additional torque; Design a loss function for training the predictor via gradient backpropagation; The vehicle state parameters input to the solver are divided into a training set and a test set; the predictor is trained using the training set to predict the initial additional torque; when the difference between the initial additional torque predicted by the trained predictor on the test set and the optimal additional torque obtained by the solver using the test set is below a specified tolerance, the initial additional torque obtained by the predictor is determined to be usable.
[0007] In conjunction with the first aspect, in one implementation, the solver construction process includes: Construct a two-degree-of-freedom dynamic model of the vehicle; Determine the optimal additional yaw moment based on the two-degree-of-freedom dynamic model; The optimal additional torque is distributed to the four wheels based on the optimal additional yaw moment.
[0008] In conjunction with the first aspect, in one implementation, the two-degree-of-freedom dynamic model is expressed as: , ; in: Represents the lateral forces of the left and right front wheels. Represents the lateral forces of the left and right rear wheels; This represents the distance from the front axle to the vehicle's center of gravity. This represents the distance from the rear axle to the vehicle's center of gravity. Represents vehicle quality. The moment of inertia representing the vehicle's rotation about its center of mass. Represents longitudinal velocity. Represents the centroid sideslip angle. Represents yaw rate. This represents the additional yaw moment.
[0009] In conjunction with the first aspect, in one implementation, the process of determining the optimal additional yaw moment based on the two-degree-of-freedom dynamic model includes: constructing an optimization objective function for the future yaw moment of the vehicle:
[0010] in: The first item represents yaw rate tracking control. The weight of the first item; Let t represent the desired yaw rate; t represents the current time. Represents the predicted duration; The second item represents the suppression of the vehicle's center of gravity sideslip angle. The weight of the second term; The third item represents the limitation on the vehicle's kinetic energy. ; The weight of the third term; The linearized objective function is obtained by performing a Taylor expansion on the optimization objective function:
[0011] in To perform a first-order Taylor expansion on the first term, To perform a second-order Taylor expansion on the second term, the stationary point corresponding to the linearized objective function... That is, the target optimal additional yaw moment .
[0012] In conjunction with the first aspect, in one implementation, the process of distributing the optimal additional torque to the four wheels according to the optimal additional yaw moment includes: Define the vector of optimal additional torque for the four wheels as follows: Define the rotational speed vector of the four wheels as , For the desired rotational speed, Indicates the additional torque of the motor. Represents slack variables. Indicates the rotational speed of the four wheels; Solving through the objective function and The objective function is:
[0013] Where s represents the current state. Represents the discrete prediction time domain. Represents the total prediction time domain; Used for optimal yaw moment tracking Used for four-wheel wheel speed tracking Used for actuation energy suppression, and must also meet the following requirements. and tire slip ratio The upper and lower bounds.
[0014] In conjunction with the first aspect, in one implementation, the process of constructing the predictor includes: Based on spatial parameters that reflect the spatial information of vehicles, a spatial information extraction network for vehicles is constructed. Based on the temporal parameters that reflect the temporal motion information of the vehicle during driving, a temporal information extraction network for the vehicle is constructed. Construct an optimal solution mapping network; use the outputs of the spatial information network and the motion information network as inputs to the optimal solution mapping network, and use the output of the optimal solution mapping network as the initial additional torque.
[0015] In conjunction with the first aspect, in one implementation, the iterative process by which the solver solves the test set is as follows:
[0016] in Indicates the initial additional torque. Custom settings; k represents the number of iterations. Iteration The optimal additional torque obtained from the subsequent test, where Q represents the quadratic programming algorithm. Indicates vehicle status parameters; The iterative process by which the solver obtains the initial additional torque solution through the training set and the predictor is as follows:
[0017]
[0018] Where j represents the iteration number, j≤k, Let W represent the optimal additional torque obtained after j iterations for training, and W represent the predictor.
[0019] In conjunction with the first aspect, in one implementation, the loss function Loss is:
[0020] The L1 loss is the output of the predictor. , for The weights; The L2 loss is the sum of the predictor output and the solver output. , for The weights; The L1 loss is the output of the solver. , for The weight.
[0021] Secondly, embodiments of this application provide a real-time torque optimization allocation device, the real-time torque optimization allocation device including a processor, a memory, and a real-time torque optimization allocation program stored in the memory and executable by the processor, wherein when the real-time torque optimization allocation program is executed by the processor, the method provided in the first aspect is implemented.
[0022] Thirdly, embodiments of this application provide a computer-readable storage medium storing a real-time torque optimization allocation program, which, when executed, implements the method provided in the first aspect.
[0023] Compared with the prior art, the advantages of this application are: This application constructs a solver for finding the optimal additional torque of the four wheels of a vehicle based on a multi-objective optimization torque vector allocation algorithm, and constructs a predictor for determining the initial additional torque based on a neural network. The predictor is continuously trained using the input (vehicle state parameters) and output (optimal additional torque) of the solver until the initial additional torque predicted by the predictor reaches below the target error.
[0024] Therefore, this application fully couples and leverages the advantages of neural networks and multi-objective optimization torque vector allocation algorithms. By coupling neural network prediction with multi-objective optimization iterative solution, it automatically predicts the appropriate initial additional torque, ensuring that the value and iterative direction are close to the optimal additional torque, thereby significantly improving the calculation speed and reducing the computing power. At the same time, the correction of the multi-objective optimization algorithm avoids the problem of incomplete control of the neural network. Attached Figure Description
[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0026] Figure 1 This is a flowchart of the real-time torque optimization allocation method in the embodiments of this application; Figure 2 This is a schematic diagram of the algorithm model of the coupled neural network in the embodiments of this application; Figure 3 This is a schematic diagram of the loss function in an embodiment of this application.
[0027] Figure 4 This is a schematic diagram of the hardware structure of the real-time torque optimization and distribution device involved in the embodiments of this application. Detailed Implementation
[0028] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0029] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.
[0030] First, the research and development principles of this application will be briefly described.
[0031] Neural networks have a strong ability to fit nonlinear functions. Through early training, neural networks can fit complex functions well and perform fast calculations during subsequent inference. However, neural networks also have problems such as poor interpretability and incomplete controllability.
[0032] The distributed drive vehicle torque vector allocation method, which couples neural networks with multi-objective optimization, combines the advantages of neural networks and MPC. By coupling neural network prediction with multi-objective optimization iterative solution, it can automatically predict the appropriate initial additional torque and solve the above-mentioned technical problems.
[0033] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.
[0034] Firstly, embodiments of this application provide a real-time torque optimization allocation method, see [link to relevant documentation]. Figure 1 As shown, the steps of this method include: S1: A solver is constructed based on a multi-objective optimization torque vector allocation algorithm to solve for the optimal additional torque of the four wheels of a vehicle. The input of the solver is the vehicle state parameters, and the output is the optimal additional torque.
[0035] Examples of the vehicle status parameters mentioned above include: vehicle motion status information and road surface adhesion information, specifically including data such as vehicle speed, steering angle, road surface adhesion rate, acceleration, and yaw rate.
[0036] S2: Construct a predictor based on a neural network to determine the initial additional torque.
[0037] S3: The loss function designed for training the predictor via gradient backpropagation.
[0038] S4: Construct the algorithm model of the coupled neural network, with the predictor as the first half of the algorithm model and the solver as the second half.
[0039] S5: Divide the vehicle state parameters input to the solver into a training set and a test set; train the predictor through the training set to obtain the initial additional torque; the solver solves based on the initial additional torque and the training set to obtain the optimal additional torque corresponding to the training set (i.e., the optimal additional torque for training); solve the test set through the solver to obtain the optimal additional torque corresponding to the test set (i.e., the optimal additional torque for testing); when the difference between the initial additional torque predicted by the trained predictor on the test set and the optimal additional torque for testing is below the specified tolerance, the initial additional torque obtained by the predictor is determined to be usable.
[0040] Therefore, this application constructs a solver for finding the optimal additional torque of the four wheels of a vehicle based on a multi-objective optimization torque vector allocation algorithm, and constructs a predictor for determining the initial additional torque based on a neural network. The predictor is continuously trained using the input (vehicle state parameters) and output (optimal additional torque) of the solver until the initial additional torque predicted by the predictor reaches below the target error.
[0041] Therefore, this application fully couples and leverages the advantages of neural networks and multi-objective optimization torque vector allocation algorithms. By coupling neural network prediction with multi-objective optimization iterative solution, it automatically predicts the appropriate initial additional torque, ensuring that the value and iterative direction are close to the optimal additional torque, thereby significantly improving the calculation speed and reducing the computing power. At the same time, the correction of the multi-objective optimization algorithm avoids the problem of incomplete control of the neural network.
[0042] In one embodiment, the construction process of the multi-objective optimized torque vector allocation algorithm in S1 includes: S101: Construct a two-degree-of-freedom dynamic model of the vehicle; S102: Determine the optimal additional yaw moment based on the two-degree-of-freedom dynamic model; S103: Distribute the optimal additional torque to the four wheels based on the optimal additional yaw moment.
[0043] Based on this, in one embodiment, the two-degree-of-freedom dynamic model is expressed as: , ; in: Represents the lateral forces of the left and right front wheels. This represents the lateral force of the left and right rear wheels. The lateral force is calculated using tire models, such as the brush model or the magic formula tire model. This represents the distance from the front axle to the vehicle's center of gravity. This represents the distance from the rear axle to the vehicle's center of gravity. Represents vehicle quality. The moment of inertia representing the vehicle's rotation about its center of mass. Represents longitudinal velocity. Represents the centroid sideslip angle. Represents yaw rate. This represents the additional yaw moment.
[0044] Furthermore, the process of determining the optimal additional yaw moment based on the two-degree-of-freedom dynamic model includes: constructing the optimization objective function for the future yaw moment of the vehicle, expressed as:
[0045] in: The first item represents yaw rate tracking control. The weight of the first item; For the desired yaw rate, The desired front wheel steering angle, derived from the driver's steering wheel input, is converted using a second-order reference model; t represents the current moment. Represents the predicted duration; The second item represents the suppression of the vehicle's center of gravity sideslip angle. The weight of the second term; The third item represents the limitation on the vehicle's kinetic energy. ; The weight of the third item.
[0046] The linearized objective function is obtained by performing a Taylor expansion on the optimization objective function, and its expression is:
[0047] in To perform a first-order Taylor expansion on the first term, To perform a second-order Taylor expansion on the second term, P1 and P2 are the mathematical notations for the Taylor expansion. The third term is retained, and it corresponds to the stationary point of the linearized objective function. That is, the target optimal additional yaw moment .
[0048]
[0049] Furthermore, the process of distributing the optimal additional torque to the four wheels based on the optimal additional yaw moment includes: Because it is necessary to comprehensively consider parameters such as the optimal additional yaw moment and tire condition, a multi-objective optimization problem is constructed to solve for the optimal four-wheel torque distribution. Therefore, the vector of the optimal additional torque for the four wheels is defined as follows: Define the rotational speed vector of the four wheels as , The desired rotational speed is synthesized from the vehicle's longitudinal velocity and the desired yaw rate. Indicates the additional torque of the motor. Represents slack variables. This indicates the rotational speed of the four wheels.
[0050] Solving through the objective function and The expression for the objective optimization function is:
[0051] Where s represents the current state. Represents the discrete prediction time domain. Represents the total prediction time domain; Used for optimal yaw moment tracking Used for four-wheel wheel speed tracking Used for actuation energy suppression, and should also meet the following requirements and tire slip ratio The upper and lower bounds.
[0052] In one embodiment, see Figure 2 As shown, the predictor construction process in S2 includes: S201: Construct a spatial information extraction network for the vehicle based on spatial parameters that reflect the vehicle's spatial information (these parameters will not change). Spatial parameters include vehicle mass, moment of inertia, wheelbase, etc.
[0053] S202: Construct a vehicle temporal information extraction network based on temporal parameters that reflect the vehicle's motion during driving (these parameters change at every moment). Temporal parameters include vehicle speed, steering wheel angle, yaw rate, etc.
[0054] S203: On top of the spatial information extraction network and the temporal motion information extraction network, construct an optimal solution mapping network; use the output of the spatial information network and the output of the motion information network as inputs to the optimal solution mapping network, and use the output of the optimal solution mapping network as the initial additional torque.
[0055] In one embodiment, the iterative process of solving for the test set and the custom initial additional torque using a solver is represented as follows:
[0056] in This indicates the initial additional torque, here. Custom settings; k represents the number of iterations. Iteration The optimal additional torque obtained after this test is Q, which stands for QP (Quadratic Programming). This indicates vehicle status parameters.
[0057] The solution principle is as follows: The above objective optimization function constructs an additional torque allocation problem under the objectives of optimal yaw moment tracking, desired four-wheel speed tracking, and kinetic energy suppression, with slip ratio and upper and lower limits of total additional torque as constraints. To solve this algorithmic problem, an iterative method to approximate the optimal solution is required. This application uses QP to iterate the objective optimization function.
[0058] Meanwhile, the iterative process of obtaining the initial additional torque from the training set and predictor through the solver can be expressed as:
[0059]
[0060] Where j represents the iteration number, j≤k, Let W represent the optimal additional torque obtained after j iterations for training, and W represent the predictor, i.e. Obtained by the predictor.
[0061] In one embodiment, the loss function Loss in S3 is:
[0062] See Figure 3 As shown: The L1 loss (absolute error loss) is the output of the predictor. , for The weight, It can bring the predicted initial additional torque closer to the value of the optimal additional torque; L2 loss (mean squared error loss) is the sum of the predictor output and the solver output. , for The weight, This loss guides the prediction iteration direction to approach the initial additional torque of the optimal additional torque, ensuring that the initial solution output by the predictor will not deviate from the optimal solution in the iteration direction.
[0063] The L1 loss is the output of the solver. , for The weight, Let the solution output by the algorithm model of the coupled neural network after j iterations approximate the solution output by the solver after k iterations.
[0064] In one embodiment, in S5, the training set comprises 60-80% of the total vehicle state parameters input to the solver, and the test set comprises 20-40% of the total vehicle state parameters input to the solver. Secondly, embodiments of this application provide a real-time torque optimization and distribution device, which can be a device with data processing capabilities such as a personal computer (PC), a laptop computer, or a server.
[0065] Reference Figure 4 , Figure 4 This is a schematic diagram of the hardware structure of the real-time torque optimization and allocation device involved in the embodiments of this application. In the embodiments of this application, the real-time torque optimization and allocation device may include a processor, a memory, a communication interface, and a communication bus.
[0066] The communication bus can be of any type and is used to interconnect the processor, memory, and communication interface.
[0067] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces used for interconnecting internal components of the real-time torque optimization distribution device, as well as interfaces used for interconnecting the real-time torque optimization distribution device with other devices (such as other computing devices or user equipment). Physical interfaces can be Ethernet interfaces, fiber optic interfaces, ATM interfaces, etc.; user equipment can be displays, keyboards, etc.
[0068] Memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.
[0069] The processor can be a general-purpose processor, which can call a real-time torque optimization allocation program stored in memory and execute the real-time torque optimization allocation method provided in the embodiments of this application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed when the real-time torque optimization allocation program is called can be referred to in various embodiments of the real-time torque optimization allocation method of this application, and will not be repeated here.
[0070] Those skilled in the art will understand that Figure 4 The hardware structure shown does not constitute a limitation of this application and may include more or fewer components than shown, or combine certain components, or have different component arrangements.
[0071] Thirdly, embodiments of this application also provide a computer-readable storage medium.
[0072] The computer-readable storage medium of this application stores a real-time torque optimization allocation program, wherein when the real-time torque optimization allocation program is executed by a processor, it implements the steps of the real-time torque optimization allocation method as described above.
[0073] The method implemented when the real-time torque optimization allocation program is executed can be referred to in various embodiments of the real-time torque optimization allocation method of this application, and will not be repeated here.
[0074] It should be noted that the sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.
[0075] The terms "comprising" and "having," and any variations thereof, in the specification, claims, and accompanying drawings of this application are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to such process, method, product, or apparatus. The terms "first," "second," and "third," etc., are used to distinguish different objects, etc., and do not indicate a sequence, nor do they limit "first," "second," and "third" to different types.
[0076] In the description of the embodiments of this application, terms such as "exemplary," "for example," or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as "exemplary," "for example," or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of terms such as "exemplary," "for example," or "for instance" is intended to present the relevant concepts in a concrete manner.
[0077] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. The "and / or" in the text is merely a description of the relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist simultaneously, and B exists alone. In addition, in the description of the embodiments of this application, "multiple" means two or more.
[0078] In some processes described in the embodiments of this application, multiple operations or steps are included in a specific order. However, it should be understood that these operations or steps may not be executed in the order they appear in the embodiments of this application, or they may be executed in parallel. The sequence number of the operation is only used to distinguish different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed sequentially or in parallel, and these operations or steps may be combined.
[0079] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes several instructions to cause a terminal device to execute the methods described in the various embodiments of this application.
[0080] The above are merely specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the scope of the claims.
Claims
1. A real-time torque optimization allocation method, characterized in that, The method includes the following steps: Construct a solver to find the optimal additional torque for the four wheels of a vehicle. The solver takes vehicle state parameters as input and outputs the optimal additional torque. Construct a predictor to determine the initial additional torque; Design a loss function for training the predictor via gradient backpropagation; The vehicle state parameters input to the solver are divided into a training set and a test set; the predictor is trained using the training set to predict the initial additional torque; when the difference between the initial additional torque predicted by the trained predictor on the test set and the optimal additional torque obtained by the solver using the test set is below a specified tolerance, the initial additional torque obtained by the predictor is determined to be usable.
2. The real-time torque optimization distribution method as described in claim 1, characterized in that, The process of constructing the solver includes: Construct a two-degree-of-freedom dynamic model of the vehicle; Determine the optimal additional yaw moment based on the two-degree-of-freedom dynamic model; The optimal additional torque is distributed to the four wheels based on the optimal additional yaw moment.
3. The real-time torque optimization distribution method as described in claim 2, characterized in that, The two-degree-of-freedom dynamic model is expressed as follows: ， ； in: Represents the lateral forces of the left and right front wheels. Represents the lateral forces of the left and right rear wheels; This represents the distance from the front axle to the vehicle's center of gravity. This represents the distance from the rear axle to the vehicle's center of gravity. Represents vehicle quality. The moment of inertia representing the vehicle's rotation about its center of mass. Represents longitudinal velocity. Represents the centroid sideslip angle. Represents yaw rate. This represents the additional yaw moment.
4. The real-time torque optimization distribution method as described in claim 3, characterized in that, The process of determining the optimal additional yaw moment based on the two-degree-of-freedom dynamic model includes: constructing an optimization objective function for the future yaw moment of the vehicle. in: The first item represents yaw rate tracking control. The weight of the first term; Let t represent the desired yaw rate; t represents the current time. Represents the predicted duration; The second item represents the suppression of the vehicle's center of gravity sideslip angle. The weight of the second term; The third item represents the limitation on the vehicle's kinetic energy. ; The weight of the third term; The linearized objective function is obtained by performing a Taylor expansion on the optimization objective function: in To perform a first-order Taylor expansion on the first term, To perform a second-order Taylor expansion on the second term, the stationary point corresponding to the linearized objective function... That is, the target optimal additional yaw moment .
5. The real-time torque optimization distribution method as described in claim 4, characterized in that, The process of distributing the optimal additional torque to the four wheels based on the optimal additional yaw moment includes: Define the vector of optimal additional torque for the four wheels as follows: Define the rotational speed vector of the four wheels as , For the desired rotational speed, Indicates the additional torque of the motor. Represents slack variables. Indicates the rotational speed of the four wheels; Solving through the objective function and The objective function is: Where s represents the current state. Represents the discrete prediction time domain. Represents the total prediction time domain; Used for optimal yaw moment tracking Used for four-wheel wheel speed tracking Used for actuation energy suppression, and must also meet the following requirements. and tire slip ratio The upper and lower bounds.
6. The real-time torque optimization distribution method as described in claim 1, characterized in that, The process of constructing the predictor includes: Based on spatial parameters that reflect the spatial information of the vehicle, a spatial information extraction network for the vehicle is constructed. Based on the temporal parameters that reflect the temporal motion information of the vehicle during driving, a temporal information extraction network for the vehicle is constructed. Construct an optimal solution mapping network; use the outputs of the spatial information network and the motion information network as inputs to the optimal solution mapping network, and use the output of the optimal solution mapping network as the initial additional torque.
7. The real-time torque optimization distribution method as described in claim 1, characterized in that, The iterative process by which the solver solves the problem using the test set is as follows: in Indicates the initial additional torque. Custom settings; k represents the number of iterations. Iteration The optimal additional torque obtained from the subsequent test, where Q represents the quadratic programming algorithm. Indicates vehicle status parameters; The iterative process by which the solver obtains the initial additional torque solution through the training set and the predictor is as follows: Where j represents the iteration number, j≤k, Let W represent the optimal additional torque obtained after j iterations for training, and W represent the predictor.
8. The real-time torque optimization distribution method as described in claim 1, characterized in that, The loss function Loss is: The L1 loss is the output of the predictor. , for The weights; The L2 loss is the sum of the predictor output and the solver output. , for The weights; The L1 loss is the output of the solver. , for The weight.
9. A real-time torque optimization distribution device, characterized in that, The real-time torque optimization allocation device includes a processor, a memory, and a real-time torque optimization allocation program stored in the memory and executable by the processor, wherein when the real-time torque optimization allocation program is executed by the processor, it implements the steps of the real-time torque optimization allocation method as described in any one of claims 1 to 8.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a real-time torque optimization allocation program, wherein when the real-time torque optimization allocation program is executed, it implements the steps of the real-time torque optimization allocation method as described in any one of claims 1 to 8.