A Vehicle Driving Style Adaptation Method Based on Online Learning

Abstract

The present invention belongs to the technical field of artificial intelligence, and discloses a method for adapting vehicle driving styles based on online learning. The method includes: during the driving process of the vehicle, obtaining vehicle driving data at a plurality of consecutive moments including the current moment; inputting the vehicle driving data into a driving style detection model obtained through online training to obtain the driving style of the driver at the next moment; performing corresponding control on the vehicle based on the driving style; the online training process of the driving style detection model includes: predicting the predicted value of the vehicle driving data at the next moment based on the vehicle driving data, generating an online feedback strategy based on the error between the vehicle driving data and the predicted value of the vehicle driving data, and performing online feedback training on the driving style detection model based on the online feedback strategy. The present invention can dynamically predict the driving style of the driver, and perform adaptive adjustment on the vehicle based on the predicted driving style, achieving the purpose of "one vehicle, one face and one person, one face".

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence, and particularly relates to a method for adapting vehicle driving styles based on online learning. Background Art

[0002] An advanced driver assistance system is a set of built-in coordination technologies that help drivers improve the safety of vehicle driving. By sensing, making decisions, and executing, it helps drivers detect potential dangers. It effectively combines driving and braking, reducing the driving pressure of drivers to a certain extent and greatly reducing the risk of traffic accidents caused by improper driver operation.

[0003] In related technologies, the driving styles of drivers are usually divided into aggressive, ordinary, and cautious types, etc., and overall, it is mainly "people adapting to the vehicle". Specifically, in related technologies, an offline learning method is used to pre-train a driving style prediction model, and the driving style of the driver is predicted through this driving style prediction model, and the predicted driving style is fixed as the driving style of this driver. For example, if the driving style of a driver is predicted to be aggressive, then the driving style of this driver is fixed as aggressive, and the vehicle is controlled according to the vehicle control method corresponding to the aggressive type. In fact, drivers usually do not always maintain an aggressive driving style, thus affecting the driving experience of drivers. Summary of the Invention

[0004] The purpose of the present invention is to dynamically predict the driving style of a driver, rather than the driving style being fixed as in related technologies. Moreover, the present invention can adaptively adjust the vehicle based on the predicted driving style, achieving the purpose of "a thousand vehicles with a thousand faces and a thousand people with a thousand faces". At the same time, the driving style detection model can be trained with online feedback to dynamically adjust the model parameters of the driving style detection model and optimize the accuracy and response speed of the overall system.

[0005] In a first aspect, an embodiment of the present invention provides a method for adapting vehicle driving styles based on online learning, and the method includes:

[0006] During the driving process of the vehicle, obtain the vehicle driving data at a plurality of consecutive moments including the current moment;

[0007] Input the vehicle driving data into a driving style detection model obtained by online training to obtain the driving style of the driver at the next moment;

[0008] Control the vehicle accordingly based on the driving style;

[0009] Among them, the online training process of the driving style detection model includes: predicting the predicted value of the vehicle driving data at the next moment based on the vehicle driving data, generating an online feedback strategy based on the error between the vehicle driving data and the predicted value of the vehicle driving data, and performing online feedback training on the driving style detection model based on the online feedback strategy.

[0010] Optionally, the process of the driving style detection model predicting the driving style of the driver at the next moment includes:

[0011] Generating vehicle driving time-series data based on the vehicle driving data, where the vehicle driving time-series data is used to characterize the time-series features and context association information of the vehicle driving data;

[0012] Extracting key time-series data in the vehicle driving time-series data and the attention weights of each piece of key time-series data based on the attention mechanism;

[0013] Converting the key time-series data into target time-series data in a time-series format;

[0014] Predicting the driving style of the driver at the next moment based on the target time-series data and the attention weights corresponding to each piece of target time-series data.

[0015] Optionally, the driving style detection model includes a time-series encoder, a hidden space attention mechanism module including a self-attention mechanism and a cross-attention mechanism, and a time-series decoder;

[0016] The generating vehicle driving time-series data based on the vehicle driving data includes:

[0017] Encoding the vehicle driving data into vehicle driving time-series data through the time-series encoder;

[0018] Correspondingly, the extracting key time-series data in the vehicle driving time-series data and the attention weights of each piece of key time-series data based on the attention mechanism includes:

[0019] Processing the vehicle driving time-series data through the hidden space attention mechanism module to obtain the key time-series data in the vehicle driving time-series data and the attention weights of each piece of key time-series data;

[0020] Correspondingly, the converting the key time-series data into target time-series data in a time-series format includes: converting the key time-series data into target time-series data in a time-series format through the time-series decoder.

[0021] Optionally, the driving style detection model further includes a pedal and steering prediction module and an online gradient network;

[0022] Predicting the predicted value of the vehicle driving data at the next moment based on the vehicle driving data, generating an online feedback strategy based on the error between the vehicle driving data and the predicted value of the vehicle driving data, and performing online feedback training on the driving style detection model based on the online feedback strategy to obtain an online trained driving style detection model, including:

[0023] Predicting the predicted value of the pedal data and the predicted value of the steering data at the next moment through the pedal and steering prediction module based on the target time series data and the attention weights corresponding to each piece of target time series data;

[0024] Calculating the error between the pedal data and the steering data included in the vehicle driving data and the predicted value of the pedal data and the predicted value of the steering data through the online gradient network, determining the online feedback strategy based on the error, and feeding back the online feedback strategy to the hidden space attention mechanism module;

[0025] The hidden space attention mechanism module performs online training based on the online feedback strategy.

[0026] Optionally, the obtaining of the vehicle driving data for consecutive multiple moments including the current moment includes:

[0027] Obtaining the initial vehicle driving data for consecutive multiple moments including the current moment, where the initial vehicle driving data includes the pedal data, steering data, driving speed, and yaw speed of the vehicle;

[0028] Preprocessing the initial vehicle driving data to obtain preprocessed vehicle driving data, where the preprocessing includes data fusion and data denoising.

[0029] Optionally, the driving style includes multiple types, and the inputting the vehicle driving data into the driving style detection model obtained through online training to obtain the driving style of the driver at the next moment includes:

[0030] Inputting the vehicle driving data into the driving style detection model obtained through online training, and outputting multiple types of driving styles and the probability corresponding to each type of driving style through the driving style detection model;

[0031] Calculating the driving style of the driver at the next moment based on the multiple types of driving styles and the probability corresponding to each type of driving style.

[0032] Optionally, the correspondingly controlling the vehicle based on the driving style includes:

[0033] Receiving the torque request input by the driver through the accelerator pedal;

[0034] Obtain the torque change curves corresponding to various types of driving styles;

[0035] Based on the probabilities corresponding to various types of driving styles, perform weighted summation on the torque change curves corresponding to various types of driving styles to obtain a weighted-sum torque change curve;

[0036] Obtain the output torque through the weighted-sum torque change curve, and control the vehicle with the output torque.

[0037] In a second aspect, an embodiment of the present invention provides a vehicle driving style adaptive device based on online learning. The device includes:

[0038] A vehicle driving data acquisition module, configured to acquire vehicle driving data at a plurality of consecutive moments including the current moment during vehicle driving;

[0039] A driving style detection module, configured to input the vehicle driving data into a driving style detection model obtained through online training to obtain the driver's driving style at the next moment;

[0040] A vehicle control module, configured to perform corresponding control on the vehicle based on the driving style;

[0041] Wherein, the online training process of the driving style detection model includes: predicting the predicted value of the vehicle driving data at the next moment based on the vehicle driving data, generating an online feedback strategy based on the error between the vehicle driving data and the predicted value of the vehicle driving data, and performing online feedback training on the driving style detection model based on the online feedback strategy.

[0042] In a third aspect, an embodiment of the present invention provides an electronic device, including:

[0043] At least one processor;

[0044] A memory for storing instructions executable by the at least one processor;

[0045] Wherein, the at least one processor is configured to execute the instructions to implement the method described in the first aspect.

[0046] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium. When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can execute the method described in the first aspect.

[0047] After obtaining the vehicle driving data, the technical solution provided by the embodiments of the present invention can input the vehicle driving data into the driving style detection model obtained by online training to obtain the driving style of the driver at the next moment, and perform corresponding control on the vehicle based on the driving style. In this way, the predicted driving style can meet the driver's current driving state, and as the continuous data stream is input, the predicted driving style will also change, unlike the related art where the driving style remains unchanged. Moreover, the present invention can perform adaptive adjustment on the vehicle based on the predicted driving style, achieving the purpose of "one vehicle, one face and one person, one face", ensuring that the response of the vehicle matches the driver's intention and the driving conditions at that time, and improving the driver's driving experience.

[0048] Furthermore, the present invention determines the online feedback strategy based on the error between the real vehicle driving data and the predicted vehicle driving data, and performs online feedback training on the driving style detection model to dynamically adjust the model parameters of the driving style detection model, optimizing the accuracy and response speed of the overall system. That is, the output driving style is more accurate, and corresponding control is performed on the vehicle based on the driver's driving style, improving the safety of the vehicle, further ensuring that the response of the vehicle matches the driver's intention and the driving conditions at that time, and further improving the driver's driving experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the overall technical solution architecture provided by the embodiments of the present invention;

[0050] Figure 2 It is a schematic diagram of a specific implementation manner provided by the embodiments of the present invention;

[0051] Figure 3 It is a flowchart of a vehicle driving style self-adaptive method based on online learning provided by the embodiments of the present invention;

[0052] Figure 4 For Figure 3 It is a flowchart of an implementation manner of S330 in

[0053] Figure 5 It is a schematic diagram of the structure of a vehicle driving style self-adaptive device based on online learning provided by the embodiments of the present invention;

[0054] Figure 6 It is a schematic diagram of the structure of an electronic device provided by the embodiments of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0055] The present invention will be described in detail below through embodiments.

[0056] Advanced Driver Assistance Systems (ADAS) are a set of built-in coordinated technologies that help drivers improve vehicle driving safety. By sensing, making decisions, and executing, they assist drivers in detecting potential dangers. It effectively combines driving and braking, reducing the driver's driving stress to a certain extent and significantly lowering the risk of traffic accidents caused by improper driver operation.

[0057] In related technologies, a driver's driving style is usually classified into aggressive, ordinary, and cautious types, etc., and overall it is mainly "human adapting to the vehicle". However, "each vehicle is unique" and "each person is unique". The purpose of the embodiments of the present invention is to achieve "the vehicle adapting to the person" to accommodate various driving styles. Different drivers have different driving styles, and many factors can affect a driver's driving behavior. Therefore, driver adaptive technology still needs further research to achieve smooth transitions between different vehicle modes, so as to adapt to different driving styles.

[0058] At present, the method of applying end-to-end large models is mostly used to improve driver adaptive technology. Through large-scale data training, control instructions can be directly generated from sensor data, avoiding delays and error accumulation in intermediate links and improving the overall performance of the system. However, the problem with this method is that the implementation complexity is high. It is necessary to split the model and coordinate the calculation results, making the model more "black box", reducing the interpretability of the network, and it is difficult to determine the contribution of each component in the model to the final goal, resulting in a decrease in model flexibility.

[0059] The present invention aims to adaptively adjust the vehicle based on the driver's driving style. Specifically, by introducing data such as the drive pedal, brake pedal, and steering during the driver's driving process to build a driving style detection model, which can also be called a driving style detection model; predicting the driver's driving style at the next moment through the driving style detection model, and controlling the vehicle to smoothly switch between a sports mode, an economy mode, and a comfort mode in real time according to the driving style. The present invention performs online learning through a local feedback network, trains and optimizes the detection model in real time, and updates the parameters during vehicle driving, steering, and braking.

[0060] As Figure 1 shown, it is a schematic diagram of the overall technical solution architecture provided by the embodiments of the present invention.

[0061] From Figure 1 it can be seen that the embodiments of the present invention use feature information such as the vehicle's pedal drive data, pedal brake data, steering data, vehicle speed, and vehicle yaw rate (which can be called vehicle driving data) as inputs. It should be emphasized that the input vehicle driving data is the vehicle driving data of multiple consecutive moments including the current moment, that is, the input vehicle driving data is a continuous data stream.

[0062] Artificial intelligence is used to detect the driving behavior of a driver and output a driving mode based on the driving behavior. This driving mode can be called a driving style, which can include cautiousness and aggressiveness, etc. Of course, in practical applications, other types of driving styles can also be included, and no examples will be given one by one here.

[0063] Among them, the dashed box represents the operations performed by the driving style detection model. The driving style prediction model predicts the driving behavior based on the above vehicle driving data, obtains the driving mode based on the driving behavior, outputs various driving modes from the driving style detection model, and the probability corresponding to each driving mode. Finally, the vehicle control system can control the vehicle based on the driving mode. Among them, the vehicle control system can include the vehicle's overall vehicle controller, electric power steering system, electro-hydraulic braking system, intelligent driving system, advanced driver assistance system, etc.

[0064] The technical solution provided by the embodiments of the present invention, when predicting the driving style of the driver at the next moment, first obtains the vehicle driving data of a continuous plurality of moments including the current moment, that is, obtains a continuous video stream. In this way, the predicted driving style can meet the driver's current driving state, and as the input continuous data stream changes, the predicted driving style will also change, unlike the related technologies where the driving style remains unchanged. Moreover, the present invention can perform adaptive adjustment on the vehicle based on the predicted driving style, achieving the purpose of "thousands of vehicles with thousands of faces and thousands of people with thousands of faces", ensuring that the vehicle's response matches the driver's intention and the current driving conditions, and improving the driver's driving experience.

[0065] As Figure 2 shown, it is a schematic diagram of a specific implementation manner provided by the embodiments of the present invention.

[0066] From Figure 2 it can be seen that first, the pedal drive data, pedal brake data, and steering data of a continuous plurality of moments including the current moment are obtained. The pedal drive data and pedal brake data of the vehicle are fused through the pedal fusion module to obtain the fused pedal data, and the steering data of the vehicle is filtered through the filter to obtain the filtered steering data.

[0067] The fused pedal data and the filtered steering data are input into the time series encoder, and the time series encoder generates vehicle driving time series data. Among them, the role of the time series encoder is to capture the change rules of the pedal data and steering data of a continuous plurality of moments over time, and convert the pedal data and steering data of a continuous plurality of moments into data that can be received by the hidden space attention mechanism.

[0068] Among them, the vehicle driving time-series data is not in a simple sequence form in terms of format, but a high-dimensional feature sequence. For example, it can be a feature vector, an embedding vector, an attention weight, etc. These representations can capture the time-series features and context information in the vehicle driving data. The feature vector, embedding vector, and attention weight will be explained separately below.

[0069] 1. Feature vector. The time-series encoder converts the input continuous data stream into a set of feature vectors, which can represent the key features of the original data.

[0070] 2. Embedding vector. Among them, the embedding vector is a method of mapping discrete data into a continuous vector space, which can capture the semantics and relationships of discrete data.

[0071] 3. Attention weight: It represents the degree of attention to different time steps when the model processes the sequence.

[0072] After obtaining the vehicle driving time-series data through the time-series encoder, the vehicle driving time-series data can be input into the hidden space attention mechanism module that includes a self-attention mechanism and a cross-attention mechanism. The hidden space attention mechanism module will filter out the key time-series data in the vehicle driving time-series data, as well as the attention weights corresponding to the critical moment data at each moment. This allows the model to establish connections between different parts of the time-series data sequence, thereby better understanding the context of the time-series data.

[0073] The time-series decoder converts the key time-series data processed by the hidden space attention mechanism module back into the time-series format for predicting the driving manner of the driver at the next moment, that is, predicting the driving style of the driver at the next moment.

[0074] From Figure 2 It can be seen that the pedal prediction module outputs the pedal prediction value and the steering prediction value at the next moment based on the data output by the hidden space attention mechanism module. Among them, the pedal prediction value includes the pedal drive data prediction value, the pedal brake data prediction value, and the steering data prediction value, that is, predicting whether the driver will accelerate or decelerate at the next moment, and whether to turn.

[0075] The core function of the feedback mechanism in the embodiment of the present invention is to predict the future operation behavior of the driver (including the above-mentioned pedal drive data prediction value, pedal brake data prediction value, and steering data prediction value), and transmit these prediction results to the online gradient network of the system to optimize the overall performance.

[0076] The input data of the online gradient network includes initial pedal drive data, pedal brake data, and steering data (data that has undergone data fusion and filtering); the input data of the online gradient network also includes the predicted values of the pedal drive data, pedal brake data, and steering data output by the pedal prediction module.

[0077] The online gradient network calculates the error between the initial pedal drive data and the predicted value of the pedal drive data, the error between the initial pedal brake data and the predicted value of the pedal brake data, and the error between the initial steering data and the predicted value of the steering data, and determines an online feedback strategy based on these three errors, and transmits the online feedback strategy to the hidden space attention mechanism module for dynamically adjusting the parameters (such as attention weights) of the hidden space attention mechanism module to optimize the accuracy and response speed of the overall system.

[0078] In the embodiment of the present invention, by using the online gradient network to calculate the error between the real pedal drive data of the initial input and the predicted pedal drive data, the real pedal brake data and the predicted pedal brake data, and the real steering data and the predicted steering data, an online feedback strategy can be determined, and online feedback training can be performed on the hidden space attention mechanism module to dynamically adjust the parameters (such as attention weights) of the hidden space attention mechanism module to optimize the accuracy and response speed of the overall system. That is, the output driving style is more accurate, and the vehicle is controlled accordingly based on the driver's driving style, improving the safety of the vehicle, ensuring that the response of the vehicle matches the driver's intention and the current driving conditions, and improving the driver's driving experience.

[0079] Next, a vehicle driving style adaptive method based on online learning provided by the embodiment of the present invention will be elaborated in detail. As Figure 3 shown, a vehicle driving style adaptive method based on online learning provided by the embodiment of the present invention may include the following steps, respectively:

[0080] S310, during the vehicle driving process, obtain the vehicle driving data of consecutive multiple moments including the current moment.

[0081] Specifically, in order to be able to adaptively predict the driver's driving style at the next moment according to the user's current driving state, during the vehicle driving process, the vehicle driving data of consecutive multiple moments including the current moment can be obtained. Among them, the vehicle driving data may include the pedal drive data, pedal brake data, steering data, vehicle speed, and vehicle yaw rate of the vehicle, etc., and the embodiment of the present invention does not make specific limitations on this.

[0082] As an implementation manner of the embodiment of the present invention, obtaining the vehicle driving data of consecutive multiple moments including the current moment may include the following steps:

[0083] Obtain initial vehicle driving data for a continuous number of moments including the current moment, where the initial vehicle driving data includes the pedal data, steering data, driving speed, and yaw rate of the vehicle.

[0084] Preprocess the initial vehicle driving data to obtain preprocessed vehicle driving data, where the preprocessing includes data fusion and data denoising.

[0085] In this implementation manner, after obtaining the initial vehicle driving data, preprocessing operations such as data fusion and denoising can be performed on the initial vehicle driving data. In this way, after inputting the vehicle driving data into the driving style detection model, it helps the driving style detection model to input a driving style with a relatively high accuracy.

[0086] S320, Input the vehicle driving data into the driving style detection model obtained by online training to obtain the driving style of the driver at the next moment.

[0087] S330, Control the vehicle accordingly based on the driving style.

[0088] Among them, the online training process of the driving style detection model includes: predicting the predicted value of the vehicle driving data at the next moment based on the vehicle driving data, generating an online feedback strategy based on the error between the vehicle driving data and the predicted value of the vehicle driving data, and performing online feedback training on the driving style detection model based on the online feedback strategy.

[0089] Specifically, after obtaining the vehicle driving data, the vehicle driving data can be input into the driving style detection model obtained by online training to obtain the driving style of the driver at the next moment, and the vehicle can be controlled accordingly based on the driving style. In this way, the predicted driving style can meet the current driving state of the driver, and as the continuous data stream input changes, the predicted driving style will also change, unlike in the related art where the driving style is fixed. Moreover, the present invention can perform adaptive adjustment on the vehicle based on the predicted driving style, achieving the purpose of "thousands of vehicles with thousands of faces and thousands of people with thousands of faces", ensuring that the response of the vehicle matches the driver's intention and the current driving conditions, and improving the driver's driving experience.

[0090] Moreover, the present invention determines an online feedback strategy based on the error between the real vehicle driving data and the predicted vehicle driving data, and performs online feedback training on the driving style detection model to dynamically adjust the model parameters of the driving style detection model, optimizing the accuracy and response speed of the overall system. That is, the output driving style is more accurate, and the vehicle is controlled accordingly based on the driver's driving style, improving the safety of the vehicle and further ensuring that the response of the vehicle matches the driver's intention and the current driving conditions, further enhancing the driver's driving experience.

[0091] Based on Figure 3 the embodiment shown, in one implementation manner, the process of the driving style detection model predicting the driver's driving style at the next moment may include the following four steps, which are respectively:

[0092] Step 1: Generate vehicle driving time series data based on the vehicle driving data.

[0093] Among them, the vehicle driving time series data is used to characterize the time series features and context correlation information of the vehicle driving data.

[0094] Step 2: Extract the key time series data in the vehicle driving time series data and the attention weights of each piece of key time series data based on the attention mechanism.

[0095] Step 3: Convert the key time series data into target time series data in time series format.

[0096] Step 4: Predict the driver's driving style at the next moment based on the target time series data and the attention weights corresponding to each piece of target time series data.

[0097] In practical applications, the driving style detection model may include a time series encoder, a hidden space attention mechanism module including a self-attention mechanism and a cross-attention mechanism, and a time series decoder.

[0098] At this time, Step 1: Generate vehicle driving time series data based on the vehicle driving data may include the following steps:

[0099] Step 11: Encode the vehicle driving data into vehicle driving time series data through the time series encoder;

[0100] Correspondingly, Step 2: Extract the key time series data in the vehicle driving time series data and the attention weights of each piece of key time series data based on the attention mechanism may include the following steps:

[0101] Step 21: Process the vehicle driving time series data through the hidden space attention mechanism module to obtain the key time series data in the vehicle driving time series data and the attention weights of each piece of key time series data;

[0102] Correspondingly, in step 3, converting the key timing data into target timing data in a time series format may include the following steps:

[0103] Step 31: Convert the key timing data into target timing data in a time series format through a time series decoder.

[0104] Since steps 1 to 4, as well as steps 11, 21, and 31 have been elaborated in detail in the above embodiments, they will not be repeated in the embodiments of the present invention.

[0105] Based on the above embodiments, the driving style detection model further includes a pedal and steering prediction module and an online gradient network.

[0106] At this time, predicting the predicted value of the vehicle driving data at the next moment based on the vehicle driving data, generating an online feedback strategy based on the error between the vehicle driving data and the predicted value of the vehicle driving data, and performing online feedback training on the driving style detection model based on the online feedback strategy to obtain the online trained driving style detection model may include the following steps:

[0107] Predict the predicted value of the pedal data and the predicted value of the steering data at the next moment through the pedal and steering prediction module based on the target timing data and the attention weights corresponding to each piece of target timing data;

[0108] Calculate the error between the pedal data and the steering data included in the vehicle driving data and the predicted value of the pedal data and the predicted value of the steering data through the online gradient network, determine the online feedback strategy based on the error, and feedback the online feedback strategy to the hidden space attention mechanism module;

[0109] The hidden space attention mechanism module performs online training based on the online feedback strategy.

[0110] The above steps are the online feedback training process of the driving style detection model, which has been elaborated in the Figure 2 illustrated embodiments and will not be repeated here.

[0111] Based on the above embodiments, in one implementation, the driving style may include multiple types. Inputting the vehicle driving data into the driving style detection model obtained through online training to obtain the driving style of the driver at the next moment may include the following steps:

[0112] Input the vehicle driving data into the driving style detection model obtained through online training, and output multiple types of driving styles and the probability corresponding to each type of driving style through the driving style detection model.

[0113] Calculate the driver's driving style at the next moment based on multiple types of driving styles and the probabilities corresponding to each type of driving style.

[0114] Specifically, there can be multiple driving styles output by the driving style detection model. For example, they can be aggressive, economical, cautious, etc., and the probabilities corresponding to each driving style can be output. For example, aggressive 20%, economical 50%, cautious 30%, and so on. Then, perform a weighted sum of the multiple types of driving styles and the corresponding probabilities to obtain the driver's driving style at the next moment.

[0115] As an implementation manner of the embodiment of the present invention, S330, perform corresponding control on the vehicle based on the driving style, such as Figure 4 shown, which can include the following steps:

[0116] S331, receive the torque request input by the driver through the accelerator pedal.

[0117] S332, obtain the torque change curves corresponding to various types of driving styles.

[0118] S333, based on the probabilities corresponding to various types of driving styles, perform a weighted sum of the torque change curves corresponding to various types of driving styles to obtain a weighted sum torque change curve.

[0119] S334, obtain the output torque through the weighted sum torque change curve, and control the vehicle through the output torque.

[0120] Specifically, after the vehicle control system receives the torque request input by the driver by stepping on the accelerator, it can obtain the torque change curves corresponding to various types of driving styles. It can be understood that the torque change curves corresponding to different types of driving styles are different. The curvature of the torque change curve of an aggressive driving style is larger, and the curvature of the torque change curve of a cautious driving style is smaller.

[0121] Since the driving style detection model outputs multiple types of driving styles and the probabilities corresponding to each type of driving style, in order to obtain the torque change curve of the driving style at the next moment, a weighted sum of the torque change curves corresponding to various types of driving styles can be performed based on the probabilities corresponding to various types of driving styles to obtain a weighted sum torque change curve. This weighted sum torque change curve is the torque change curve of the driving style at the next moment. Finally, obtain the output torque through the weighted sum torque change curve, and control the vehicle through the output torque.

[0122] In the embodiments of the present invention, the predicted driving style at the next moment is not fixed, but changes with the continuous vehicle driving data stream input. Moreover, the present invention can perform adaptive adjustment on the vehicle based on the predicted driving style, achieving the purpose of "a thousand vehicles with a thousand faces and a thousand people with a thousand faces", ensuring that the response of the vehicle matches the driver's intention and the current driving conditions, and improving the driver's driving experience.

[0123] In a second aspect, an embodiment of the present invention provides a vehicle driving style adaptive device 50 based on online learning, as Figure 5 shown, the device includes:

[0124] A vehicle driving data acquisition module 510, configured to acquire vehicle driving data at a plurality of consecutive moments including the current moment during vehicle driving;

[0125] A driving style detection module 520, configured to input the vehicle driving data into a driving style detection model obtained by online training to obtain the driver's driving style at the next moment;

[0126] A vehicle control module 530, configured to perform corresponding control on the vehicle based on the driving style;

[0127] Wherein, the online training process of the driving style detection model includes: predicting a predicted value of the vehicle driving data at the next moment based on the vehicle driving data, generating an online feedback strategy based on the error between the vehicle driving data and the predicted value of the vehicle driving data, and performing online feedback training on the driving style detection model based on the online feedback strategy.

[0128] In a third aspect, an embodiment of the present invention provides an electronic device 600, as Figure 6 shown, including:

[0129] At least one processor 601;

[0130] A memory 602 for storing executable instructions of the at least one processor;

[0131] Wherein, the at least one processor is configured to execute the instructions to implement the method described in the first aspect.

[0132] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, when the instructions in the computer-readable storage medium are executed by a processor of an electronic device, enabling the electronic device to execute the method described in the first aspect.

[0133] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention without departing from the principles and spirit of the present invention.

Claims

1. A vehicle driving style adaptation method based on online learning, characterized in that: The method comprises: During the driving process of the vehicle, obtaining the vehicle driving data at a plurality of consecutive moments including the current moment; Inputting the vehicle driving data into a driving style detection model obtained through online training to obtain the driver's driving style at the next moment; controlling the vehicle accordingly based on the driving style; The online training process of the driving style detection model includes: predicting a predicted value of vehicle driving data at the next moment based on the vehicle driving data, generating an online feedback strategy based on an error between the vehicle driving data and the predicted value of the vehicle driving data, and performing online feedback training on the driving style detection model based on the online feedback strategy; The process of the driving style detection model predicting the driver's driving style at the next moment includes: Generate vehicle driving time series data based on the vehicle driving data, wherein the vehicle driving time series data is used to characterize time series characteristics and context association information of the vehicle driving data; Extracting key time series data from the vehicle driving time series data and the attention weight of each key time series data based on the attention mechanism; Converting the key time series data into target time series data in a time series format; Predicting the driver's driving style at the next moment based on the target time series data and the attention weights corresponding to each target time series data; The driving style detection model outputs multiple types of driving styles and the probability corresponding to each type of driving style; and the vehicle is controlled accordingly based on the driving style, including: receiving a torque request input by a driver via an accelerator pedal; Obtain torque variation curves corresponding to various types of driving styles; Based on the probabilities corresponding to the various types of driving styles, weighted summation is performed on the torque change curves corresponding to the various types of driving styles to obtain the torque change curve after the weighted summation; The output torque is obtained through the weighted summed torque variation curve, and the vehicle is controlled through the output torque.

2. The method according to claim 1, characterized in that The driving style detection model includes a time series encoder, a latent space attention mechanism module including a self-attention mechanism and a cross-attention mechanism, and a time series decoder; The generating of vehicle driving time series data based on the vehicle driving data comprises: Encoding the vehicle driving data into vehicle driving time series data by the time series encoder; Accordingly, the key time series data in the vehicle driving time series data and the attention weight of each key time series data are extracted based on the attention mechanism, including: The vehicle driving time series data is processed by the latent space attention mechanism module to obtain key time series data in the vehicle driving time series data and the attention weight of each key time series data; Correspondingly, the converting of the key time series data into target time series data in a time series format includes: converting the key time series data into target time series data in a time series format by using the time series decoder.

3. The method according to claim 2, characterized in that The driving style detection model also includes a pedal and steering prediction module and an online gradient network; The method predicts a predicted value of vehicle driving data at the next moment based on the vehicle driving data, generates an online feedback strategy based on an error between the vehicle driving data and the predicted value of the vehicle driving data, and performs online feedback training on a driving style detection model based on the online feedback strategy to obtain a driving style detection model after online training, including: Based on the target time series data and the attention weights corresponding to each target time series data, predicting the pedal data prediction value and the steering data prediction value at the next moment through the pedal and steering prediction module; Calculating the pedal data and steering data included in the vehicle driving data by the online gradient network, and the errors between the pedal data prediction value and the steering data prediction value, determining an online feedback strategy based on the errors, and feeding back the online feedback strategy to the latent space attention mechanism module; The latent space attention mechanism module is trained online based on the online feedback strategy.

4. The method according to any one of claims 1 to 3, characterized in that: The step of obtaining vehicle driving data at a plurality of consecutive moments including the current moment includes: Acquire initial vehicle driving data at a plurality of consecutive moments including the current moment, wherein the initial vehicle driving data includes pedal data, steering data, driving speed, and yaw speed of the vehicle; The initial vehicle driving data is preprocessed to obtain preprocessed vehicle driving data, wherein the preprocessing includes data fusion and data denoising.

5. The method according to any one of claims 1 to 3, characterized in that: The step of inputting the vehicle driving data into a driving style detection model obtained through online training to obtain the driver's driving style at the next moment includes: The vehicle driving data is input into a driving style detection model obtained through online training, and the driver's driving style at the next moment is calculated based on the multiple types of driving styles output by the driving style detection model and the probability corresponding to each type of driving style.

6. A vehicle driving style adaptive device based on online learning, characterized in that: The device comprises: The vehicle driving data acquisition module is used to acquire the vehicle driving data of a plurality of consecutive moments including the current moment during the vehicle driving process; A driving style detection module, used for inputting the vehicle driving data into a driving style detection model obtained through online training to obtain the driver's driving style at the next moment; A vehicle control module, configured to control the vehicle accordingly based on the driving style; The online training process of the driving style detection model includes: predicting a predicted value of vehicle driving data at the next moment based on the vehicle driving data, generating an online feedback strategy based on an error between the vehicle driving data and the predicted value of the vehicle driving data, and performing online feedback training on the driving style detection model based on the online feedback strategy; The process of the driving style detection model predicting the driver's driving style at the next moment includes: Generate vehicle driving time series data based on the vehicle driving data, wherein the vehicle driving time series data is used to characterize time series characteristics and context association information of the vehicle driving data; Extracting key time series data from the vehicle driving time series data and the attention weight of each key time series data based on the attention mechanism; Converting the key time series data into target time series data in a time series format; Predicting the driver's driving style at the next moment based on the target time series data and the attention weights corresponding to each target time series data; The driving style detection model outputs multiple types of driving styles and the probability corresponding to each type of driving style; the vehicle control module is specifically used to: receiving a torque request input by a driver via an accelerator pedal; Obtain torque variation curves corresponding to various types of driving styles; Based on the probabilities corresponding to the various types of driving styles, weighted summation is performed on the torque change curves corresponding to the various types of driving styles to obtain the torque change curve after the weighted summation; The output torque is obtained through the weighted summed torque variation curve, and the vehicle is controlled through the output torque.

7. An electronic device, characterized in that: include: at least one processor; a memory for storing the at least one processor-executable instruction; The at least one processor is configured to execute the instructions to implement the method according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the method according to any one of claims 1 to 5.

Images