An electronic power steering control method and device, electronic equipment and storage medium

Abstract

The present application relates to the field of electronic power steering control, and particularly relates to an electronic power steering control method and device, electronic equipment and storage medium. The present application obtains steering request and driving data in the process of vehicle driving; the steering request carries target steering information; a control current prediction model is used to predict the control current of the assist motor based on the target steering information and the driving data, and the target control current of the assist motor is obtained; the control current prediction model is obtained by training a plane neural network based on historical target steering information, historical driving data and corresponding historical control current; and the steering wheel is controlled based on the target control current of the assist motor. In this way, the non-linear steering of the electric power steering device in the full speed range can be effectively assisted, the safety and stability of the vehicle during operation are ensured, and the traffic accidents are reduced and the steering conditions are improved.

Description

Technical Field

[0001] This application relates to the field of electronic power steering control, and in particular to an electronic power steering control method, device, electronic equipment and storage medium. Background Technology

[0002] With increasing demands for vehicle handling performance and the development of large-scale integrated circuit technology, automotive steering systems are evolving from simple mechanical steering to electronic power steering (EPS) systems. EPS improves driver comfort while also enhancing the performance and reliability of the steering system. Simultaneously, with the implementation of energy-saving and environmental protection concepts, the demand for low-power, high-precision, and high-reliability EPS systems is growing.

[0003] The development of automotive power steering has been a long process, evolving from hydraulic power steering to electro-hydraulic power steering, and finally to EPS systems. The selection of control strategies and algorithms within EPS systems are crucial factors determining their performance. A well-designed EPS system can significantly reduce the steering force required by the driver, thereby improving the overall system's ease of use, responsiveness, and real-time performance. Domestic and international scholars have conducted extensive and in-depth research on improving EPS system performance, primarily simplifying the EPS system into a linear system, such as using traditional proportional-integral-derivative (PID) controllers or improved PID controllers for system control and regulation.

[0004] However, due to the variable road environment and the uncontrollable external interference factors such as the friction of the tires, there will inevitably be errors and uncertainties between the electric power steering device established under ideal conditions and the actual controlled object. Therefore, for nonlinear electric power steering systems, the control algorithm based on traditional modern control theory cannot achieve its accurate control state, resulting in poor robustness of the control system. Summary of the Invention

[0005] To address the problems of existing technologies, this application provides an electronic power steering control method, device, electronic equipment, and storage medium. The technical solution is as follows:

[0006] On the one hand, an electronic power steering control method is provided, the method comprising:

[0007] Acquire steering requests and driving data during vehicle operation; the steering request carries target steering information;

[0008] Using a control current prediction model, the control current of the power assist motor is predicted based on the target steering information and driving data to obtain the target control current of the power assist motor. The control current prediction model is obtained by training a planar neural network based on historical target steering information, historical driving data and corresponding historical control current.

[0009] The steering wheel is assisted to control the steering based on the target control current of the power steering motor.

[0010] In one feasible implementation, the control current prediction model includes an input layer, an intermediate layer, and an output layer connected in sequence; the intermediate layer includes a feature layer and an enhancement layer; the input layer consists of input data input to the control current prediction model; the output layer is connected to the feature layer and the enhancement layer respectively; the method for training the control current prediction model includes:

[0011] Obtain the training set; the training set includes the historical steering parameters at a first number of historical moments and the historical control current at each historical moment; the historical steering parameters include historical target steering information and historical driving data;

[0012] The first nonlinear activation function is used to perform feature mapping on all historical turning parameters in the training set to generate the feature matrix of the feature layer.

[0013] The feature matrix of the feature layer is extracted using the second nonlinear activation function to generate the feature matrix of the enhancement layer.

[0014] The target weight parameters of the control current prediction model are determined based on the feature matrix of the feature layer, the feature matrix of the enhancement layer, and the corresponding historical control currents in the training set, thus obtaining the control current prediction model.

[0015] In one feasible implementation, a first nonlinear activation function is used to perform feature mapping processing on all historical turning parameters in the training set to generate a feature matrix for the feature layer, including:

[0016] Obtain the first random weight coefficient and the first random bias; the first random weight coefficient follows a Gaussian distribution.

[0017] The first parameter matrix is ​​determined based on the first random weight coefficient, the first random bias, and all historical turning parameters in the training set;

[0018] The first nonlinear activation function is used to perform feature mapping on the first parameter matrix to generate the feature matrix of the feature layer;

[0019] In one feasible implementation, a second nonlinear activation function is used to perform feature extraction processing on the feature matrix of the feature layer to generate the feature matrix of the enhancement layer, including:

[0020] Obtain the second random weight coefficient and the second random bias; the second random weight coefficient follows a Gaussian distribution.

[0021] The second parameter matrix is ​​determined based on the second random weight coefficients, the second random bias, and the feature matrix of the feature layer;

[0022] The second nonlinear activation function is used to perform feature extraction on the second parameter matrix to obtain the feature matrix of the enhancement layer.

[0023] In one feasible implementation, the target weight parameters of the control current prediction model are determined based on the feature matrix of the feature layer, the feature matrix of the enhancement layer, and the historical control currents corresponding to the training set. After obtaining the control current prediction model, the method further includes:

[0024] Obtain a test set, which includes the historical steering parameters at a second number of historical moments and the historical control current at each historical moment;

[0025] The historical data of the second number of historical moments are used as input parameters to the control current prediction model to obtain the predicted control current for each historical moment.

[0026] Determine the error value between the predicted control current and the corresponding historical control current for each historical steering parameter in the test set.

[0027] The model accuracy value is determined based on the error values ​​corresponding to each historical steering parameter in the test set;

[0028] If the model accuracy value is greater than the preset threshold, the number of enhancement nodes in the enhancement layer of the control current prediction model is increased, and the control current prediction model is retrained until the model accuracy value of the control current prediction model is less than or equal to the preset threshold.

[0029] In one feasible implementation, obtaining the training set includes:

[0030] Acquire historical steering wheel information and driver's historical steering force at multiple historical moments;

[0031] The historical steering wheel information and the driver's historical steering force at multiple historical moments are input into the steering driving style prediction model to obtain the driver's steering driving style. The steering driving style prediction model is obtained by training a planar neural network based on the steering wheel information, the driver's steering force, and the corresponding driving style.

[0032] Acquire the historical steering parameters and historical control current for each historical moment, including the historical target steering information and historical driving data.

[0033] For each historical moment, the historical control current is adjusted based on the driver's steering style to generate the target historical control current.

[0034] The training set is constructed based on the historical turning parameters and the corresponding target historical control current of each historical moment in the first set of historical moments.

[0035] In one feasible implementation, the characteristic is that acquiring steering requests and driving data during vehicle operation includes:

[0036] Obtain the distance between the vehicle and the obstacles on both sides of the vehicle;

[0037] If the distance between the vehicle and obstacles on both sides of the vehicle is less than or equal to a first preset distance value, then in response to the user's first control command on the control interface, a steering request is generated based on the distance value, and driving data is acquired during the vehicle's driving process; or, in response to the user's second control command on the control interface, the steering request and driving data are acquired during the vehicle's driving process; the first control command indicates that the vehicle is allowed to enter the assisted driving mode; the second control command indicates that the vehicle maintains the manual driving mode.

[0038] On the other hand, an electronic power steering control device is provided, the device comprising:

[0039] The acquisition module is used to acquire steering requests and driving data during vehicle operation; the steering request carries target steering information.

[0040] The prediction module is used to predict the control current of the power assist motor based on the target steering information and driving data using the control current prediction model, so as to obtain the target control current of the power assist motor. The control current prediction model is obtained by training a planar neural network based on historical target steering information, historical driving data and corresponding historical control current.

[0041] The control module is used to control the steering wheel with power steering based on the target control current of the power steering motor.

[0042] On the other hand, an electronic device is provided, including a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the electronic power steering control method of any of the above aspects.

[0043] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction or at least one program is stored therein, the at least one instruction or the at least one program being loaded and executed by a processor to implement the electronic power steering control method as described above.

[0044] This application embodiment acquires steering requests and driving data during vehicle operation; the steering request carries target steering information; using a control current prediction model, the control current of the power steering motor is predicted based on the target steering information and driving data to obtain the target control current of the power steering motor; the control current prediction model is obtained by training a planar neural network based on historical target steering information, historical driving data, and corresponding historical control currents; based on the target control current of the power steering motor, the steering wheel is assisted in steering control. In this way, nonlinear steering of the electric power steering device can be effectively assisted across the entire speed range. This not only solves the problems of complex control algorithms and poor stability in traditional electronic power steering systems, but also ensures the safety and stability of the vehicle during operation, playing an important role in reducing traffic accidents and improving steering conditions, and also improving the user experience. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of this application, 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 this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a schematic diagram of an implementation environment provided in an embodiment of this application;

[0047] Figure 2 This is a schematic flowchart of an electronic power steering control method provided in an embodiment of this application;

[0048] Figure 3 This is a schematic diagram of the operation of a power assist motor provided in an embodiment of this application;

[0049] Figure 4 This is a simplified flowchart illustrating how to determine the control current under different driving modes, as provided in an embodiment of this application.

[0050] Figure 5 This is a road scene diagram provided in an embodiment of this application;

[0051] Figure 6 This is a schematic diagram of a process for training a control current prediction model provided in an embodiment of this application;

[0052] Figure 7 This is a schematic diagram of a width learning model provided in an embodiment of this application;

[0053] Figure 8 This is a schematic diagram of another training control current prediction model provided in an embodiment of this application;

[0054] Figure 9 This is a structural block diagram of an electronic power steering device provided in an embodiment of this application. Detailed Implementation

[0055] 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, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0056] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0057] It is understood that in the specific embodiments of this application, data such as user information are involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0058] Please see Figure 1The diagram illustrates an implementation environment provided in this application embodiment. This environment includes an electronic power steering control system, which comprises a steering wheel, sensors, an electronic control unit (ECU), a power steering motor, a gear mechanism, radar, and wheels. Sensors (which may be torque angle sensors) are mounted on a connecting rod under the steering wheel to collect steering wheel angle and angular velocity, and to obtain steering requests carrying target steering information. The ECU acquires driving data during vehicle operation and uses a control current prediction model to predict the control current of the power steering motor based on the target steering information and driving data, thus obtaining the target control current of the power steering motor. The control current prediction model is obtained by training a planar neural network based on historical target steering information, historical driving data, and corresponding historical control currents. Subsequently, the target control current is converted into steering torque to control the gear mechanism and the wheels connected to the gear mechanism to steer, thereby achieving power steering operation.

[0059] Optionally, the electronic control unit can be an independent control device used to analyze various data on the vehicle, process multiple modules, and handle various operating conditions. However, this introduces a combinatorial explosion problem into data processing. On the other hand, since there may be processing time delays and communication delays during the transmission of commands for execution and feedback of execution status, functional safety issues such as intelligent control delays may arise. To avoid the above problems, the electronic control unit can also be a processing unit located in the power steering system. It can act not only as an actuator but also as a processor, capable of independently processing intelligent steering judgments, avoiding information delay problems, and improving the safety and stability of intelligent driving vehicles.

[0060] Optionally, the electronic control unit can also be deployed in a terminal or server. The terminal can be a physical device such as a smartphone, computer (e.g., desktop computer, tablet computer, laptop computer), digital assistant, intelligent voice interaction device (e.g., smart speaker), smart wearable device, or in-vehicle terminal, or software running on the physical device, such as a computer program. The operating system corresponding to the terminal can be Android, iOS (a mobile operating system developed by Apple), Linux, Microsoft Windows, etc.

[0061] The server involved in this application embodiment can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, and big data and artificial intelligence platforms.

[0062] Please see Figure 2 The diagram shown is a flowchart illustrating an electronic power steering control method provided in an embodiment of this application. This method can be applied to... Figure 1 The electronic control unit in the system. It should be noted that this specification provides the operational steps of the methods described in the embodiments or flowcharts, but based on conventional or non-inventive labor, more or fewer operational steps may be included. The order of steps listed in the embodiments is merely one possible execution order among many and does not represent the only execution order. In actual system or product execution, the methods shown in the embodiments or drawings can be executed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment). Specifically, as shown... Figure 2 As shown, the method may include:

[0063] S201: Obtain steering requests and driving data during vehicle operation; the steering request carries target steering information.

[0064] In one feasible implementation, see Figure 3 , Figure 3This is a schematic diagram illustrating the operation of an assist motor according to an embodiment of this application. When the driver sends an ignition command, the battery provides power and the EPS system initializes. If the EPS system self-test passes, it begins normal operation. When the driver turns the steering wheel, if the data acquisition module collects data on vehicle speed, road curvature, steering wheel angle, and steering wheel angular velocity, the collected and received data is stored and processed accordingly. When the EPS electronic control unit is working normally and receives a steering request from the control device, carrying a target steering angle or target steering torque, the target control current is obtained by processing the acquired data. Based on this target control current, the assist motor is driven to achieve steering. The control device can be another control unit in the vehicle's local area network. The data acquisition module includes a torque angle sensor, a speed acquisition module, and a road data acquisition module. The torque angle sensor is used to collect the steering wheel angle and steering wheel angular velocity, and to obtain the steering request. The speed acquisition module is used to collect the vehicle's current speed. The road data acquisition module is used to obtain the road curvature. Subsequently, by inputting the aforementioned vehicle speed, road curvature, steering wheel angle and angular velocity, target steering angle, or target steering torque into the control current prediction module for calculation, the target control current is obtained. Based on this target control current, the power steering motor is driven to work, thereby achieving power steering. The parameters related to the control current of the power steering motor required in this application are not limited to the aforementioned vehicle speed and road curvature; parameters can also be added or reduced according to the vehicle model or actual needs.

[0065] In one feasible implementation, this application can be applied to assisted driving scenarios, see [reference]. Figure 4 and Figure 5 , Figure 4 This is a simplified flowchart of a method for determining the control current under different driving modes, provided in an embodiment of this application. Figure 5This is a road scene diagram provided in an embodiment of this application. Step S201 may include: obtaining the distance value between the vehicle and obstacles on both sides of the vehicle; if the distance value between the vehicle and obstacles on both sides of the vehicle is less than or equal to a first preset distance value, then in response to a first control command from the user on the control interface, generating a steering request based on the distance value during vehicle driving, and obtaining driving data during vehicle driving; the first control command indicates that the vehicle is allowed to enter the assisted driving mode. For example, when the distance between the vehicle and obstacles on the left and right sides or the vehicle is less than a first preset threshold, the vehicle system will give a warning through voice prompts and remind the user to make a choice in the form of a question reminder, such as "Enter assisted driving mode?". If the user responds "yes", or the user clicks the "confirm" button on the control interface, then after obtaining the confirmation command (i.e., the first control command), the electronic steering system will enter the assisted driving mode to help the driver avoid obstacles in an emergency and steer. Specifically, when entering the assisted driving mode, the electronic control unit can calculate the target steering angle or target steering torque based on the distance value, and generate a corresponding steering request based on the target steering angle or target steering torque, so as to predict the control current of the subsequent power assist motor based on the steering request.

[0066] The electronic control unit responds to the user's second control command on the control interface, acquiring steering requests and driving data during vehicle operation; the second control command indicates that the vehicle maintains manual driving mode. Continuing the above example, if the user responds "no," or the user clicks the "cancel" button on the control interface, the driver continues to operate the steering wheel, and the vehicle does not enter assisted driving mode.

[0067] In another feasible implementation, to further ensure driver safety and improve user experience, step S201 may further include: obtaining the distance value between the vehicle and obstacles on both sides of the vehicle; if the distance value between the vehicle and obstacles on both sides of the vehicle is less than or equal to a second preset distance value, generating a steering request based on the distance value during vehicle operation, and obtaining driving data during vehicle operation; the second preset distance value is less than a first preset distance value; if the distance value between the vehicle and obstacles on both sides of the vehicle is greater than the second preset distance value and less than or equal to the first preset distance value, generating a steering request based on the distance value during vehicle operation, and obtaining driving data during vehicle operation.

[0068] In this embodiment, a two-level threshold is set to determine whether to enter the assisted driving mode. When the distance value is less than or equal to the second preset distance value, it indicates that the vehicle is highly likely to collide with vehicles on both sides, and the driver may not be able to react in time. Therefore, in this case, a warning can be given through voice prompts, and the electronic power steering system will directly enter the assisted driving mode to help the driver avoid obstacles and steer. When the distance value is greater than the second preset distance value but less than the first preset distance value, the vehicle system will give a warning through voice prompts and remind the user to make a choice in the form of a question. The system will determine whether to enter the assisted driving mode based on the user's choice. This allows the user to have the right to choose in relatively safe situations and ensures safety in more dangerous situations.

[0069] The aforementioned distance value can be based on radar detection. Optionally, the radar can be installed on the connecting axle connecting the two wheels, with one radar on each side of the vehicle.

[0070] In this assisted driving mode, the target control current of the power steering motor can also be determined using the control current prediction model described below. However, to ensure the accuracy of the estimated control current, the training set needs to include the parameters and corresponding control currents for the various assisted driving modes mentioned above when training the control current prediction model. The obtained control current also needs to be used to calculate the corresponding steering torque based on the following formula, thereby achieving steering through the drive gear mechanism:

[0071] Steering torque = Control current × (Torque constant + Compensation value)

[0072] Among them, the control current represents the current of the q-axis of the assist motor; the compensation value represents the current of the d-axis of the assist motor, which is obtained by looking up the table, and the torque constant can be taken as 0.052.

[0073] S203: Using a control current prediction model, the control current of the power assist motor is predicted based on the target steering information and driving data to obtain the target control current of the power assist motor; the control current prediction model is obtained by training a planar neural network based on historical target steering information, historical driving data and corresponding historical control current.

[0074] In one feasible implementation, see Figure 6-7 , Figure 6 This is a schematic diagram of a process for training a control current prediction model provided in an embodiment of this application; Figure 7 This is a schematic diagram of a width learning model provided in an embodiment of this application. The control current prediction model includes an input layer, an intermediate layer, and an output layer connected in sequence; the intermediate layer includes a feature layer and an enhancement layer; the input layer consists of input data for the control current prediction model; the output layer is connected to the feature layer and the enhancement layer respectively; the method for controlling the current prediction model in step S203 includes:

[0075] S601: Obtain the training set; the training set includes the historical steering parameters of the first number of historical moments and the historical control current of each historical moment; the historical steering parameters include historical target steering information and historical driving data.

[0076] In a feasible implementation, step S601 may involve acquiring the historical steering parameters and historical control currents for each historical moment from multiple historical moments to form a training set. The historical steering parameters can be used as inputs to the control current prediction model to be trained subsequently, and the historical control currents can be used as labels corresponding to each input in the control current prediction model. The model is then trained based on the following steps to obtain the model parameters of the control current prediction model, thus completing the training of the model and obtaining the control current prediction model.

[0077] In another feasible implementation, the electronic power steering control system can also provide personalized power steering based on the driver's steering style. Step S601 may include: acquiring historical steering information and the driver's historical steering force at multiple historical moments; inputting the historical steering information and the driver's historical steering force at multiple historical moments into a steering style prediction model to obtain the driver's steering style; the steering style prediction model is trained on a planar neural network based on the steering information, the driver's steering force, and the corresponding driving style; acquiring historical steering parameters and historical control current at a first number of historical moments; the historical steering parameters include historical target steering information and historical driving data; adjusting the historical control current based on the driver's steering style for each historical moment to generate a target historical control current; and constructing a training set based on the historical steering parameters and the corresponding target historical control current at each of the first number of historical moments.

[0078] In this embodiment, the driver's steering force can be sensed through the steering wheel's tactile feedback. A torque angle sensor collects steering information, including steering wheel rotation speed and angle. This steering force and information are input into a trained steering style model to determine the driver's steering style. Optionally, the driver's steering style can be categorized into three states: "rapid," "normal," and "slow." Based on the driver's steering style, the electronic steering system provides a corresponding personalized steering assist method. The training process for this steering style prediction model is similar to that of the control current prediction model. The main difference lies in the training data. The training set for the steering style prediction model includes the driver's steering force, steering wheel information, and corresponding driving style labels.

[0079] Once the driver's steering style is obtained, the historical control current in the above steps is adjusted according to the preset adjustment rules. The general adjustment rules are: when the steering style is normal, no adjustment of the historical control current is required; when the steering style is fast, the historical control current is reduced; and when the steering style is slow, the historical control current is increased, thereby improving the user experience while ensuring the driver's safety.

[0080] To improve the speed of data processing in the model, the data in the training set can be normalized first. Specifically, based on the historical control current, the training set can be divided into multiple sub-sample sets, each sub-sample set corresponding to a historical control current (e.g., all 3 amps). Each sub-sample set includes steering parameters at multiple times. Each parameter in each sub-sample set is normalized to obtain the normalized target training set. Subsequent steps will also be based on the target training set.

[0081] For example, a subsample set X = [X1, X2, ..., Xn], where n is the number of data types, such as 5 types as mentioned above, then n = 5; for each parameter, such as X1 = [x1, x2, ..., xn], ... m ], where m is the number of parameters; the formula for normalizing the data in dataset X1 is as follows:

[0082]

[0083] Where, x i 'for x i The result after normalization, x i For the original data, X min Let X be the minimum value in dataset X1. max The maximum value in dataset X1.

[0084] The normalized subset X'1 = [x'1, x'2, ..., x' m The normalization method for the remaining subsample set is the same, and will not be repeated here.

[0085] S603: Using the first nonlinear activation function, feature mapping is performed on all historical turning parameters in the training set to generate the feature matrix of the feature layer.

[0086] See Figure 7 As mentioned above, the control current prediction model is a wide-learning model, which is a planar neural network evolved from function chain neural networks. It typically includes an input layer, an intermediate layer, and an output layer connected sequentially. The intermediate layer includes a feature layer and an enhancement layer. The input layer consists of input data, and the number of input nodes equals the dimension of the input data. The feature layer includes multiple feature nodes, and the enhancement layer includes multiple enhancement nodes, such as... Figure 7 As shown, the feature layer includes features Z1, Z2, ..., Zn, where n is an integer greater than or equal to 3; the enhancement feature layer includes enhancement features H1, H2, ..., Hm, where m is an integer greater than or equal to 3, and each feature corresponds to a node. The output layer is determined based on the feature nodes of the feature layer and the enhancement nodes of the enhancement layer; typically, the number of input nodes does not exceed 10, and the number of feature nodes and enhancement nodes can be set to 10-20, with the number of enhancement nodes being greater than the number of feature nodes; see below for details. Feature nodes are extracted using the φ() feature extraction function, and then enhancement nodes are generated from the feature nodes using the ε() function. Finally, the two types of nodes together form the intermediate layer.

[0087] In one feasible implementation, see Figure 8 , Figure 8 This is a schematic flowchart of another training control current prediction model provided in an embodiment of this application. Step S603 may include:

[0088] S6031: Obtain the first random weight coefficient and the first random bias; the first random weight coefficient follows a Gaussian distribution.

[0089] In this embodiment, the first random weight coefficient is denoted as We, and the first random bias is denoted as βe. We is a random weight matrix that follows a Gaussian distribution.

[0090] S6033: Determine the first parameter matrix based on the first random weight coefficient, the first random bias, and all historical turning parameters in the training set.

[0091] In this embodiment, the first parameter matrix is ​​the sum of the product of the input data (i.e., the historical target turning parameters) and the corresponding first random weight system and the first random bias, denoted as the first parameter matrix = XWe + βe.

[0092] S6035: Using the first nonlinear activation function, perform feature mapping on the first parameter matrix to generate the feature matrix of the feature layer.

[0093] In this embodiment, the first nonlinear activation function is denoted as φ(), which can be the nonlinear activation function ReLU; the feature corresponding to the feature node in the feature layer is Zi = φ(XWei + βei), i = 1, 2, ..., n, where Zi represents the feature of the i-th feature node; the features of these feature nodes are concatenated to form the feature matrix Z = [Z1, Z2, ..., Zn].

[0094] S605: Using the second nonlinear activation function, feature extraction is performed on the feature matrix of the feature layer to generate the feature matrix of the enhancement layer.

[0095] In one feasible implementation, see Figure 8 Step S605 may include:

[0096] S6051: Obtain the second random weight coefficient and the second random bias; the second random weight coefficient follows a Gaussian distribution.

[0097] In this embodiment, the second random weight coefficient is denoted as Wh, and the second random bias is denoted as βh. Wh is a random weight matrix with a Gaussian distribution.

[0098] S6053: Determine the second parameter matrix based on the second random weight coefficient, the second random bias, and the feature matrix of the feature layer.

[0099] In this embodiment, the second parameter matrix is ​​the sum of the product of the feature matrix Z of the feature layer and the corresponding second random weight system and the second random bias, denoted as the second parameter matrix = ZWh + βh.

[0100] S6055: Using the second nonlinear activation function, feature extraction is performed on the second parameter matrix to obtain the feature matrix of the enhancement layer.

[0101] In this embodiment, the second nonlinear activation function is denoted as ε(), which can be the nonlinear activation function ReLU; the feature Hj corresponding to the enhancement node in the enhancement layer is ε(XWhj+βhj), j=1,2,...,n, and Hj represents the feature of the j-th enhancement node; the features of these enhancement nodes are concatenated to form the feature matrix H=[H1,H2,...,Hm].

[0102] S607: Based on the feature matrix of the feature layer, the feature matrix of the enhancement layer, and the corresponding historical control current in the training set, determine the target weight parameters of the control current prediction model to obtain the control current prediction model.

[0103] In this embodiment, the feature matrix of the feature layer and the feature matrix of the enhancement layer can be merged into an intermediate layer A = [Z|H], and the vertical line indicates that the feature matrix of the feature layer and the feature matrix of the enhancement layer are merged into one row; the output of the control current prediction model can be denoted as: Y T =AW; where W represents the weight coefficient of the feature node connecting the output layer to the feature layer and the enhancement node of the enhancement layer, and Y is the output of the model, which is the known historical control current. Therefore, the model's weight parameters W = A. -1 Y. Specifically, ridge regression and pseudo-inverse operations can be used to solve for the weight matrix.

[0104] During training, the objective function of the model can be expressed as follows:

[0105]

[0106] Where AW is the predicted value output by the control current prediction model; Y is the actual value of the control current prediction model (i.e., the historical control current mentioned above); λ is used to minimize training error; λ is the regularization coefficient used to prevent the model from overfitting.

[0107] By solving the above objective function, the objective weight parameters W = (λI + AA) of the model can be obtained. T ) - 1 A T Y;

[0108] Where I is the identity matrix; A T W is the transpose of A, and W is obtained by transposing A. + The ridge regression approximation is obtained according to the following formula:

[0109] A + It is the pseudo-inverse of A.

[0110] The above process is a one-step process. If the data and model structure are fixed, the optimal parameters, i.e. the optimal prediction result, can be found directly.

[0111] In one feasible implementation, after step S607, the method further includes: acquiring a test set, which includes a second number of historical steering parameters at historical moments and historical control currents at each historical moment; inputting the second number of historical steering parameters at historical moments into a control current prediction model to obtain the predicted control current at each historical moment; determining the error value between the predicted control current and the corresponding historical control current for each historical steering parameter in the test set; determining the model accuracy value based on the error value corresponding to each historical steering parameter in the test set; if the model accuracy value is greater than a preset threshold, increasing the number of enhancement nodes in the enhancement layer of the control current prediction model and retraining the control current prediction model until the model accuracy value of the control current prediction model is less than or equal to the preset threshold. This further improves the accuracy of the model.

[0112] In this embodiment, the historical control current corresponding to the model output can be used. (i.e., historical control current) and predicted control current y i The difference is used as the error value between the predicted control current and the corresponding historical control current for the corresponding historical steering parameter, that is, the error value between each predicted control current and the corresponding historical control current.

[0113] The model accuracy value RMSE can be expressed by the following formula:

[0114]

[0115] Where k represents the number of samples in the test set, which is the same as the number of historical control currents in the test set. Generally speaking, the smaller the RMSE, the closer the model's predicted value is to the true value, and the better the model's prediction performance.

[0116] In this embodiment, the dataset can be obtained in step S601, and a portion of it can be used as the training set and the other portion as the subsequent test set according to a certain ratio.

[0117] When the model accuracy exceeds a preset threshold, indicating poor prediction performance and insufficient model fitting ability, the number of nodes can be increased to reduce the loss function. Specifically, a column 'a' can be added to matrix A to represent the newly added augmentation nodes, resulting in [A|a]. A new weight matrix, [A|a], then needs to be calculated. -1 Therefore, the problem is transformed into the generalized inverse problem of a block matrix, resulting in [A|a]. -1 Then the updated weight matrix W new for [A|a] -1 Y, the specific solution form is as follows, as can be seen, W n+1 The previous weight matrix W was used in the process. nThis effectively reduces the computational cost of updating weights.

[0118]

[0119] In the formula, c = aA n d

[0120] Of course, when it is necessary to update the above control current prediction model, such as when the types of input parameters of the model increase, the model structure may also need to be updated, such as increasing the number of feature nodes. In this case, the previous calculation results and the newly added data can be used to obtain the updated weight data with only a small amount of calculation. When updating the pseudo-inverse of the output layer, only the newly added enhancement nodes need to be calculated.

[0121] S205: Power steering control of the steering wheel based on the target control current of the power steering motor.

[0122] As can be seen from the above technical solutions of the embodiments of this application, the embodiments of this application collect information such as vehicle speed, steering wheel angle and angular velocity, road curvature, and radar signals, input them into the width learning model for training, and then output the power assist motor compensation amount, that is, the control current of the power assist motor, to realize the compensation of EPS steering assist characteristics and friction characteristics, so as to control the power assist motor to work, and solve the problems of system accuracy loss and control strategy complexity caused by simplifying the EPS system into a linear system for control in the traditional way.

[0123] The method provided in this application avoids complex and difficult physical modeling. By deeply mining the vehicle's environmental and state data, it establishes an efficient control current prediction model, which features fast response speed, good steering effect, and adaptability to different environments. It directly obtains the optimal weight coefficients and biases of the model using matrix pseudo-inversion. When the network accuracy does not meet expectations, the network width can be increased by adding enhancement nodes to improve network performance. The designed EPS system can effectively assist electric power steering in nonlinear steering across the entire speed range. It not only solves the problems of complex control algorithms and poor stability in traditional electronic power steering systems, but also ensures the safety and stability of the vehicle during operation, playing an important role in reducing traffic accidents and improving steering conditions.

[0124] In addition, the electronic steering system with driver assistance function proposed in this invention has personalized settings, intelligent sensing steering driving style, and provides power assist mode; and uses radar to sense the distance to the vehicle on the left and right sides to actively provide power assist, thereby improving safety and comfort.

[0125] Corresponding to the electronic power steering control methods provided in the above embodiments, this application also provides an electronic power steering control device. Since the electronic power steering control device provided in this application corresponds to the electronic power steering control methods provided in the above embodiments, the implementation methods of the aforementioned electronic power steering control methods are also applicable to the electronic power steering control device provided in this embodiment, and will not be described in detail in this embodiment.

[0126] Please see Figure 9 The diagram shown is a structural schematic of an electronic power steering control device provided in an embodiment of this application. This device has the function of implementing the electronic power steering control method described in the above-described method embodiments. This function can be implemented by hardware or by hardware executing corresponding software. Figure 9 As shown, the device may include:

[0127] The acquisition module 901 is used to acquire steering requests and driving data during vehicle operation; the steering request carries target steering information.

[0128] The prediction module 903 is used to predict the control current of the power assist motor based on the target steering information and driving data using a control current prediction model, so as to obtain the target control current of the power assist motor; the control current prediction model is obtained by training a planar neural network based on historical target steering information, historical driving data and corresponding historical control current.

[0129] The control module 905 is used to perform power steering control on the steering wheel based on the target control current of the power steering motor.

[0130] In one feasible implementation, the control current prediction model includes an input layer, an intermediate layer, and an output layer connected in sequence; the intermediate layer includes a feature layer and an enhancement layer; the input layer consists of input data for the control current prediction model; the output layer is connected to both the feature layer and the enhancement layer; the device further includes:

[0131] The training dataset acquisition module is used to acquire the training set; the training set includes the historical steering parameters of a first number of historical moments and the historical control current of each historical moment; the historical steering parameters include historical target steering information and historical driving data;

[0132] The first feature extraction module is used to perform feature mapping processing on all historical turning parameters in the training set using the first nonlinear activation function to generate the feature matrix of the feature layer.

[0133] The second feature extraction module is used to perform feature extraction processing on the feature matrix of the feature layer using the second nonlinear activation function to generate the feature matrix of the enhancement layer.

[0134] The first determining module is used to determine the target weight parameters of the control current prediction model based on the feature matrix of the feature layer, the feature matrix of the enhancement layer, and the corresponding historical control current in the training set, so as to obtain the control current prediction model.

[0135] In one feasible implementation, the first feature extraction module is used to obtain a first random weight coefficient and a first random bias; the first random weight coefficient is Gaussian distributed; a first parameter matrix is ​​determined based on the first random weight coefficient, the first random bias and all historical turning parameters in the training set; and a first nonlinear activation function is used to perform feature mapping processing on the first parameter matrix to generate the feature matrix of the feature layer.

[0136] In one feasible implementation, the second feature extraction module is used to obtain the second random weight coefficient and the second random bias; the second random weight coefficient is Gaussian distributed; the second parameter matrix is ​​determined based on the second random weight coefficient, the second random bias and the feature matrix of the feature layer; and the second parameter matrix is ​​processed by the second nonlinear activation function to obtain the feature matrix of the enhancement layer.

[0137] In one feasible implementation, the device further includes:

[0138] The test set acquisition module is used to acquire a test set, which includes the historical steering parameters of a second number of historical moments and the historical control current of each historical moment.

[0139] The input module is used to input the historical turning parameters of the second number of historical moments into the control current prediction model to obtain the predicted control current for each historical moment.

[0140] The second determining module is used to determine the error value between the predicted control current and the corresponding historical control current for each historical steering parameter in the test set.

[0141] The third determination module is used to determine the model accuracy value based on the error values ​​corresponding to each historical steering parameter in the test set;

[0142] The judgment module is used to increase the number of enhancement nodes in the enhancement layer of the control current prediction model and retrain the control current prediction model if the model accuracy value is greater than a preset threshold, until the model accuracy value of the control current prediction model is less than or equal to the preset threshold.

[0143] In one feasible implementation, the training set acquisition module is used to acquire historical steering wheel steering information and the driver's historical steering force at multiple historical moments.

[0144] The historical steering wheel information and the driver's historical steering force at multiple historical moments are input into the steering driving style prediction model to obtain the driver's steering driving style. The steering driving style prediction model is obtained by training a planar neural network based on the steering wheel information, the driver's steering force, and the corresponding driving style.

[0145] Acquire the historical steering parameters and historical control current for each historical moment, including the historical target steering information and historical driving data.

[0146] For each historical moment, the historical control current is adjusted based on the driver's steering style to generate the target historical control current.

[0147] The training set is constructed based on the historical turning parameters and the corresponding target historical control current of each historical moment in the first set of historical moments.

[0148] In one feasible implementation, the acquisition module is used to acquire the distance values ​​between the vehicle and obstacles on both sides of the vehicle;

[0149] If the distance between the vehicle and obstacles on both sides of the vehicle is less than or equal to a first preset distance value, then in response to the user's first control command on the control interface, a steering request is generated based on the distance value, and driving data is acquired during the vehicle's driving process; or, in response to the user's second control command on the control interface, the steering request and driving data are acquired during the vehicle's driving process; the first control command indicates that the vehicle is allowed to enter the assisted driving mode; the second control command indicates that the vehicle maintains the manual driving mode.

[0150] It should be noted that the apparatus provided in the above embodiments is only illustrated by the division of the above functional modules when implementing its functions. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0151] This application provides an electronic device including a processor and a memory. The memory stores at least one instruction or at least one program segment, which is loaded and executed by the processor to implement any of the electronic power steering control methods provided in the above method embodiments.

[0152] Memory can be used to store software programs and modules. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory. Memory can primarily include a program storage area and a data storage area. The program storage area can store the operating system, application programs required for the functions, etc.; the data storage area can store data created based on the use of the device, etc. Furthermore, memory can include high-speed random access memory, and can also include non-volatile memory, such as at least one disk storage device, flash memory device, or other volatile solid-state storage device. Accordingly, memory can also include a memory controller to provide the processor with access to the memory.

[0153] Embodiments of this application also provide a computer-readable storage medium, which can be disposed in an electronic device to store at least one instruction or at least one program related to implementing an electronic power steering control method. The at least one instruction or the at least one program is loaded and executed by the processor to implement any of the electronic power steering control methods provided in the above-described method embodiments.

[0154] Embodiments of this application also provide a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of an electronic device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the electronic device to perform any of the electronic power steering control methods provided in the above-described method embodiments.

[0155] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0156] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, specific embodiments have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0157] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0158] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0159] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. An electronic power steering control method, characterized in that, The method includes: Acquire steering requests and driving data during vehicle operation; the steering requests carry target steering information. Using a control current prediction model, the control current of the power assist motor is predicted based on the target steering information and driving data to obtain the target control current of the power assist motor; the control current prediction model is obtained by training a planar neural network based on historical target steering information, historical driving data and corresponding historical control current. Based on the target control current of the power steering motor, power steering control is performed on the steering wheel; The control current prediction model includes an input layer, an intermediate layer, and an output layer connected in sequence; the intermediate layer includes a feature layer and an enhancement layer; the input layer consists of input data to the control current prediction model; the output layer is connected to the feature layer and the enhancement layer respectively; the method for training the control current prediction model includes: Obtain a training set; the training set includes a first number of historical steering parameters at historical moments and historical control currents at each historical moment; the historical steering parameters include historical target steering information and historical driving data; The first nonlinear activation function is used to perform feature mapping on all historical turning parameters in the training set to generate the feature matrix of the feature layer; The feature matrix of the feature layer is processed by using a second nonlinear activation function to generate the feature matrix of the enhancement layer. The target weight parameters of the control current prediction model are determined based on the feature matrix of the feature layer, the feature matrix of the enhancement layer, and the historical control currents corresponding to the training set, thus obtaining the control current prediction model.

2. The electronic power steering control method according to claim 1, characterized in that, The step of using a first nonlinear activation function to perform feature mapping processing on all historical turning parameters in the training set to generate the feature matrix of the feature layer includes: Obtain the first random weight coefficient and the first random bias; the first random weight coefficient follows a Gaussian distribution. The first parameter matrix is ​​determined based on the first random weight coefficient, the first random bias, and all historical turning parameters in the training set; The first nonlinear activation function is used to perform feature mapping on the first parameter matrix to generate the feature matrix of the feature layer.

3. The electronic power steering control method according to claim 2, characterized in that, The step of using a second nonlinear activation function to perform feature extraction processing on the feature matrix of the feature layer to generate the feature matrix of the enhancement layer includes: Obtain the second random weight coefficient and the second random bias; the second random weight coefficient follows a Gaussian distribution. The second parameter matrix is ​​determined based on the second random weight coefficient, the second random bias, and the feature matrix of the feature layer; The second nonlinear activation function is used to perform feature extraction on the second parameter matrix to obtain the feature matrix of the enhancement layer.

4. The electronic power steering control method according to claim 2, characterized in that, After determining the target weight parameters of the control current prediction model based on the feature matrix of the feature layer, the feature matrix of the enhancement layer, and the historical control currents corresponding to the training set, and obtaining the control current prediction model, the method further includes: Obtain a test set, which includes a second number of historical steering parameters at historical moments and historical control currents at each historical moment; Input the historical turning parameters of the second number of historical moments into the control current prediction model to obtain the predicted control current for each historical moment; Determine the error value between the predicted control current and the corresponding historical control current for each historical steering parameter in the test set; The model accuracy value is determined based on the error values ​​corresponding to each historical steering parameter in the test set. If the model accuracy value is greater than a preset threshold, the number of enhancement nodes in the enhancement layer of the control current prediction model is increased, and the control current prediction model is retrained until the model accuracy value of the control current prediction model is less than or equal to the preset threshold.

5. The electronic power steering control method according to claim 2, characterized in that, The acquisition of the training set includes: Acquire historical steering wheel information and driver's historical steering force at multiple historical moments; The historical steering wheel information and the driver's historical steering force at multiple historical moments are input into the steering driving style prediction model to obtain the driver's steering driving style; the steering driving style prediction model is obtained by training a planar neural network based on the steering wheel information, the driver's steering force, and the corresponding driving style. Acquire a first number of historical steering parameters and historical control currents for each historical moment; the historical steering parameters include historical target steering information and historical driving data; For each historical control current at each historical moment, the historical control current is adjusted based on the driver's steering and driving style to generate a target historical control current; The training set is constructed based on the historical steering parameters of each historical moment in the first number of historical moments and the corresponding target historical control current.

6. The electronic power steering control method according to any one of claims 1-5, characterized in that, The acquisition of steering requests and driving data during vehicle operation includes: Obtain the distance values ​​between the vehicle and the obstacles on both sides of the vehicle; If the distance between the vehicle and the obstacles on both sides of the vehicle is less than or equal to a first preset distance value, then in response to a first control command from the user on the control interface, a steering request is generated based on the distance value, and driving data is acquired during the vehicle's driving process; or, in response to a second control command from the user on the control interface, the steering request and driving data are acquired during the vehicle's driving process; the first control command indicates that the vehicle is allowed to enter the assisted driving mode; the second control command indicates that the vehicle maintains the manual driving mode.

7. An electronic power steering control device, characterized in that, The device includes: The acquisition module is used to acquire steering requests and driving data during vehicle operation; the steering request carries target steering information. The prediction module is used to predict the control current of the power assist motor based on the target steering information and driving data using a control current prediction model, thereby obtaining the target control current of the power assist motor. The control current prediction model is obtained by training a planar neural network based on historical target steering information, historical driving data, and corresponding historical control currents. The control current prediction model includes an input layer, an intermediate layer, and an output layer connected in sequence. The intermediate layer includes a feature layer and an enhancement layer. The input layer consists of the input data input to the control current prediction model. The output layer is connected to both the feature layer and the enhancement layer. The control module is used to perform power steering control on the steering wheel based on the target control current of the power steering motor; The training dataset acquisition module is used to acquire the training set; the training set includes the historical steering parameters of a first number of historical moments and the historical control current of each historical moment; the historical steering parameters include historical target steering information and historical driving data; The first feature extraction module is used to perform feature mapping processing on all historical turning parameters in the training set using the first nonlinear activation function to generate the feature matrix of the feature layer. The second feature extraction module is used to perform feature extraction processing on the feature matrix of the feature layer using the second nonlinear activation function to generate the feature matrix of the enhancement layer. The first determining module is used to determine the target weight parameters of the control current prediction model based on the feature matrix of the feature layer, the feature matrix of the enhancement layer, and the corresponding historical control current in the training set, so as to obtain the control current prediction model.

8. An electronic device, characterized in that, It includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, the at least one instruction or the at least one program being loaded and executed by the processor to implement the electronic power steering control method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction or at least one program, which is loaded and executed by a processor to implement the electronic power steering control method as described in any one of claims 1 to 6.

Images