Vehicle steering control method and vehicle steering control device

Abstract

The invention relates to a vehicle steering control method and a vehicle steering control device. The method comprises the steps that under the condition that a steering instruction for a target vehicle is received, vehicle environment information, vehicle state information and driver state information of the target vehicle are obtained; inputting the vehicle environment information, the vehicle state information and the driver state information into a pre-trained steering intention prediction model to obtain target turning angle information and turning angle speed information of the target vehicle, and sending the target turning angle information and the turning angle speed information to a steering execution system; and the steering execution system is used for executing a steering executor corresponding to the target vehicle according to the target steering angle information and the steering angle velocity information so as to control steering of the target vehicle. By adopting the method, the turning angle and the turning angle velocity can be analyzed and predicted in combination with information such as the vehicle environment, the vehicle state and the driver state, vehicle steering is accurately controlled through the steering execution system based on the turning angle and the turning angle velocity, and the accuracy of vehicle steering control is improved.

Description

Technical Field

[0001] The present application relates to the field of automobile control technology, and in particular to a vehicle steering control method, device, computer equipment, computer-readable storage medium and computer program product. Background Art

[0002] Vehicle steering control refers to adjusting the angle of the vehicle's front wheels through the driver's operation or the instructions of the automatic driving system, thereby changing the vehicle's driving direction. The steering control system directly affects the vehicle's handling performance, driving safety and comfort.

[0003] In traditional technology, vehicle steering control mainly uses data fitting to determine control parameters to achieve vehicle steering control. However, traditional technology is limited by factors such as sensor accuracy, data drift, and nonlinear friction interference. When steering control is performed based on data fitting, there is a problem of low control accuracy, which is difficult to meet the requirements of complex traffic conditions and is not conducive to improving the accuracy of vehicle steering control. Summary of the invention

[0004] Based on this, it is necessary to provide a vehicle steering control method, device, computer equipment, computer-readable storage medium and computer program product that can improve the accuracy of vehicle steering control in order to solve the above technical problems.

[0005] In a first aspect, the present application provides a vehicle steering control method, comprising:

[0006] Upon receiving a steering instruction for a target vehicle, obtaining vehicle environment information, vehicle state information, and driver state information of the target vehicle;

[0007] The vehicle environment information, the vehicle state information and the driver state information are input into a pre-trained steering intention prediction model to obtain the target turning angle information and turning angle velocity information of the target vehicle, and the target turning angle information and the turning angle velocity information are sent to a steering execution system; the steering execution system is used to activate the steering actuator corresponding to the target vehicle according to the target turning angle information and the turning angle velocity information to control the steering of the target vehicle.

[0008] In one embodiment, the vehicle environment information, the vehicle state information, and the driver state information are input into a pre-trained steering intention prediction model to obtain the target turning angle information and the turning angle velocity information of the target vehicle, including:

[0009] Inputting the vehicle environment information, the vehicle state information and the driver state information into a pre-trained steering intention prediction model to obtain predicted turning angle information and turning angle velocity information of the target vehicle;

[0010] The steering wheel angle information of the target vehicle is acquired from the vehicle state information, and the target angle information of the target vehicle is obtained by using the predicted angle information and the steering wheel angle information.

[0011] In one embodiment, obtaining target turning angle information of the target vehicle using the predicted turning angle information and the steering wheel angle information includes:

[0012] Determining weight information of the steering wheel angle information and weight information of the predicted angle information by using the predicted angle information and the steering wheel angle information;

[0013] According to the weight information of the steering wheel angle information and the weight information of the predicted angle information, the steering wheel angle information and the predicted angle information are weighted and summed to obtain the target angle information of the target vehicle.

[0014] In one embodiment, the turning intention prediction model includes a first branch and a second branch;

[0015] The vehicle environment information, the vehicle state information and the driver state information are input into a pre-trained steering intention prediction model to obtain the predicted turning angle information and turning angle velocity information of the target vehicle, including:

[0016] Inputting the vehicle environment information, the vehicle state information and the driver state information into the steering intention prediction model, and obtaining the predicted turning angle information and turning angle velocity information through the first branch;

[0017] The method further comprises:

[0018] Obtaining target current information of the road sensor motor of the target vehicle through the second branch;

[0019] The road sense motor target current information is sent to a road sense execution system; the road sense execution system is used to control the road sense motor corresponding to the target vehicle according to the road sense motor target current information, so as to feedback road conditions to the driver of the target vehicle.

[0020] In one embodiment, before obtaining the vehicle environment information, vehicle state information and driver state information of the target vehicle, the method further includes:

[0021] Obtaining sample vehicle environment information, sample vehicle state information, and sample driver state information of a sample vehicle;

[0022] Determine, based on the sample vehicle environment information, the sample vehicle state information and the sample driver state information, vehicle environment coding information corresponding to the sample vehicle environment information, vehicle state coding information corresponding to the sample vehicle state information and driver state coding information corresponding to the sample driver state information;

[0023] Inputting the vehicle environment coding information, the vehicle state coding information and the driver state coding information into the steering intention prediction model to be trained to obtain the road sense motor current information of the sample vehicle;

[0024] The sample road sense motor current information of the sample vehicle is obtained, and the second branch in the steering intention prediction model to be trained is trained according to the road sense motor current information of the sample vehicle and the sample road sense motor current information.

[0025] In one embodiment, before obtaining the vehicle environment information, vehicle state information and driver state information of the target vehicle, the method further includes:

[0026] Obtaining sample vehicle environment information, sample vehicle state information, and sample driver state information of a sample vehicle;

[0027] Determine, based on the sample vehicle environment information, the sample vehicle state information and the sample driver state information, vehicle environment coding information corresponding to the sample vehicle environment information, vehicle state coding information corresponding to the sample vehicle state information and driver state coding information corresponding to the sample driver state information;

[0028] Inputting the vehicle environment coding information, the vehicle state coding information and the driver state coding information into the steering intention prediction model to be trained to obtain the predicted turning angle information and the predicted turning angle angular velocity information of the sample vehicle;

[0029] Obtain sample turning angle information and sample turning angle angular velocity information of the sample vehicle, and train a first branch in the steering intention prediction model to be trained based on the predicted turning angle information of the sample vehicle, the predicted turning angle angular velocity information of the sample vehicle, the sample turning angle information and the sample turning angle angular velocity information.

[0030] In one embodiment, after obtaining the sample turning angle information and the sample turning angle angular velocity information of the sample vehicle, the method further includes:

[0031] Acquire quality information of the sample vehicle and gender information and preference information of the driver corresponding to the sample vehicle;

[0032] The model structure and model parameters of the turning intention prediction model to be trained are determined according to the quality information, the gender information and the preference information.

[0033] In one embodiment, determining, based on the sample vehicle environment information, the sample vehicle state information, and the sample driver state information, vehicle environment coding information corresponding to the sample vehicle environment information, vehicle state coding information corresponding to the sample vehicle state information, and driver state coding information corresponding to the sample driver state information includes:

[0034] Determining a vehicle environment time series according to the sample vehicle environment information, determining a vehicle state time series according to the sample vehicle state information, and determining a driver state time series according to the sample driver state information;

[0035] According to preset sequence length information, the vehicle environment time series, the vehicle state time series and the driver state time series are divided respectively to obtain sequence blocks corresponding to the vehicle environment time series, sequence blocks corresponding to the vehicle state time series and sequence blocks corresponding to the driver state time series;

[0036] Convolution is performed on the sequence blocks corresponding to the vehicle environment time series, the sequence blocks corresponding to the vehicle state time series, and the sequence blocks corresponding to the driver state time series, and the convolution results are embedded into the vector corresponding to the sample vehicle environment information, the vector corresponding to the sample vehicle state information, and the vector corresponding to the sample driver state information to obtain the vehicle environment coding information, the vehicle state coding information, and the driver state coding information.

[0037] In one embodiment, the steering intention prediction model to be trained includes at least one temporal convolutional network, the temporal convolutional network includes a depthwise separable convolutional layer, a pointwise convolutional layer and a normalization layer, the vehicle environment encoding information, the vehicle state encoding information and the driver state encoding information are input into the steering intention prediction model to be trained, and the predicted turning angle information and the predicted turning angle angular velocity information of the sample vehicle are obtained, including:

[0038] Through the depthwise separable convolutional layer in the temporal convolutional network, the vehicle environment coding information, the vehicle state coding information and the driver state coding information are convolved according to the convolution kernel corresponding to the depthwise separable convolutional layer to obtain an output result of the depthwise separable convolutional layer;

[0039] By using the point-by-point convolution layer in the temporal convolutional network, the output result of the depthwise separable convolution layer is convolved according to the convolution kernel corresponding to the point-by-point convolution layer to obtain the output result of the point-by-point convolution layer;

[0040] Normalizing the output result of the point-by-point convolution layer through the normalization layer in the temporal convolution network to obtain the output result of the normalization layer;

[0041] The predicted turning angle information and predicted turning angle angular velocity information of the sample vehicle are determined according to the vehicle environment coding information, the vehicle state coding information, the driver state coding information and the output result of the normalization layer.

[0042] In one embodiment, the temporal convolutional network further includes a first convolutional layer and a second convolutional layer, the convolutional kernel corresponding to the first convolutional layer is different from the convolutional kernel corresponding to the second convolutional layer, and after obtaining the output result of the depthwise separable convolutional layer, the method further includes:

[0043] Performing convolution processing on the output result of the depthwise separable convolution layer through the convolution kernel corresponding to the first convolution layer in the temporal convolution network, so as to convolve the feature channels corresponding to the vehicle environment encoding information, the vehicle state encoding information and the driver state encoding information, respectively, to obtain the output result of the first convolution layer;

[0044] Performing convolution processing on the output result of the first convolution layer through the convolution kernel corresponding to the second convolution layer in the temporal convolution network, so as to convolve the vehicle environment encoding information, the vehicle state encoding information and the driver state encoding information, and obtain the output result of the second convolution layer;

[0045] The predicted turning angle information and predicted turning angle angular velocity information of the sample vehicle are determined according to the output result of the second convolutional layer, the vehicle environment encoding information, the vehicle state encoding information and the driver state encoding information.

[0046] In one embodiment, obtaining vehicle environment information, vehicle state information, and driver state information of the target vehicle includes:

[0047] Acquire road type information, visibility information, terrain information, vehicle quantity information, lane quantity information, and traffic complexity information of the environment in which the target vehicle is located through a camera of the target vehicle, and determine vehicle environment information of the target vehicle based on the road type information, visibility information, terrain information, vehicle quantity information, lane quantity information, and traffic complexity information;

[0048] Determining vehicle state information of the target vehicle according to vehicle speed information, acceleration information, yaw angular velocity information, wheel angle information, steering wheel angle information, temperature information and fault information of the target vehicle;

[0049] The driver status information of the target vehicle is determined according to the behavior information of the driver corresponding to the target vehicle.

[0050] In a second aspect, the present application also provides a vehicle steering control device, comprising:

[0051] An acquisition module, configured to acquire vehicle environment information, vehicle state information, and driver state information of the target vehicle upon receiving a steering instruction for the target vehicle;

[0052] A control module is used to input the vehicle environment information, the vehicle state information and the driver state information into a pre-trained steering intention prediction model to obtain the target turning angle information and the turning angle angular velocity information of the target vehicle, and send the target turning angle information and the turning angle angular velocity information to a steering execution system; the steering execution system is used to activate the steering actuator corresponding to the target vehicle according to the target turning angle information and the turning angle angular velocity information to control the steering of the target vehicle.

[0053] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the steps of the above method are implemented.

[0054] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the above method when executed by a processor.

[0055] In a fifth aspect, the present application further provides a computer program product, wherein the computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0056] The above-mentioned vehicle steering control method, device, computer equipment, computer-readable storage medium and computer program product obtain vehicle environment information, vehicle state information and driver state information of the target vehicle when receiving a steering instruction for the target vehicle; thereby timely responding to the steering instruction and obtaining information such as the vehicle environment, vehicle state and driver state, so as to subsequently control the vehicle steering based on the vehicle environment, vehicle state and driver state; inputting the vehicle environment information, vehicle state information and driver state information into a pre-trained steering intention prediction model to obtain the target turning angle information and turning angle angular velocity information of the target vehicle; thereby combining the vehicle environment, vehicle state and driver state information in real time based on the pre-trained steering intention prediction model. The system uses information such as vehicle status and driver status to accurately analyze the turning angle information and turning angle velocity information used to control the vehicle steering; the target turning angle information and turning angle velocity information are sent to the steering execution system; the steering execution system is used to control the steering actuator corresponding to the target vehicle according to the target turning angle information and turning angle velocity information to control the steering of the target vehicle, thereby utilizing the vehicle's steering execution system to accurately control the vehicle steering based on the target turning angle and turning angle velocity, and can utilize a pre-trained model combined with various information such as vehicle environment, vehicle status and driver status to analyze and predict the turning angle and turning angle velocity, and accurately control the vehicle steering based on the turning angle and turning angle velocity through the steering execution system, thereby improving the accuracy of vehicle steering control. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0058] Figure 1 A schematic flow chart of a vehicle steering control method in one embodiment;

[0059] Figure 2 A schematic diagram of inputting data for a model in one embodiment;

[0060] Figure 3 A schematic diagram of a convolution calculation in one embodiment;

[0061] Figure 4 is a schematic diagram of a temporal convolutional network in one embodiment;

[0062] Figure 5 is a schematic diagram of a vehicle steering control system in one embodiment;

[0063] Figure 6is a structural block diagram of a vehicle steering control device in one embodiment;

[0064] Figure 7 The figure is a diagram of the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0065] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0066] In an exemplary embodiment, Figure 1 As shown, a vehicle steering control method is provided, which is described by taking the method applied to a vehicle-mounted controller as an example, and includes the following steps S102 to S104. Among them:

[0067] Step S102, when a steering instruction for a target vehicle is received, vehicle environment information, vehicle state information and driver state information of the target vehicle are acquired.

[0068] The steering instruction may refer to a command / instruction for controlling the target vehicle to change the driving direction of the vehicle. In practical applications, the steering instruction may be triggered by the driver of the target vehicle by turning the steering wheel of the target vehicle.

[0069] Among them, vehicle environment information may refer to information that characterizes the environment in which the target vehicle is located. In actual applications, vehicle environment information may include but is not limited to information that characterizes the road conditions of the road on which the target vehicle is traveling, information that characterizes the visibility of the environment in which the target vehicle is located, information that characterizes the number of vehicles in the environment in which the target vehicle is located, information that characterizes the number of lanes in the environment in which the target vehicle is located, etc.

[0070] Among them, vehicle status information may refer to information characterizing the operating status of the target vehicle. In actual applications, vehicle status information may include but is not limited to information characterizing the vehicle speed, longitudinal acceleration, lateral acceleration, yaw angular velocity, wheel angle, steering wheel angle, fault conditions, and temperatures of various parts / components of the vehicle.

[0071] Among them, the driver status information may refer to information that characterizes the physical condition, psychological state, and behavioral performance of the driver of the target vehicle during driving. In practical applications, the driver status information may be obtained by analyzing the images / videos captured by the camera of the target vehicle against the driver.

[0072] As an example, when the driver of the target vehicle is driving the target vehicle, when the driver controls the steering of the target vehicle by turning the steering wheel or other means, the on-board controller can monitor the steering instructions issued by the driver to the target vehicle. When the on-board controller detects the steering instructions to the target vehicle, the on-board controller can obtain sensor data through sensors such as cameras of the target vehicle, and determine the vehicle environment information, vehicle status information and driver status information of the target vehicle by analyzing the sensor data.

[0073] Step S104, input the vehicle environment information, vehicle state information and driver state information into a pre-trained steering intention prediction model to obtain the target turning angle information and turning angle velocity information of the target vehicle, and send the target turning angle information and turning angle velocity information to the steering execution system; the steering execution system is used to activate the steering actuator corresponding to the target vehicle according to the target turning angle information and turning angle velocity information to control the steering of the target vehicle.

[0074] Among them, the pre-trained steering intention prediction model may refer to a model used to predict / analyze the control method of controlling the steering of a target vehicle based on vehicle environment information, vehicle state information and driver state information. In practical applications, the pre-trained steering intention prediction model can learn information such as motor parameters in the vehicle steering control process through pre-training, so as to output / determine information such as the initial turning angle and turning angular velocity based on vehicle environment information, vehicle state information and driver state information.

[0075] The target turning angle information may refer to information used to control the turning angle of the target vehicle (such as steering wheel angle / wheel angle, etc.).

[0076] The turning angle velocity information may refer to information on the rate at which the steering wheel or tire angle changes over time during the vehicle steering process. In practical applications, the turning angle velocity may be used to describe the speed at which the turning angle changes.

[0077] The steering execution system may refer to a system for controlling the steering of a vehicle. In practical applications, the steering execution system may control the steering actuator corresponding to the target vehicle according to the target turning angle information and the turning angle angular velocity information to control the steering of the target vehicle. In specific implementations, the steering actuator may refer to a mechanical device for controlling the steering of a vehicle. The steering actuator may include, but is not limited to, various motors.

[0078] As an example, the on-board controller may input vehicle environment information, vehicle state information, and driver state information into a pre-trained steering intention prediction model. Since the pre-trained steering intention prediction model has pre-learned information such as motor parameters in the vehicle steering control process, the pre-trained steering intention prediction model may determine the target turning angle information and turning angle angular velocity information of the target vehicle by analyzing the vehicle environment information, vehicle state information, and driver state information of the target vehicle. Afterwards, the on-board controller may send the target turning angle information and turning angle angular velocity information to the steering execution system of the target vehicle. The steering execution system of the target vehicle includes multiple / multiple steering actuators. The steering execution system of the target vehicle may generate control instructions for each steering actuator based on the target turning angle information and turning angle angular velocity information, so as to control the steering actuator corresponding to the target vehicle, thereby controlling the steering of the target vehicle.

[0079] In the above-mentioned vehicle steering control method, when receiving a steering command for a target vehicle, the vehicle environment information, vehicle state information and driver state information of the target vehicle are obtained; thereby timely responding to the steering command, obtaining information such as the vehicle environment, vehicle state and driver state, so as to subsequently control the vehicle steering based on the vehicle environment, vehicle state and driver state; the vehicle environment information, vehicle state information and driver state information are input into a pre-trained steering intention prediction model to obtain the target turning angle information and turning angle angular velocity information of the target vehicle; thereby, based on the pre-trained steering intention prediction model, the vehicle environment, vehicle state and driver state information are combined in real time , accurately analyze the turning angle information and turning angle angular velocity information used to control the vehicle steering; send the target turning angle information and turning angle angular velocity information to the steering execution system; the steering execution system is used to control the steering actuator corresponding to the target vehicle according to the target turning angle information and turning angle angular velocity information to control the steering of the target vehicle, thereby using the vehicle's steering execution system to accurately control the vehicle steering based on the target turning angle and turning angle angular velocity, and can use a pre-trained model combined with various information such as vehicle environment, vehicle status and driver status to analyze and predict the turning angle and turning angle velocity, and accurately control the vehicle steering based on the turning angle and turning angle velocity through the steering execution system, thereby improving the accuracy of vehicle steering control.

[0080] In an exemplary embodiment, vehicle environment information, vehicle state information and driver state information are input into a pre-trained steering intention prediction model to obtain target turning angle information and turning angle velocity information of the target vehicle, including: inputting vehicle environment information, vehicle state information and driver state information into a pre-trained steering intention prediction model to obtain predicted turning angle information and turning angle velocity information of the target vehicle; acquiring steering wheel angle information of the target vehicle from the vehicle state information, and obtaining the target turning angle information of the target vehicle using the predicted turning angle information and the steering wheel angle information.

[0081] The predicted turning angle information may refer to information used to control the turning angle of the target vehicle (such as steering wheel angle / wheel angle, etc.).

[0082] The steering wheel angle information may refer to information representing the rotation angle of the steering wheel of the target vehicle.

[0083] As an example, after the on-board controller inputs the vehicle environment information, vehicle state information and driver state information into the pre-trained steering intention prediction model, since the pre-trained steering intention prediction model has pre-learned information such as motor parameters in the vehicle steering control process, the pre-trained steering intention prediction model can determine the initial turning angle information and turning angle angular velocity information of the target vehicle by analyzing the vehicle environment information, vehicle state information and driver state information of the target vehicle. After the on-board controller obtains the initial turning angle information and turning angle angular velocity information output by the pre-trained steering intention prediction model, the on-board controller can obtain the steering wheel angle information of the target vehicle from the vehicle state information, and calculate the parameters such as the sum / difference between the initial turning angle information and the steering wheel angle information to determine the target turning angle information of the target vehicle.

[0084] In this embodiment, the predicted turning angle information and turning angle velocity information of the target vehicle are obtained by inputting the vehicle environment information, the vehicle state information and the driver state information into a pre-trained steering intention prediction model; the steering wheel angle information of the target vehicle is obtained from the vehicle state information, and the target turning angle information of the target vehicle is obtained using the predicted turning angle information and the steering wheel angle information. The accurate target turning angle can be obtained by combining the steering wheel angle of the target vehicle and the initial turning angle output by the model, so that the vehicle's steering execution system can be used to accurately control the vehicle steering based on the target turning angle and turning angle velocity, thereby improving the accuracy of the vehicle steering control.

[0085] In an exemplary embodiment, target turning angle information of a target vehicle is obtained using predicted turning angle information and steering wheel angle information, including: using the predicted turning angle information and steering wheel angle information to determine weight information of the steering wheel angle information and weight information of the predicted turning angle information; taking a weighted sum of the steering wheel angle information and the predicted turning angle information according to the weight information of the steering wheel angle information and the weight information of the predicted turning angle information to obtain the target turning angle information of the target vehicle.

[0086] As an example, the vehicle controller can calculate the weight information of the steering wheel angle information and the weight information of the predicted angle information based on the predicted angle information and the steering wheel angle information, combined with the preset angle relationship coefficient, and then the vehicle controller can weight the steering wheel angle information and the predicted angle information according to the weight information of the steering wheel angle information and the weight information of the predicted angle information, and obtain the target angle information of the target vehicle. In practical applications, the steering wheel angle information can be expressed as θ, the angle relationship coefficient can be expressed as γ, the predicted angle information can be expressed as ε, and the target angle information can be expressed as δ, then the weight information α of the predicted angle information can be expressed as:

[0087]

[0088] The weight information of the steering wheel angle information can be expressed as 1-α, and the target angle information δ can be expressed as:

[0089] δ=(1-α)·γθ+α·ε.

[0090] In practical applications, ε can represent the predicted turning angle at the next moment, γθ can represent the turning intention at the current moment, the weight information of the steering wheel angle information and the weight information of the predicted turning angle information can affect the degree of change of the turning angle of the target vehicle at the next moment. According to the weight information of the steering wheel angle information and the weight information of the predicted turning angle information, the steering wheel angle information and the predicted turning angle information are weighted and summed to obtain the target turning angle information of the target vehicle, so as to achieve a smooth transition of the turning angle at the next moment. In special driving situations, a sudden change in the steering angle will cause the execution system to shake, which will give the driver an uncomfortable driving experience. Angle prediction and weighted determination of the target angle can help alleviate the uncomfortable driving experience of the vehicle caused by drastic changes in the angle.

[0091] In this embodiment, the predicted turning angle information and the steering wheel angle information are used to determine the weight information of the steering wheel angle information and the weight information of the predicted turning angle information; based on the weight information of the steering wheel angle information and the weight information of the predicted turning angle information, the steering wheel angle information and the predicted turning angle information are weighted and summed to obtain the target turning angle information of the target vehicle; the steering wheel angle information and the predicted turning angle information can be weighted based on the weight to obtain an accurate target turning angle, so as to accurately control the vehicle steering based on the turning angle and the turning angular velocity, thereby improving the accuracy of the vehicle steering control.

[0092] In an exemplary embodiment, the steering intention prediction model includes a first branch and a second branch; the vehicle environment information, the vehicle state information and the driver state information are input into the pre-trained steering intention prediction model to obtain the predicted turning angle information and the turning angle velocity information of the target vehicle, including: the vehicle environment information, the vehicle state information and the driver state information are input into the steering intention prediction model, and the predicted turning angle information and the turning angle velocity information are obtained through the first branch; the above method also includes: obtaining the target current information of the road sense motor of the target vehicle through the second branch; sending the target current information of the road sense motor to the road sense execution system; the road sense execution system is used to control the road sense motor corresponding to the target vehicle according to the target current information of the road sense motor, so as to feedback the road conditions to the driver of the target vehicle.

[0093] Among them, the road sense motor target current information may refer to information characterizing the current of the road sense motor used to control the target vehicle. In practical applications, the road sense motor may refer to a device used to help the driver turn the steering wheel more easily and simulate the road feedback transmitted to the driver through the steering wheel in a traditional mechanical or hydraulic steering system.

[0094] Among them, the road sense execution system may refer to a system used to generate and adjust the road surface feedback felt by the driver through the steering wheel. In actual applications, the road sense execution system can control the road sense motor corresponding to the target vehicle according to the target current information of the road sense motor to feedback the road conditions to the driver of the target vehicle.

[0095] As an example, the steering intention prediction model may include a first branch and a second branch. When the vehicle environment information, vehicle state information and driver state information are input into the pre-trained steering intention prediction model, the first branch in the pre-trained steering intention prediction model may analyze the vehicle environment information, vehicle state information and driver state information to determine the predicted turning angle information and turning angle velocity information. The second branch in the pre-trained steering intention prediction model may analyze the vehicle environment information, vehicle state information and driver state information to determine the target current information of the road sense motor of the target vehicle. After that, the on-board controller may send the target current information of the road sense motor of the target vehicle to the road sense execution system. The road sense execution system may control the road sense motor corresponding to the target vehicle according to the target current information of the road sense motor, thereby feeding back the road conditions to the driver of the target vehicle.

[0096] In this embodiment, vehicle environment information, vehicle state information and driver state information are input into a steering intention prediction model, and predicted turning angle information and turning angle angular velocity information are obtained through the first branch of the steering intention prediction model; the target current information of the road sense motor of the target vehicle is obtained through the second branch of the steering intention prediction model; the target current information of the road sense motor is sent to the road sense execution system; the road sense execution system is used to control the road sense motor corresponding to the target vehicle according to the target current information of the road sense motor, so as to feedback the road condition to the driver of the target vehicle, and can determine the current information used to control the road sense motor while determining the turning angle and turning angle velocity, so as to accurately control the road sense motor and accurately feedback the road condition to the driver.

[0097] In some embodiments, before obtaining the vehicle environment information, vehicle status information and driver status information of the target vehicle, the above method also includes: obtaining sample vehicle environment information, sample vehicle status information and sample driver status information of the sample vehicle; determining the vehicle environment coding information corresponding to the sample vehicle environment information, the vehicle status coding information corresponding to the sample vehicle status information and the driver status coding information corresponding to the sample driver status information based on the sample vehicle environment information, the sample vehicle status information and the sample driver status information; inputting the vehicle environment coding information, the vehicle status coding information and the driver status coding information into the steering intention prediction model to be trained to obtain the road sense motor current information of the sample vehicle; obtaining the sample road sense motor current information of the sample vehicle, and training the second branch in the steering intention prediction model to be trained based on the road sense motor current information and the sample road sense motor current information of the sample vehicle.

[0098] The sample vehicle may be the same as the target vehicle.

[0099] Among them, the sample vehicle environment information may refer to information that characterizes the environment in which the sample vehicle is located. In actual applications, the sample vehicle environment information may include but is not limited to information that characterizes the road conditions of the road on which the sample vehicle is traveling, information that characterizes the visibility of the environment in which the sample vehicle is located, information that characterizes the number of vehicles in the environment in which the sample vehicle is located, information that characterizes the number of lanes in the environment in which the sample vehicle is located, etc.

[0100] Among them, the sample vehicle status information may refer to information characterizing the operating status of the sample vehicle. In actual applications, the sample vehicle status information may include but is not limited to information characterizing the vehicle speed, longitudinal acceleration, lateral acceleration, yaw angular velocity, wheel angle, steering wheel angle, fault conditions, and temperatures of various parts / components of the vehicle.

[0101] Among them, the sample driver status information may refer to information that characterizes the physical condition, psychological state, and behavioral performance of the driver of the sample vehicle during driving. In practical applications, the sample driver status information may be obtained by analyzing the images / videos captured by the camera of the sample vehicle of the driver.

[0102] The vehicle environment coding information may refer to information obtained by encoding the sample vehicle environment information.

[0103] The vehicle status encoding information may refer to information obtained by encoding the sample vehicle status information.

[0104] The driver status coding information may refer to information obtained by encoding sample driver status information.

[0105] Among them, the road sense motor current information of the sample vehicle may refer to the information output by the steering intention prediction model to be trained by analyzing the sample vehicle environment information, sample vehicle state information and sample driver state information (or vehicle environment coding information, vehicle state coding information and driver state coding information), which is used to control the current of the road sense motor of the sample vehicle. In practical applications, the road sense motor current information of the sample vehicle can be used as a prediction value.

[0106] The sample road sense motor current information may refer to information representing the current of the road sense motor that actually controls the sample vehicle. In practical applications, the sample road sense motor current information may be used as an actual value / true value.

[0107] As an example, in order to train the steering intention prediction model, the on-board controller may first obtain the sample vehicle environment information, sample vehicle state information and sample driver state information of the sample vehicle, and then the on-board controller may encode the sample vehicle environment information, the sample vehicle state information and the sample driver state information, and determine the vehicle environment encoding information corresponding to the sample vehicle environment information, the vehicle state encoding information corresponding to the sample vehicle state information and the driver state encoding information corresponding to the sample driver state information, and then the on-board controller may input the vehicle environment encoding information, the vehicle state encoding information and the driver state encoding information into the steering intention prediction model to be trained, and the second branch in the steering intention prediction model to be trained may determine the road sense motor current information of the sample vehicle by analyzing the vehicle environment encoding information, the vehicle state encoding information and the driver state encoding information. The on-board controller may obtain the real sample road sense motor current information of the sample vehicle, and train the second branch in the steering intention prediction model to be trained according to the road sense motor current information and the sample road sense motor current information of the sample vehicle.

[0108] In this embodiment, by obtaining sample vehicle environment information, sample vehicle status information and sample driver status information of a sample vehicle; determining the vehicle environment coding information corresponding to the sample vehicle environment information, the vehicle status coding information corresponding to the sample vehicle status information and the driver status coding information corresponding to the sample driver status information according to the sample vehicle environment information, the sample vehicle status information and the sample driver status information; inputting the vehicle environment coding information, the vehicle status coding information and the driver status coding information into the steering intention prediction model to be trained to obtain the road sense motor current information of the sample vehicle; obtaining the sample road sense motor current information of the sample vehicle, and training the second branch in the steering intention prediction model to be trained according to the road sense motor current information and the sample road sense motor current information of the sample vehicle, the second branch of the steering intention prediction model can be accurately trained based on the sample information, and the second branch of the steering intention prediction model can be optimized so that the second branch of the steering intention prediction model can output accurate road sense motor current information, thereby improving the accuracy of the road sense motor current information.

[0109] In some embodiments, before obtaining the vehicle environment information, vehicle status information and driver status information of the target vehicle, the method further includes: obtaining sample vehicle environment information, sample vehicle status information and sample driver status information of a sample vehicle; determining the vehicle environment coding information corresponding to the sample vehicle environment information, the vehicle status coding information corresponding to the sample vehicle status information and the driver status coding information corresponding to the sample driver status information based on the sample vehicle environment information, the sample vehicle status information and the sample driver status information; inputting the vehicle environment coding information, the vehicle status coding information and the driver status coding information into the steering intention prediction model to be trained to obtain the predicted turning angle information and the predicted turning angle angular velocity information of the sample vehicle; obtaining the sample turning angle information and the sample turning angle velocity information of the sample vehicle, and training the first branch in the steering intention prediction model to be trained based on the predicted turning angle information of the sample vehicle, the predicted turning angle angular velocity information of the sample vehicle, the sample turning information and the sample turning angle velocity information.

[0110] The sample vehicle may be the same as the target vehicle.

[0111] Among them, the sample vehicle environment information may refer to information that characterizes the environment in which the sample vehicle is located. In actual applications, the sample vehicle environment information may include but is not limited to information that characterizes the road conditions of the road on which the sample vehicle is traveling, information that characterizes the visibility of the environment in which the sample vehicle is located, information that characterizes the number of vehicles in the environment in which the sample vehicle is located, information that characterizes the number of lanes in the environment in which the sample vehicle is located, etc.

[0112] Among them, the sample vehicle status information may refer to information characterizing the operating status of the sample vehicle. In actual applications, the sample vehicle status information may include but is not limited to information characterizing the vehicle speed, longitudinal acceleration, lateral acceleration, yaw angular velocity, wheel angle, steering wheel angle, fault conditions, and temperatures of various parts / components of the vehicle.

[0113] Among them, the sample driver status information may refer to information that characterizes the physical condition, psychological state, and behavioral performance of the driver of the sample vehicle during driving. In practical applications, the sample driver status information may be obtained by analyzing the images / videos captured by the camera of the sample vehicle of the driver.

[0114] The vehicle environment coding information may refer to information obtained by encoding the sample vehicle environment information.

[0115] The vehicle status encoding information may refer to information obtained by encoding the sample vehicle status information.

[0116] The driver status coding information may refer to information obtained by encoding sample driver status information.

[0117] Among them, the predicted turning angle information of the sample vehicle may refer to the information used to control the turning angle of the sample vehicle output by the steering intention prediction model to be trained by analyzing the sample vehicle environment information, sample vehicle state information and sample driver state information (or vehicle environment coding information, vehicle state coding information and driver state coding information). In practical applications, the predicted turning angle information of the sample vehicle can be used as a prediction value.

[0118] Among them, the predicted turning angle angular velocity information of the sample vehicle may refer to the information output by the steering intention prediction model to be trained by analyzing the sample vehicle environment information, sample vehicle state information and sample driver state information (or vehicle environment coding information, vehicle state coding information and driver state coding information), which is used to control the rate of change of the turning angle of the sample vehicle. In practical applications, the predicted turning angle angular velocity information of the sample vehicle can be used as a prediction value.

[0119] The sample turning angle information may refer to information representing the turning angle of the actual controlled sample vehicle. In practical applications, the sample turning angle information may be used as an actual value / real value.

[0120] The sample turning angle velocity information may refer to information representing the rate of change of the turning angle of the actual control sample vehicle. In practical applications, the sample turning angle velocity information may be used as an actual value / true value.

[0121] As an example, in order to train the steering intention prediction model, the on-board controller may first obtain the sample vehicle environment information, sample vehicle state information and sample driver state information of the sample vehicle, and then the on-board controller may encode the sample vehicle environment information, the sample vehicle state information and the sample driver state information, determine the vehicle environment encoding information corresponding to the sample vehicle environment information, the vehicle state encoding information corresponding to the sample vehicle state information and the driver state encoding information corresponding to the sample driver state information, and then the on-board controller may input the vehicle environment encoding information, the vehicle state encoding information and the driver state encoding information into the steering intention prediction model to be trained, and the first branch in the steering intention prediction model to be trained may obtain the predicted turning angle information and the predicted turning angle angular velocity information of the sample vehicle by analyzing the vehicle environment encoding information, the vehicle state encoding information and the driver state encoding information. The on-board controller may obtain the real sample turning angle information and the sample turning angle angular velocity information of the sample vehicle, and train the first branch in the steering intention prediction model to be trained according to the predicted turning angle information of the sample vehicle, the predicted turning angle angular velocity information of the sample vehicle, the sample turning angle information and the sample turning angle angular velocity information. It can be understood that after the first branch and the second branch are trained, a trained turning intention prediction model can be obtained, and the trained turning intention prediction model can be used as a pre-trained turning intention prediction model.

[0122] In this embodiment, by obtaining sample vehicle environment information, sample vehicle state information and sample driver state information of a sample vehicle; determining the vehicle environment coding information corresponding to the sample vehicle environment information, the vehicle state coding information corresponding to the sample vehicle state information and the driver state coding information corresponding to the sample driver state information according to the sample vehicle environment information, the sample vehicle state information and the sample driver state information; inputting the vehicle environment coding information, the vehicle state coding information and the driver state coding information into the steering intention prediction model to be trained to obtain the predicted turning angle information and the predicted turning angle angular velocity information of the sample vehicle; obtaining the sample turning angle information and the sample turning angle angular velocity information of the sample vehicle, and training the first branch in the steering intention prediction model to be trained according to the predicted turning angle information of the sample vehicle, the predicted turning angle angular velocity information of the sample vehicle, the sample turning angle information and the sample turning angle velocity information; the first branch of the steering intention prediction model to be trained can be accurately trained based on the sample information, and the first branch of the steering intention prediction model can be optimized so that the first branch of the steering intention prediction model can output accurate turning angle information and turning angle angular velocity information, thereby improving the accuracy of the turning angle information and the turning angle angular velocity information.

[0123] In some embodiments, after obtaining the sample turning angle information and sample turning angle angular velocity information of the sample vehicle, the above method also includes: obtaining the mass information of the sample vehicle and the gender information and preference information of the corresponding driver of the sample vehicle; determining the model structure and model parameters of the steering intention prediction model to be trained based on the mass information, gender information and preference information.

[0124] The mass information may refer to information characterizing the weight of the sample vehicle.

[0125] The driver's gender information may refer to information representing the gender of the driver of the sample vehicle.

[0126] The driver's preference information may refer to information characterizing the driving habits of the driver of the sample vehicle.

[0127] As an example, different vehicles and drivers will affect the actual process of vehicle steering control. Therefore, the on-board controller can first obtain the quality information of the sample vehicle and the gender information and preference information of the corresponding driver of the sample vehicle. Then, the on-board controller can determine the model structure and model parameters of the steering intention prediction model to be trained based on the quality information, gender information and preference information. In practical applications, the quality information, gender information and preference information can be used as a set of hyperparameters, and each set of hyperparameters can have corresponding model structures and model parameters. Therefore, after determining the quality of the vehicle and the gender and preference of the driver, the on-board controller can first preliminarily determine the model structure and model parameters of the steering intention prediction model to be trained based on the quality of the vehicle and the gender and preference of the driver. Then, the on-board controller can train the steering intention prediction model to be trained based on the sample information until the trained steering intention prediction model is obtained. In the specific implementation, the hyperparameters can be used as key data other than the input data, and will be set in advance in the model to adjust the model structure and related important parameters. Different hyperparameter designs will affect the output results of the model. Information such as vehicle quality, driver gender, and driver preference is used as a model adjustable hyperparameter. The vehicle mass can be mapped to a reasonable range of the feature matrix so that the vehicle mass can be mapped to the range of (0-1). The mapping method can be expressed as:

[0128]

[0129] Where m is the vehicle mass, b is m is the mass offset, usually 1000kg, x m is the mass of the vehicle after mapping.

[0130] Similarly, driver gender and driver preference can also be mapped to a reasonable range of the feature matrix. The mapping of driver gender can be expressed as:

[0131]

[0132] Among them, x s is the mapped gender of the driver.

[0133] The mapping of driver preferences can be expressed as:

[0134]

[0135] Among them, x model It is the mapped driver preference, wherein the driver preference may be pre-set with preference setting types such as comfort, standard, and sport.

[0136] In this embodiment, by obtaining the quality information of the sample vehicle and the gender information and preference information of the corresponding driver of the sample vehicle; determining the model structure and model parameters of the steering intention prediction model to be trained according to the quality information, gender information and preference information, the model structure and model parameters of the steering intention prediction model can be accurately determined based on the quality of the vehicle and the gender and preference of the driver, so that the steering intention prediction model can adapt to different vehicles and different drivers to achieve personalized steering control.

[0137] In some embodiments, based on sample vehicle environment information, sample vehicle status information and sample driver status information, vehicle environment coding information corresponding to the sample vehicle environment information, vehicle status coding information corresponding to the sample vehicle status information and driver status coding information corresponding to the sample driver status information are determined, including: determining a vehicle environment time series based on the sample vehicle environment information, determining a vehicle status time series based on the sample vehicle status information, and determining a driver status time series based on the sample driver status information; dividing the vehicle environment time series, the vehicle status time series and the driver status time series according to preset sequence length information to obtain sequence blocks corresponding to the vehicle environment time series, sequence blocks corresponding to the vehicle status time series and sequence blocks corresponding to the driver status time series; convolving the sequence blocks corresponding to the vehicle environment time series, the sequence blocks corresponding to the vehicle status time series and the sequence blocks corresponding to the driver status time series, and embedding the convolution results into a vector corresponding to the sample vehicle environment information, a vector corresponding to the sample vehicle status information and a vector corresponding to the sample driver status information to obtain vehicle environment coding information, vehicle status coding information and driver status coding information.

[0138] The vehicle environment time series may refer to information obtained by sorting sample vehicle environment information in chronological order.

[0139] The vehicle status time series may refer to information obtained by sorting sample vehicle status information in chronological order.

[0140] The driver status time series may refer to information obtained by sorting sample driver status information in chronological order.

[0141] The preset sequence length information may refer to a preset sequence length.

[0142] The sequence block may refer to a sequence obtained by dividing the time series according to preset sequence length information. In practical applications, at least one sequence block may be obtained by dividing the time series according to preset sequence length information.

[0143] Among them, the convolution result may refer to the information obtained after convolving the sequence blocks corresponding to the vehicle environment time series, the sequence blocks corresponding to the vehicle status time series, and the sequence blocks corresponding to the driver status time series.

[0144] Herein, a vector may refer to information obtained after mapping information.

[0145] As an example, in order to encode vehicle environment information, vehicle state information and driver state information, the on-board controller may first obtain sample vehicle environment information, sample vehicle state information and sample driver state information of the sample vehicle at different time points, and sort the sample vehicle environment information at different time points in chronological order to obtain a vehicle environment time series. Similarly, the on-board controller may determine the vehicle state time series according to the sample vehicle state information, and determine the driver state time series according to the sample driver state information. Afterwards, the on-board controller may divide the vehicle environment time series, the vehicle state time series and the driver state time series respectively according to the preset sequence length information to obtain a sequence block corresponding to the vehicle environment time series, a sequence block corresponding to the vehicle state time series and a sequence block corresponding to the driver state time series. Then, the on-board controller may convolve the sequence block corresponding to the vehicle environment time series, the sequence block corresponding to the vehicle state time series and the sequence block corresponding to the driver state time series, and embed the convolution result into the vector corresponding to the sample vehicle environment information, the vector corresponding to the sample vehicle state information and the vector corresponding to the sample driver state information to obtain vehicle environment coding information, vehicle state coding information and driver state coding information. In practical applications, for any one of the vehicle environment time series, vehicle state time series and driver state time series, the on-board controller can take the information at three consecutive time points as the sequence block to be convolved, and perform convolution processing on the sequence block to be convolved to obtain the convolution result corresponding to the sequence block to be convolved. The on-board controller can embed the convolution result corresponding to the sequence block to be convolved into the vector corresponding to the sample vehicle environment information, the vector corresponding to the sample vehicle state information and the vector corresponding to the sample driver state information to obtain the vehicle environment coding information, the vehicle state coding information and the driver state coding information.

[0146] In this embodiment, a vehicle environment time series is determined according to sample vehicle environment information, a vehicle state time series is determined according to sample vehicle state information, and a driver state time series is determined according to sample driver state information; the vehicle environment time series, the vehicle state time series, and the driver state time series are divided according to preset sequence length information to obtain sequence blocks corresponding to the vehicle environment time series, sequence blocks corresponding to the vehicle state time series, and sequence blocks corresponding to the driver state time series; the sequence blocks corresponding to the vehicle environment time series, the sequence blocks corresponding to the vehicle state time series, and the sequence blocks corresponding to the driver state time series are convolved, and the convolution results are embedded into the vector corresponding to the sample vehicle environment information, the vector corresponding to the sample vehicle state information, and the vector corresponding to the sample driver state information to obtain vehicle environment coding information, vehicle state coding information, and driver state coding information. It is possible to interact and encode multiple information to obtain accurate vehicle environment coding information, vehicle state coding information, and driver state coding information, so that the steering intention prediction model outputs accurate information for controlling vehicle steering based on the vehicle environment coding information, the vehicle state coding information, and the driver state coding information, thereby improving the accuracy of vehicle steering control.

[0147] In some embodiments, the steering intention prediction model to be trained includes at least one temporal convolutional network, the temporal convolutional network includes a depthwise separable convolutional layer, a pointwise convolutional layer and a normalization layer, the vehicle environment coding information, the vehicle state coding information and the driver state coding information are input into the steering intention prediction model to be trained, and the predicted turning angle information and the predicted turning angle angular velocity information of the sample vehicle are obtained, including: through the depthwise separable convolutional layer in the temporal convolutional network, the vehicle environment coding information, the vehicle state coding information and the driver state coding information are compared according to the convolution kernel corresponding to the depthwise separable convolutional layer. The information is convolved to obtain the output result of the depthwise separable convolution layer; the output result of the depthwise separable convolution layer is convolved according to the convolution kernel corresponding to the pointwise convolution layer through the point-by-point convolution layer in the temporal convolution network to obtain the output result of the point-by-point convolution layer; the output result of the point-by-point convolution layer is normalized through the normalization layer in the temporal convolution network to obtain the output result of the normalization layer; the predicted turning angle information and the predicted turning angle angular velocity information of the sample vehicle are determined according to the vehicle environment coding information, the vehicle state coding information, the driver state coding information and the output result of the normalization layer.

[0148] Among them, the temporal convolutional network can refer to a deep learning model for processing sequence data.

[0149] Among them, the depthwise separable convolution layer (Depthwise Separable Convolution, DWConv) may refer to a model for implementing deep convolution processing.

[0150] Among them, the pointwise convolution layer (Pointwise Convolution, PWConv) may refer to a model for implementing pointwise convolution processing.

[0151] The normalization layer may refer to a model used to implement normalization processing.

[0152] As an example, the steering intention prediction model to be trained includes at least one temporal convolutional network, and the temporal convolutional network includes a depth-separable convolutional layer, a point-by-point convolutional layer, and a normalization layer. The step of determining the turning angle information and the turning angle angular velocity information based on the vehicle environment coding information, the vehicle state coding information, and the driver state coding information by the steering intention prediction model to be trained may include: the vehicle controller performs convolution processing on the vehicle environment coding information, the vehicle state coding information, and the driver state coding information through the depth-separable convolutional layer in the temporal convolutional network according to the convolution kernel corresponding to the depth-separable convolutional layer, and obtains the output result of the depth-separable convolutional layer; then the vehicle controller performs convolution processing on the output result of the depth-separable convolutional layer through the point-by-point convolutional layer in the temporal convolutional network according to the convolution kernel corresponding to the point-by-point convolutional layer, and obtains the output result of the point-by-point convolutional layer; then the vehicle controller normalizes the output result of the point-by-point convolutional layer through the normalization layer in the temporal convolutional network, and obtains the output result of the normalization layer. The temporal convolutional network can also have a residual structure, so the on-board controller can determine the predicted turning angle information and predicted turning angle velocity information of the sample vehicle based on the vehicle environment encoding information, vehicle state encoding information, driver state encoding information and the output results of the normalization layer.

[0153] In this embodiment, the vehicle environment coding information, the vehicle state coding information and the driver state coding information are convolved according to the convolution kernel corresponding to the depthwise separable convolution layer through the depthwise separable convolution layer to obtain the output result of the depthwise separable convolution layer; the output result of the depthwise separable convolution layer is convolved according to the convolution kernel corresponding to the pointwise convolution layer through the pointwise convolution layer in the time convolution network to obtain the output result of the pointwise convolution layer; the output result of the pointwise convolution layer is normalized through the normalization layer in the time convolution network to obtain the output result of the normalization layer; according to the vehicle environment coding information, the vehicle state coding information, the driver state coding information and the output result of the normalization layer, the predicted turning angle information and the predicted turning angle angular velocity information of the sample vehicle are determined, and based on the timewise convolution network in the steering intention prediction model, different convolution kernels can be used to convolve the vehicle environment coding information, the vehicle state coding information and the driver state coding information respectively, so as to realize the deep interaction of multiple information.

[0154] In some embodiments, the temporal convolutional network also includes a first convolutional layer and a second convolutional layer, and the convolution kernel corresponding to the first convolutional layer is different from the convolution kernel corresponding to the second convolutional layer. After obtaining the output result of the depthwise separable convolutional layer, the method also includes: performing convolution processing on the output result of the depthwise separable convolutional layer by the convolution kernel corresponding to the first convolutional layer in the temporal convolutional network, so as to convolve the feature channels corresponding to the vehicle environment coding information, the vehicle environment coding information and the driver state coding information, and obtain the output result of the first convolutional layer; performing convolution processing on the output result of the first convolutional layer by the convolution kernel corresponding to the second convolutional layer in the temporal convolutional network, so as to convolve the vehicle environment coding information, the vehicle state coding information and the driver state coding information, and obtain the output result of the second convolutional layer; determining the predicted turning angle information and predicted turning angle angular velocity information of the sample vehicle according to the output result of the second convolutional layer, the vehicle environment coding information, the vehicle state coding information and the driver state coding information.

[0155] As an example, the temporal convolutional network may also include a first convolutional layer and a second convolutional layer, and the convolutional kernel corresponding to the first convolutional layer is different from the convolutional kernel corresponding to the second convolutional layer. After the on-board controller obtains the output result of the depth-separable convolutional layer, the on-board controller may convolve the output result of the depth-separable convolutional layer through the convolutional kernel corresponding to the first convolutional layer in the temporal convolutional network, so as to convolve the feature channels corresponding to the vehicle environment coding information, the vehicle environment coding information and the driver state coding information, and obtain the output result of the first convolutional layer, and then the on-board controller may convolve the output result of the first convolutional layer through the convolutional kernel corresponding to the second convolutional layer in the temporal convolutional network, so as to convolve the vehicle environment coding information, the vehicle state coding information and the driver state coding information, and obtain the output result of the second convolutional layer. The temporal convolutional network may also have a residual structure, so the on-board controller may determine the predicted turning angle information and the predicted turning angle angular velocity information of the sample vehicle according to the output result of the second convolutional layer, the vehicle environment coding information, the vehicle state coding information and the driver state coding information.

[0156] In this embodiment, the output result of the depthwise separable convolutional layer is convolved by the convolution kernel corresponding to the first convolutional layer in the temporal convolutional network, so as to convolve the feature channels corresponding to the vehicle environment coding information, the vehicle environment coding information and the driver state coding information, and obtain the output result of the first convolutional layer; the output result of the first convolutional layer is convolved by the convolution kernel corresponding to the second convolutional layer in the temporal convolutional network, so as to convolve the vehicle environment coding information, the vehicle state coding information and the driver state coding information, and obtain the output result of the second convolutional layer; according to the output result of the second convolutional layer, the vehicle environment coding information, the vehicle state coding information and the driver state coding information, the predicted turning angle information and the predicted turning angle angular velocity information of the sample vehicle are determined, and based on the temporal convolutional network in the steering intention prediction model, different convolution kernels can be used to convolve the vehicle environment coding information, the vehicle state coding information and the driver state coding information, respectively, to achieve deep interaction of multiple information.

[0157] In some embodiments, vehicle environment information, vehicle status information and driver status information of a target vehicle are obtained, including: obtaining road surface type information, visibility information, terrain information, vehicle quantity information, lane quantity information and traffic complexity information of the target vehicle's environment through a camera of the target vehicle, and determining the vehicle environment information of the target vehicle based on the road surface type information, visibility information, terrain information, vehicle quantity information, lane quantity information and traffic complexity information; determining the vehicle status information of the target vehicle based on the vehicle speed information, acceleration information, yaw angular velocity information, wheel angle information, steering wheel angle information, temperature information and fault information of the target vehicle; and determining the driver status information of the target vehicle based on the behavior information of the driver corresponding to the target vehicle.

[0158] The road surface type information may refer to information characterizing the type of road surface on which the vehicle is traveling. In practical applications, the road surface type characterized by the road surface type information may include, but is not limited to, a high-grade road surface, a medium-grade road surface, a low-grade road surface, and a dangerous road surface. The high-grade road surface is a road surface with excellent anti-skid conditions such as a highway or a first-class highway built with high-grade asphalt concrete materials, with more than 4 lanes and a relatively large road width of 40-70m; the second-grade road surface is a third- or fourth-class highway built with asphalt-penetrated gravel and asphalt surface technology, mostly suburban roads, with more than 4 lanes and a moderate road width (30-6 0m); Intermediate road surface is a branch road of Class IV highway built with mud-bound or graded crushed stone, irregular block stone materials, usually a local road, with no less than 2 lanes and a narrow road width (20-40m); Low-grade road surface is a rural road built with various slag soil, gravel soil and sand-gravel soil, with a speed limit of no more than 30kph, no less than 2 lanes, a road width of 16-30m, and no dividing strip; Dangerous road surface is a slippery mud road surface with low friction coefficient such as rolling sand and gravel, soft soil, and snowmelt cover with poor driving conditions, with less than 2 lanes, extremely narrow roads, and no dividing strips or other traffic facilities.

[0159] Among them, visibility information may refer to information that characterizes the visibility of the vehicle's environment. In actual applications, the visibility types represented by the visibility information may include but are not limited to high visibility and low visibility. A high visibility environment refers to an environment where visibility is greater than 500m, and the driver has a clear line of sight and can accurately judge the road conditions and obstacles; a low visibility environment refers to an environment where visibility is less than 200m, and the driver's line of sight is obstructed, making it difficult to accurately judge the road conditions and obstacles.

[0160] Among them, terrain information can refer to information that characterizes the terrain of the vehicle's environment. In actual applications, the terrain types represented by the terrain information can include but are not limited to simple terrain and complex terrain. Among them, simple terrain refers to a road environment with flat terrain and no obstacles; complex terrain refers to a road environment with multiple slopes, multiple bends and multiple obstacles.

[0161] Among them, the vehicle quantity information may refer to information representing the number of vehicles in the environment in which the vehicle is located. In actual applications, the vehicle quantity information may represent the total number of vehicles within a range with the vehicle itself as the center and a preset distance as the radius.

[0162] The lane quantity information may refer to information representing the number of lanes in the environment where the vehicle is located.

[0163] The traffic complexity information may be information indicating the complexity of the traffic conditions in the vehicle's environment. In practical applications, the traffic complexity information may represent information such as the number of vehicles, the number of lanes, and the type of lanes in the vehicle's environment.

[0164] The temperature information may refer to information characterizing the temperature of various locations / components of the vehicle. In practical applications, the temperature information may include but is not limited to the average temperature of the vehicle as a whole, the system temperature of various systems in the vehicle, etc.

[0165] The fault information may refer to information indicating whether there is a fault in the vehicle and the type of the fault in the vehicle.

[0166] The driver's behavior information may refer to information representing the driver's behavior. In practical applications, the driver's behavior information may represent the driver's concentration level, mental state, etc.

[0167] As an example, the model input of the steering intention prediction model can be multiple time series (such as vehicle environment information, vehicle status information and driver status information), and the time series can constitute a feature matrix. Each time series contains the changes in specific variables (i.e., specific types of information, such as vehicle environment information, vehicle status information and driver status information) in the corresponding time period. The entire feature matrix can feedback the system / vehicle status information, which can be used as the input feature of the deep learning model (such as the steering intention prediction model) and obtain the prediction result through the time series network model (such as the steering intention prediction model). In practical applications, the on-board controller can comprehensively predict the road conditions based on the vehicle's forward camera and suspension road preview results. As the road type information, the road type information can be determined based on the road specifications and road environment. The relationship between the road type information and the road specifications and road environment can be expressed as shown in Table 1 below:

[0168] Table 1

[0169]

[0170] Among them, the road surface type information can determine the road surface classification of A, B, C, D, and E. The road surface classifications A, B, C, D, and E correspond to characteristic values ​​5, 15, 25, 35, and 45, respectively, and the road surface conditions decrease in sequence.

[0171] The on-board controller can determine the traffic complexity based on information such as the number of vehicles, the number of lanes, and the lane type. The relationship between the traffic complexity and the number of vehicles, the number of lanes, and the lane type can be expressed as shown in the following Table 2:

[0172] Table 2

[0173]

[0174] Among them, the traffic complexity can be divided into levels A, B, C, D, and E. The traffic complexity levels A, B, C, D, and E correspond to characteristic values ​​10, 20, 30, 40, and 50 respectively, and the traffic complexity increases successively.

[0175] The onboard controller can obtain the vehicle speed, longitudinal acceleration, lateral acceleration, yaw rate, wheel angle, and steering wheel angle values ​​as vehicle state input. The vehicle speed can be processed and mapped to (-10, 20), the acceleration is mapped to (-5, 5), and the yaw rate is mapped to (-5, 5). As the input sequence of the deep learning model, the mapping method can be expressed as:

[0176]

[0177] Where x is the vehicle speed, longitudinal acceleration, lateral acceleration, yaw rate, wheel angle, steering wheel angle, steering wheel angular velocity, etc. c and d are the mapping range breakpoints, and c>d. λ is the mapping coefficient, which can control the scaling degree of the data during the mapping process. x c It can be mapped data.

[0178] The on-board controller can determine the system safety level based on the vehicle's fault information and temperature information. The system safety level can be divided into A, B, C, D, and E levels, and the safety level decreases in sequence, corresponding to characteristic values ​​1, 2, 3, 4, and 5 respectively. The system safety level can be rated from the aspects of system temperature and system failure. Among them, the system temperature comes from the controller and the motor temperature sensor signal, which can be divided into (overheating, hot, normal temperature, cold); the fault information comes from the system component fault information and the system fault comprehensive judgment logic, which can be classified into (fatal fault, serious fault, general fault, minor fault and no fault). The relationship between the system safety level and the temperature information and fault information is shown in the following Table 3:

[0179] Table 3

[0180]

[0181] The on-board controller can also determine the driver status information based on the driver status fed back by the driver monitoring system (DMS) inside the vehicle (normal, tired, drowsy, and comatose correspond to characteristic values ​​1, 2, 3, and 4, respectively).

[0182] In this embodiment, the road surface type information, visibility information, terrain information, vehicle quantity information, lane quantity information and traffic complexity information of the environment in which the target vehicle is located are obtained through the camera of the target vehicle, and the vehicle environment information of the target vehicle is determined according to the road surface type information, visibility information, terrain information, vehicle quantity information, lane quantity information and traffic complexity information; the vehicle state information of the target vehicle is determined according to the vehicle speed information, acceleration information, yaw angular velocity information, wheel angle information, steering wheel angle information, temperature information and fault information of the target vehicle; the driver state information of the target vehicle is determined according to the behavior information of the driver corresponding to the target vehicle, and it is possible to determine accurate vehicle environment information based on road surface type, visibility, terrain, vehicle quantity, lane quantity and traffic complexity, determine accurate vehicle state information based on vehicle speed, acceleration, yaw angular velocity, wheel angle, steering wheel angle, temperature and fault, and determine accurate driver state information based on the driver's behavior, thereby improving the accuracy of vehicle environment information, vehicle state information and driver state information.

[0183] In one exemplary embodiment, the on-board controller may first determine hyperparameters based on vehicle mass, driver gender and driver preference, and use the hyperparameters to determine a steering intention prediction model, after which the on-board controller may use the time series corresponding to the road type, visibility, terrain, number of vehicles, number of lanes, traffic complexity, vehicle speed, acceleration, yaw rate, wheel angle, steering wheel angle, temperature, fault and driver status (such as driver condition) as model input data, and then input the model input data into the steering intention prediction model, the steering intention prediction model includes two branches: a steering execution system control branch and a road sense execution system control branch, wherein the steering execution system control branch is used to determine the target angle and target angle velocity at the next moment, and the road sense execution system control branch is used to determine the road sense feedback level and the road sense motor target current at the next moment. The steering intention prediction model can perform convolution calculations on different variables on the time series based on the calculation method of the Modern TCN network for time series information, so as to realize information interaction and prediction across variable dimensions. In practical applications, in order to realize the backward prediction of time series by convolutional network, an empty sequence part is designed before the model input data t0, and the data is filled with 0, corresponding to time t -1 ,t -2 , the length is the convolution kernel length minus 1, such as Figure 2 As shown in Figure 1, a schematic diagram of the model input data is provided. When the convolution kernel moves to the left, the network will perform convolution calculations on the input time series data, as shown in Figure 1. Figure 3 As shown, a schematic diagram of convolution calculation is provided.

[0184] The steering intention prediction model can represent the time series X in the model input data as a sequence of M time series variables of length L, and divide it into N sequence blocks of size P. Finally, the above sequence blocks are embedded into a vector of dimension D in a convolutional manner to realize the encoding of the time series X, which is expressed as X emb Where X emb The dimension is M*N*D, and the encoded input X emb The network is processed by stacking three TCN network blocks (RU time convolutional network), each of which is composed of DWConv, PWConv and normalization layer with residual structure and is recorded as Block TCN ,like Figure 4As shown in the figure, a schematic diagram of a temporal convolutional network is provided. The input is convolved with the features between variables separately through DWConv convolution, so that it can independently learn the temporal dependency of each univariate time series. At this time, the feature dimension is (M*D)*N. Conv_n represents a convolutional module composed of n 1*1 convolutions, ConvBlock k=M and ConvBlock k=D represent convolutional networks composed of convolutional modules of different sizes, respectively, which convolve the feature channels and feature variables, thereby realizing independent interaction between channels and variables. The entire network backbone can be expressed as:

[0185] X i =Block TCN (X i-1 )+X i-1 i=0, 1, 2.

[0186] The network backbone is divided into two branches, which correspond to the output prediction results of the steering execution system and the road sense execution system. Specifically, each branch is a 50*2 and 10*2 linear layer that outputs the prediction results of the branch for two variables. Specifically, it is the target turning angle at the next moment, the turning angle velocity at the next moment, the road sense feedback level, and the road sense motor control current at the next moment (such as the road sense motor target current information).

[0187] like Figure 5 As shown, a schematic diagram of a vehicle steering control system is provided. The vehicle steering control system can input road surface grade, traffic grade, vehicle speed, longitudinal acceleration, lateral acceleration, yaw rate, wheel angle, steering wheel angle, system condition and driver condition as system controller inputs. The system controller includes a control algorithm and control logic including a deep learning model, and an intention fusion algorithm to fuse the actual steering wheel steering intention and the predicted driver steering intention. The intention fusion algorithm can be expressed as:

[0188] δ=(1-α)·γθ+α·ε.

[0189] Among them, θ is the steering wheel angle, γ is the angle relationship coefficient and is inversely proportional to the transmission ratio, ε is the target angle predicted by the deep learning model at the next moment, δ is the target angle received by the steering execution system, and α is the fusion weight coefficient. The calculation expression of the fusion weight coefficient can be expressed as:

[0190]

[0191]

[0192] The next-moment target current output by the control branch of the road sense execution system in the deep learning model (such as the steering intention prediction model) can be used as the direct control current value of the road sense motor to achieve real-time control of the road sense motor. It is understandable that the deep learning model (such as the steering intention prediction model) needs to be pre-trained using training samples. The training samples may include vehicle condition data collected by sensors. The vehicle condition data can be used to construct data feature maps to achieve deep learning model training. Using the deep learning model to learn the motor structure parameters can avoid system errors caused by parameter identification. Using the convolutional network to learn the system timing information can achieve the driver's steering intention prediction, and the system signal output is combined with the steering system's action compensation to reduce the frustration of the steering mechanism and steering wheel caused by repeated steering.

[0193] In this embodiment, multiple system signals are applied to the convolutional network to achieve deep interaction of multiple signals, and the model is trained to make the model learn the motor control parameters to avoid errors caused by system modeling; at the same time, the convolutional network can capture cross-variable dependencies, achieve efficient prediction of system timing information, accurately predict the driver's steering intention based on the current vehicle state information, environmental information and driver state information, and realize reasonable steering control strategy and road "sense control" strategy in combination with the current steering wheel steering signal captured by the system sensor. For example, the control of variables such as steering angle, steering rate, and steering change rate, so as to predict driving intention by deep learning, which can be combined with the steering system steering wheel steering signal to achieve closed loop and fusion of driving intention, which is conducive to alleviating the frustration of repeated steering of the steering wheel (upward turn) and achieving improved steering smoothness (downward turn), ensuring that the system can complete the driver's intended steering action and good operating experience, thereby accurately controlling the vehicle's wire control system based on the timing convolutional network and improving the accuracy of vehicle steering control.

[0194] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0195] Based on the same inventive concept, the embodiment of the present application also provides a vehicle steering control device for implementing the vehicle steering control method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more vehicle steering control device embodiments provided below can refer to the limitations of the vehicle steering control method above, and will not be repeated here.

[0196] In an exemplary embodiment, Figure 6 As shown, a vehicle steering control device is provided, including: an acquisition module 602 and a control module 604, wherein:

[0197] The acquisition module 602 is used to acquire vehicle environment information, vehicle state information and driver state information of the target vehicle when a steering instruction for the target vehicle is received.

[0198] The control module 604 is used to input the vehicle environment information, the vehicle state information and the driver state information into a pre-trained steering intention prediction model to obtain the target turning angle information and the turning angle velocity information of the target vehicle, and send the target turning angle information and the turning angle velocity information to the steering execution system; the steering execution system is used to activate the steering actuator corresponding to the target vehicle according to the target turning angle information and the turning angle velocity information to control the steering of the target vehicle.

[0199] In one of the exemplary embodiments, the control module 604 is specifically used to input the vehicle environment information, the vehicle state information and the driver state information into a pre-trained steering intention prediction model to obtain the predicted turning angle information and turning angle velocity information of the target vehicle; obtain the steering wheel angle information of the target vehicle from the vehicle state information, and obtain the target turning angle information of the target vehicle using the predicted turning angle information and the steering wheel angle information.

[0200] In one of the exemplary embodiments, the control module 604 is specifically used to use the predicted turning angle information and the steering wheel angle information to determine the weight information of the steering wheel angle information and the weight information of the predicted turning angle information; based on the weight information of the steering wheel angle information and the weight information of the predicted turning angle information, the steering wheel angle information and the predicted turning angle information are weighted and summed to obtain the target turning angle information of the target vehicle.

[0201] In one of the exemplary embodiments, the steering intention prediction model includes a first branch and a second branch; the control module 604 is specifically used to input the vehicle environment information, the vehicle state information and the driver state information into the steering intention prediction model, and obtain the predicted turning angle information and turning angle velocity information through the first branch.

[0202] In one of the exemplary embodiments, the steering intention prediction model includes a first branch and a second branch; the device also includes a road sense module, which is specifically used to obtain the target current information of the road sense motor of the target vehicle through the second branch; send the target current information of the road sense motor to the road sense execution system; the road sense execution system is used to control the road sense motor corresponding to the target vehicle according to the target current information of the road sense motor, so as to feedback the road conditions to the driver of the target vehicle.

[0203] In one of the exemplary embodiments, the device also includes a second branch training module, which is specifically used to obtain sample vehicle environment information, sample vehicle state information and sample driver state information of a sample vehicle; determine the vehicle environment coding information corresponding to the sample vehicle environment information, the vehicle state coding information corresponding to the sample vehicle state information and the driver state coding information corresponding to the sample driver state information based on the sample vehicle environment information, the sample vehicle state information and the sample driver state information; input the vehicle environment coding information, the vehicle state coding information and the driver state coding information into the steering intention prediction model to be trained to obtain the road sense motor current information of the sample vehicle; obtain the sample road sense motor current information of the sample vehicle, and train the second branch in the steering intention prediction model to be trained based on the road sense motor current information of the sample vehicle and the sample road sense motor current information.

[0204] In one of the exemplary embodiments, the device also includes a first branch training module, which is specifically used to obtain sample vehicle environment information, sample vehicle state information and sample driver state information of a sample vehicle; determine the vehicle environment coding information corresponding to the sample vehicle environment information, the vehicle state coding information corresponding to the sample vehicle state information and the driver state coding information corresponding to the sample driver state information based on the sample vehicle environment information, the sample vehicle state information and the sample driver state information; input the vehicle environment coding information, the vehicle state coding information and the driver state coding information into the steering intention prediction model to be trained to obtain the predicted turning angle information and predicted turning angle angular velocity information of the sample vehicle; obtain the sample turning angle information and the sample turning angle angular velocity information of the sample vehicle, and train the first branch in the steering intention prediction model to be trained based on the predicted turning angle information of the sample vehicle, the predicted turning angle angular velocity information of the sample vehicle, the sample turning information and the sample turning angle angular velocity information.

[0205] In one of the exemplary embodiments, the first branch training module is specifically used to obtain the quality information of the sample vehicle and the gender information and preference information of the corresponding driver of the sample vehicle; based on the quality information, the gender information and the preference information, determine the model structure and model parameters of the steering intention prediction model to be trained.

[0206] In one of the exemplary embodiments, the first branch training module is specifically used to determine the vehicle environment time series based on the sample vehicle environment information, determine the vehicle state time series based on the sample vehicle state information, and determine the driver state time series based on the sample driver state information; according to the preset sequence length information, the vehicle environment time series, the vehicle state time series and the driver state time series are divided respectively to obtain the sequence blocks corresponding to the vehicle environment time series, the sequence blocks corresponding to the vehicle state time series and the sequence blocks corresponding to the driver state time series; convolve the sequence blocks corresponding to the vehicle environment time series, the sequence blocks corresponding to the vehicle state time series and the sequence blocks corresponding to the driver state time series, and embed the convolution results into the vector corresponding to the sample vehicle environment information, the vector corresponding to the sample vehicle state information and the vector corresponding to the sample driver state information to obtain the vehicle environment coding information, the vehicle state coding information and the driver state coding information.

[0207] In one of the exemplary embodiments, the steering intention prediction model to be trained includes at least one temporal convolutional network, and the temporal convolutional network includes a depthwise separable convolutional layer, a pointwise convolutional layer and a normalization layer. The first branch training module is specifically used to perform convolution processing on the vehicle environment coding information, the vehicle state coding information and the driver state coding information through the depthwise separable convolutional layer in the temporal convolutional network according to the convolution kernel corresponding to the depthwise separable convolutional layer to obtain the output result of the depthwise separable convolutional layer; perform convolution processing on the output result of the depthwise separable convolutional layer through the pointwise convolutional layer in the temporal convolutional network according to the convolution kernel corresponding to the pointwise convolutional layer to obtain the output result of the pointwise convolutional layer; normalize the output result of the pointwise convolutional layer through the normalization layer in the temporal convolutional network to obtain the output result of the normalization layer; determine the predicted turning angle information and predicted turning angle angular velocity information of the sample vehicle according to the vehicle environment coding information, the vehicle state coding information, the driver state coding information and the output result of the normalization layer.

[0208] In one of the exemplary embodiments, the temporal convolutional network also includes a first convolutional layer and a second convolutional layer, and the first branch training module is specifically used to perform convolution processing on the output result of the depthwise separable convolutional layer through the convolution kernel corresponding to the first convolutional layer in the temporal convolutional network, so as to convolve the feature channels corresponding to the vehicle environment coding information, the vehicle state coding information and the driver state coding information, and obtain the output result of the first convolutional layer; perform convolution processing on the output result of the first convolutional layer through the convolution kernel corresponding to the second convolutional layer in the temporal convolutional network, so as to convolve the vehicle environment coding information, the vehicle state coding information and the driver state coding information, and obtain the output result of the second convolutional layer; determine the predicted turning angle information and predicted turning angle angular velocity information of the sample vehicle according to the output result of the second convolutional layer, the vehicle environment coding information, the vehicle state coding information and the driver state coding information.

[0209] In one of the exemplary embodiments, the acquisition module 602 is specifically used to acquire the road type information, visibility information, terrain information, vehicle quantity information, lane quantity information and traffic complexity information of the environment in which the target vehicle is located through the camera of the target vehicle, and determine the vehicle environment information of the target vehicle based on the road type information, the visibility information, the terrain information, the vehicle quantity information, the lane quantity information and the traffic complexity information; determine the vehicle status information of the target vehicle based on the vehicle speed information, acceleration information, yaw angular velocity information, wheel angle information, steering wheel angle information, temperature information and fault information of the target vehicle; and determine the driver status information of the target vehicle based on the behavior information of the driver corresponding to the target vehicle.

[0210] Each module in the above vehicle steering control device can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute operations corresponding to each module.

[0211] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 7 As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface, the display unit and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a vehicle steering control method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0212] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0213] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0214] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0215] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0216] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0217] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0218] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0219] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A vehicle steering control method, characterized in that: The method comprises: Upon receiving a steering instruction for a target vehicle, obtaining vehicle environment information, vehicle state information, and driver state information of the target vehicle; The vehicle environment information, the vehicle state information and the driver state information are input into a pre-trained steering intention prediction model to obtain the target turning angle information and turning angle velocity information of the target vehicle, and the target turning angle information and the turning angle velocity information are sent to a steering execution system; the steering execution system is used to activate the steering actuator corresponding to the target vehicle according to the target turning angle information and the turning angle velocity information to control the steering of the target vehicle.

2. The method according to claim 1, characterized in that The vehicle environment information, the vehicle state information and the driver state information are input into a pre-trained steering intention prediction model to obtain target turning angle information and turning angle velocity information of the target vehicle, including: Inputting the vehicle environment information, the vehicle state information and the driver state information into a pre-trained steering intention prediction model to obtain predicted turning angle information and turning angle velocity information of the target vehicle; The steering wheel angle information of the target vehicle is acquired from the vehicle state information, and the target angle information of the target vehicle is obtained by using the predicted angle information and the steering wheel angle information.

3. The method according to claim 2, characterized in that Obtaining target turning angle information of the target vehicle using the predicted turning angle information and the steering wheel turning angle information includes: Determining weight information of the steering wheel angle information and weight information of the predicted angle information by using the predicted angle information and the steering wheel angle information; According to the weight information of the steering wheel angle information and the weight information of the predicted angle information, the steering wheel angle information and the predicted angle information are weighted and summed to obtain the target angle information of the target vehicle.

4. The method according to claim 2, characterized in that: The turning intention prediction model includes a first branch and a second branch; The vehicle environment information, the vehicle state information and the driver state information are input into a pre-trained steering intention prediction model to obtain the predicted turning angle information and turning angle velocity information of the target vehicle, including: Inputting the vehicle environment information, the vehicle state information and the driver state information into the steering intention prediction model, and obtaining the predicted turning angle information and turning angle velocity information through the first branch; The method further comprises: Obtaining target current information of the road sensor motor of the target vehicle through the second branch; The road sense motor target current information is sent to a road sense execution system; the road sense execution system is used to control the road sense motor corresponding to the target vehicle according to the road sense motor target current information, so as to feedback road conditions to the driver of the target vehicle.

5. The method according to claim 2, characterized in that: Before acquiring the vehicle environment information, vehicle state information and driver state information of the target vehicle, the method further includes: Obtaining sample vehicle environment information, sample vehicle state information, and sample driver state information of a sample vehicle; Determine, based on the sample vehicle environment information, the sample vehicle state information and the sample driver state information, vehicle environment coding information corresponding to the sample vehicle environment information, vehicle state coding information corresponding to the sample vehicle state information and driver state coding information corresponding to the sample driver state information; Inputting the vehicle environment coding information, the vehicle state coding information and the driver state coding information into the steering intention prediction model to be trained to obtain the road sense motor current information of the sample vehicle; The sample road sense motor current information of the sample vehicle is obtained, and the second branch in the steering intention prediction model to be trained is trained according to the road sense motor current information of the sample vehicle and the sample road sense motor current information.

6. The method according to claim 2, characterized in that Before acquiring the vehicle environment information, vehicle state information and driver state information of the target vehicle, the method further includes: Obtaining sample vehicle environment information, sample vehicle state information, and sample driver state information of a sample vehicle; Determine, based on the sample vehicle environment information, the sample vehicle state information and the sample driver state information, vehicle environment coding information corresponding to the sample vehicle environment information, vehicle state coding information corresponding to the sample vehicle state information and driver state coding information corresponding to the sample driver state information; Inputting the vehicle environment coding information, the vehicle state coding information and the driver state coding information into the steering intention prediction model to be trained to obtain the predicted turning angle information and the predicted turning angle angular velocity information of the sample vehicle; Obtain sample turning angle information and sample turning angle angular velocity information of the sample vehicle, and train a first branch in the steering intention prediction model to be trained based on the predicted turning angle information of the sample vehicle, the predicted turning angle angular velocity information of the sample vehicle, the sample turning angle information and the sample turning angle angular velocity information.

7. The method according to claim 6, characterized in that After obtaining the sample turning angle information and the sample turning angle angular velocity information of the sample vehicle, the method further includes: Acquire quality information of the sample vehicle and gender information and preference information of the driver corresponding to the sample vehicle; The model structure and model parameters of the turning intention prediction model to be trained are determined according to the quality information, the gender information and the preference information.

8. The method according to claim 6, characterized in that Determining, according to the sample vehicle environment information, the sample vehicle state information, and the sample driver state information, vehicle environment coding information corresponding to the sample vehicle environment information, vehicle state coding information corresponding to the sample vehicle state information, and driver state coding information corresponding to the sample driver state information, including: Determining a vehicle environment time series according to the sample vehicle environment information, determining a vehicle state time series according to the sample vehicle state information, and determining a driver state time series according to the sample driver state information; According to preset sequence length information, the vehicle environment time series, the vehicle state time series and the driver state time series are divided respectively to obtain sequence blocks corresponding to the vehicle environment time series, sequence blocks corresponding to the vehicle state time series and sequence blocks corresponding to the driver state time series; Convolution is performed on the sequence blocks corresponding to the vehicle environment time series, the sequence blocks corresponding to the vehicle state time series, and the sequence blocks corresponding to the driver state time series, and the convolution results are embedded into the vector corresponding to the sample vehicle environment information, the vector corresponding to the sample vehicle state information, and the vector corresponding to the sample driver state information to obtain the vehicle environment coding information, the vehicle state coding information, and the driver state coding information.

9. The method according to claim 8, characterized in that The steering intention prediction model to be trained includes at least one temporal convolutional network, and the temporal convolutional network includes a depthwise separable convolutional layer, a pointwise convolutional layer, and a normalization layer. The vehicle environment encoding information, the vehicle state encoding information, and the driver state encoding information are input into the steering intention prediction model to be trained to obtain the predicted turning angle information and the predicted turning angle angular velocity information of the sample vehicle, including: Through the depthwise separable convolutional layer in the temporal convolutional network, the vehicle environment coding information, the vehicle state coding information and the driver state coding information are convolved according to the convolution kernel corresponding to the depthwise separable convolutional layer to obtain an output result of the depthwise separable convolutional layer; By using the point-by-point convolution layer in the temporal convolutional network, the output result of the depthwise separable convolution layer is convolved according to the convolution kernel corresponding to the point-by-point convolution layer to obtain the output result of the point-by-point convolution layer; Normalizing the output result of the point-by-point convolution layer through the normalization layer in the temporal convolution network to obtain the output result of the normalization layer; The predicted turning angle information and predicted turning angle angular velocity information of the sample vehicle are determined according to the vehicle environment coding information, the vehicle state coding information, the driver state coding information and the output result of the normalization layer.

10. The method according to claim 9, characterized in that The temporal convolutional network further includes a first convolutional layer and a second convolutional layer, a convolutional kernel corresponding to the first convolutional layer is different from a convolutional kernel corresponding to the second convolutional layer, and after obtaining an output result of the depthwise separable convolutional layer, the method further includes: Performing convolution processing on the output result of the depthwise separable convolution layer through the convolution kernel corresponding to the first convolution layer in the temporal convolution network, so as to convolve the feature channels corresponding to the vehicle environment encoding information, the vehicle state encoding information and the driver state encoding information, respectively, to obtain the output result of the first convolution layer; Performing convolution processing on the output result of the first convolution layer through the convolution kernel corresponding to the second convolution layer in the temporal convolution network, so as to convolve the vehicle environment encoding information, the vehicle state encoding information and the driver state encoding information, and obtain the output result of the second convolution layer; The predicted turning angle information and predicted turning angle angular velocity information of the sample vehicle are determined according to the output result of the second convolutional layer, the vehicle environment encoding information, the vehicle state encoding information and the driver state encoding information.

11. The method according to claim 1, characterized in that: Acquiring vehicle environment information, vehicle status information, and driver status information of the target vehicle, including: Acquire road type information, visibility information, terrain information, vehicle quantity information, lane quantity information, and traffic complexity information of the environment in which the target vehicle is located through a camera of the target vehicle, and determine vehicle environment information of the target vehicle based on the road type information, visibility information, terrain information, vehicle quantity information, lane quantity information, and traffic complexity information; Determining vehicle state information of the target vehicle according to vehicle speed information, acceleration information, yaw angular velocity information, wheel angle information, steering wheel angle information, temperature information and fault information of the target vehicle; The driver status information of the target vehicle is determined according to the behavior information of the driver corresponding to the target vehicle.

12. A vehicle steering control device, characterized in that: The device comprises: An acquisition module, configured to acquire vehicle environment information, vehicle state information, and driver state information of the target vehicle upon receiving a steering instruction for the target vehicle; A control module is used to input the vehicle environment information, the vehicle state information and the driver state information into a pre-trained steering intention prediction model to obtain the target turning angle information and the turning angle angular velocity information of the target vehicle, and send the target turning angle information and the turning angle angular velocity information to a steering execution system; the steering execution system is used to activate the steering actuator corresponding to the target vehicle according to the target turning angle information and the turning angle angular velocity information to control the steering of the target vehicle.

Images