Driving intention recognition method and device, equipment and storage medium

Abstract

The embodiment of the invention provides a driving intention recognition method and device, equipment and a storage medium, and relates to the technical field of intelligent driving. The method comprises the steps of obtaining and preprocessing historical vehicle power data and historical operation data of a real vehicle to obtain standardized data, performing screening processing on the standardized data to obtain target data, establishing a short-term driving data prediction model based on a Mamba module, training the short-term driving data prediction model by using the target data, and obtaining a short-term driving data prediction result. Obtaining a target short-time driving data prediction model; obtaining vehicle power data and operation data of a real vehicle, and inputting the vehicle power data and operation data into the target short-time driving data prediction model for processing to obtain a prediction result; and obtaining preset rules corresponding to the plurality of driving intentions, and identifying the prediction result by using the preset rules to obtain a target driving intention. The method is used for achieving the effects of reducing the calculation cost in the driving intention recognition process and improving the accuracy and reliability of driving intention recognition.

Description

Technical Field

[0001] The present application relates to the field of intelligent driving technology, and in particular to a driving intention recognition method, device, equipment and storage medium. Background Art

[0002] With the continuous progress of the automotive industry, advanced driver assistance systems have become a key component of intelligent connected vehicles, which greatly improves the convenience and safety of driving operations. In order to further develop these systems, it is particularly important to accurately identify driving intentions. By understanding and predicting the driver's behavior, driving intention recognition technology can make vehicle driving safer and smoother. Therefore, developing an efficient and accurate driving intention recognition system is an important goal to promote technological innovation in intelligent connected vehicles.

[0003] Among existing driving intention recognition technologies, deep learning demonstrates higher recognition accuracy for drivers' subtle movements and potential intentions than rule-based methods. It can parse and predict the driver's next behavior through complex algorithm analysis. At the same time, compared with shallow machine learning technology, the advanced structure of deep learning enables it to effectively perform long-term series analysis, thereby accurately extracting the driver's true intentions from continuous driving data.

[0004] However, deep learning models such as the Transformer model rely on the attention mechanism and require a large amount of parallel computing, which increases the computing cost in the prediction stage. In addition, the RNN model must be trained sequentially and cannot be trained in parallel, which makes the training process more time-consuming and consumes relatively large computing resources. At the same time, in the actual data collection process, all driving signal data for driving intention recognition training will have a large amount of interference data, which will affect the structure and relevance of the driving data, resulting in poor accuracy and reliability of driving intention recognition. Summary of the invention

[0005] The embodiments of the present application provide a driving intention recognition method, device, equipment and storage medium to reduce the computational cost in the driving intention recognition process and improve the accuracy and reliability of driving intention recognition.

[0006] In a first aspect, an embodiment of the present application provides a driving intention recognition method, comprising:

[0007] Acquire historical vehicle power data and historical operation data of the actual vehicle, pre-process the historical vehicle power data and historical operation data to obtain standardized data, and filter and process the standardized data to obtain target data, where the target data is used to represent data corresponding to any driving scenario of the actual vehicle from start to end;

[0008] A short-time driving data prediction model is established based on the Mamba module, and the short-time driving data prediction model is trained using the target data to obtain a target short-time driving data prediction model;

[0009] Obtaining vehicle power data and operation data of the actual vehicle, inputting the vehicle power data and operation data into the target short-term driving data prediction model for processing, and obtaining a prediction result, which refers to the speed, longitudinal acceleration and lateral acceleration of the actual vehicle in a preset time period in the future;

[0010] The preset rules corresponding to multiple driving intentions are obtained, and the prediction results are identified using the preset rules to obtain the target driving intention.

[0011] In a possible implementation, the target short-time driving data prediction model includes an embedding layer, a random inactivation layer, a first Mamba module, a second Mamba module, a first normalization layer, a feedforward network, a second normalization layer and a mapping layer. The vehicle power data and the operation data are input into the target short-time driving data prediction model for processing to obtain a prediction result, including:

[0012] Inputting the vehicle power data and the operation data into the embedding layer, processing the vehicle power data and the operation data using the embedding layer to obtain a first result, and inputting the first result into the random deactivation layer to obtain a second result;

[0013] Inputting the second result into the first Mamba module for processing to obtain a third result;

[0014] Inputting the second result into the second Mamba module for processing to obtain a fourth result;

[0015] Add the third result and the fourth result to obtain a calculation result;

[0016] The calculation results are sequentially input into the first normalization layer, the feedforward network, the second normalization layer and the mapping layer to obtain the prediction results.

[0017] In a possible implementation manner, the second result is input into the second Mamba module for processing to obtain a fourth result, including:

[0018] Flipping the second result to obtain the first processing result, and processing the first processing result using the second Mamba module to obtain the second processing result;

[0019] The second processing result is flipped to obtain a fourth result.

[0020] In a possible implementation, the vehicle power data and the operation data are processed by using the embedding layer to obtain a first result, including:

[0021] The vehicle power data and the operation data are labeled according to the variable scale using the linear layer in the embedding layer to obtain a first result.

[0022] In a possible implementation, historical vehicle power data and historical operation data of a real vehicle are obtained, the historical vehicle power data and historical operation data are preprocessed to obtain standardized data, and the standardized data is screened to obtain target data, including:

[0023] Analyze historical vehicle dynamics data and historical operation data to determine multiple signal fields corresponding to driving intent;

[0024] According to the signal field, historical vehicle power data and historical operation data are screened to obtain first driving data, and the first driving data is preprocessed to obtain standardized data;

[0025] Determine the driving scene type corresponding to the standardized data, select the first standardized data corresponding to any segment of the driving scene type, and divide the first standardized data according to the complete start-stop rule to obtain the target data.

[0026] In a possible implementation, the multiple driving intentions include acceleration driving intention, deceleration driving intention and turning intention, and the preset rules corresponding to the multiple driving intentions are obtained. The prediction results are identified using the preset rules to obtain the target driving intention, including:

[0027] Respectively obtaining a first preset rule corresponding to an acceleration driving intention, a second preset rule corresponding to a deceleration driving intention, and a third preset rule corresponding to a turning intention;

[0028] According to the first preset rule, the second preset rule and the third preset rule, the prediction result is identified and processed to determine the target driving intention corresponding to the prediction result.

[0029] In a possible implementation, a short-time driving data prediction model is established based on the Mamba module, and the short-time driving data prediction model is trained using target data to obtain a target short-time driving data prediction model, including:

[0030] A short-term driving data prediction model is established based on the Mamba module, the target data is divided into a plurality of window data according to a preset time period, and the plurality of window data are sequentially input into the short-term driving data prediction model according to a first preset time period by using a sliding window method to obtain a plurality of first prediction results;

[0031] Obtaining true values ​​corresponding to the plurality of first prediction results, and using a time series loss function to respectively calculate the plurality of first prediction results and the true values ​​to obtain a first calculation result;

[0032] According to the first calculation result, the short-time driving data prediction model is optimized to obtain a target short-time driving data prediction model, wherein the time series loss function is a first-order difference introduced on the basis of the mean square error.

[0033] In a second aspect, an embodiment of the present application provides a driving intention recognition device, comprising:

[0034] A processing module is used to obtain historical vehicle power data and historical operation data of the real vehicle, pre-process the historical vehicle power data and historical operation data to obtain standardized data, and filter and process the standardized data to obtain target data, which is used to represent data corresponding to any driving scene of the real vehicle from start to end;

[0035] A training module, used for establishing a short-time driving data prediction model based on the Mamba module, training the short-time driving data prediction model using target data, and obtaining a target short-time driving data prediction model;

[0036] A prediction module is used to obtain vehicle power data and operation data of the actual vehicle, input the vehicle power data and operation data into the target short-term driving data prediction model for processing, and obtain a prediction result, which refers to the speed, longitudinal acceleration and lateral acceleration of the actual vehicle in a preset time period in the future;

[0037] The recognition module is used to obtain preset rules corresponding to multiple driving intentions, identify the prediction results using the preset rules, and obtain the target driving intention.

[0038] In a possible implementation, the processing module is further configured to:

[0039] Analyze historical vehicle dynamics data and historical operation data to determine multiple signal fields corresponding to driving intent;

[0040] According to the signal field, historical vehicle power data and historical operation data are screened to obtain first driving data, and the first driving data is preprocessed to obtain standardized data;

[0041] Determine the driving scene type corresponding to the standardized data, select any segment of first standardized data corresponding to the driving scene type, and divide the first standardized data according to the complete start-stop rule to obtain target data.

[0042] In a possible implementation, the training module is further used to:

[0043] A short-term driving data prediction model is established based on the Mamba module, the target data is divided into a plurality of window data according to a preset time period, and the plurality of window data are sequentially input into the short-term driving data prediction model according to a first preset time period by using a sliding window method to obtain a plurality of first prediction results;

[0044] Obtaining true values ​​corresponding to the plurality of first prediction results, and using a time series loss function to respectively calculate the plurality of first prediction results and the true values ​​to obtain calculation results;

[0045] According to the calculation results, the short-time driving data prediction model is optimized to obtain the target short-time driving data prediction model, in which the time series loss function introduces the first-order difference on the basis of the mean square error.

[0046] In a possible implementation, the prediction module is further used to:

[0047] The target short-term driving data prediction model includes an embedding layer, a random inactivation layer, a first Mamba module, a second Mamba module, a first normalization layer, a feedforward network, a second normalization layer, and a mapping layer;

[0048] Inputting the vehicle power data and the operation data into the embedding layer, processing the vehicle power data and the operation data using the embedding layer to obtain a first result, and inputting the first result into the random deactivation layer to obtain a second result;

[0049] Inputting the second result into the first Mamba module for processing to obtain a third result;

[0050] Inputting the second result into the second Mamba module for processing to obtain a fourth result;

[0051] Add the third result and the fourth result to obtain a calculation result;

[0052] The calculation results are sequentially input into the first normalization layer, the feedforward network, the second normalization layer and the mapping layer to obtain the prediction results.

[0053] In a possible implementation, the prediction module is further used to:

[0054] Flipping the second result to obtain the first processing result, and processing the first processing result using the second Mamba module to obtain the second processing result;

[0055] The second processing result is flipped to obtain a fourth result.

[0056] In a possible implementation, the prediction module is further used to:

[0057] The vehicle power data and the operation data are labeled according to the variable scale using the linear layer in the embedding layer to obtain a first result.

[0058] In a possible implementation, the identification module is further used to:

[0059] Respectively obtaining a first preset rule corresponding to an acceleration driving intention, a second preset rule corresponding to a deceleration driving intention, and a third preset rule corresponding to a turning intention;

[0060] According to the first preset rule, the second preset rule and the third preset rule, the prediction result is identified and processed to determine the target driving intention corresponding to the prediction result.

[0061] In a third aspect, an embodiment of the present application provides a driving intention recognition device, including: a memory, a processor;

[0062] The memory stores computer-executable instructions;

[0063] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementations of the first aspect.

[0064] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the first aspect above and / or various possible implementations of the first aspect.

[0065] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the above first aspect and / or various possible implementation methods of the first aspect.

[0066] The driving intention recognition method, device, equipment and storage medium provided in the embodiment of the present application obtain target data by preprocessing and screening historical vehicle power data and historical operation data, reducing the influence of interference data, making the data have a clear structure in the time series and improving the correlation between data. A short-term driving data prediction model is established based on the simple calculation structure of the Mamba module, and the short-term driving data prediction model is trained using the target data to obtain a target short-term driving data prediction model. By obtaining the current vehicle power data and operation data of the actual vehicle, these data are input into the target short-term driving data prediction model to obtain a prediction result. And by obtaining the preset rules corresponding to multiple driving intentions, the prediction results are identified using the preset rules to obtain the target driving intention, thereby achieving the technical effect of reducing the calculation cost and improving the prediction accuracy and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0068] Figure 1 A schematic diagram of a driving intention recognition method provided in an embodiment of the present application Figure 1 ;

[0069] Figure 2 A schematic diagram of the results of Kalman velocity filtering provided in an embodiment of the present application;

[0070] Figure 3 A schematic diagram of a driving intention recognition method provided in an embodiment of the present application Figure 2 ;

[0071] Figure 4 A schematic diagram of the structure of standardized data division provided in an embodiment of the present application;

[0072] Figure 5 A schematic diagram of a flow chart of a first preset rule provided in an embodiment of the present application;

[0073] Figure 6 A schematic diagram of a flow chart of a second preset rule provided in an embodiment of the present application;

[0074] Figure 7 A schematic diagram of a flow chart of a third preset rule provided in an embodiment of the present application;

[0075] Figure 8 A schematic diagram of a driving intention recognition method provided in an embodiment of the present application Figure 3 ;

[0076] Fig. 9 A schematic diagram of the structure of a target short-term driving data prediction model provided in an embodiment of the present application;

[0077] Fig.10 A schematic diagram of the structure of a driving intention recognition device provided in an embodiment of the present application;

[0078] Fig.11 A schematic diagram of the structure of a driving intention recognition device provided in an embodiment of the present application.

[0079] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0080] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0081] In the current field of driving intention recognition technology, deep learning can analyze and predict the driver's next behavior by analyzing the driver's subtle movements and potential intentions with the help of complex algorithms. At the same time, the advanced structure of deep learning enables it to effectively perform long-term series analysis, thereby extracting the driver's true intention from continuous driving data. However, deep learning models such as the Transformer model rely on the attention mechanism to perform a large number of parallel calculations, which increases the computational cost of the prediction stage. In addition, models such as RNN need to be executed sequentially during the training process and cannot perform parallel calculations, resulting in a time-consuming training process and high consumption of computing resources. At the same time, when actually collecting data, using all driving signal data for driving intention recognition training will result in a lot of interference data, which will affect the structure and correlation of the driving data, thereby reducing the accuracy and reliability of driving intention recognition.

[0082] A driving intention recognition method, device, equipment and storage medium provided by the embodiment of the present application obtains target data by preprocessing and screening historical vehicle power data and historical operation data. A target short-term driving data prediction model that can predict driving data in a short period of time is obtained by using the target data and the Mamba module, and the current vehicle power data and operation data of the actual vehicle are input into the target short-term driving data prediction model for processing to obtain the prediction results of the driver's speed, longitudinal acceleration and lateral acceleration in the future preset time period. By obtaining the preset rules corresponding to multiple driving intentions, the prediction results are identified using the preset rules to obtain the target driving intention. Therefore, by preprocessing and screening the historical driving data, the influence of interference data is reduced, so that the data has a clear structure in the time series and the correlation between the data is improved, so that the subsequent short-term driving data prediction model learns the correlation and regularity between different driving behaviors, thereby more accurately and reliably identifying the driving intention, and at the same time using the Mamba model to create the target short-term driving data prediction model reduces the computational complexity and cost in the training and prediction process, and improves the prediction efficiency.

[0083] The technical solution of the present application and how the technical solution of the present application solves the above-mentioned technical problems are described in detail below with specific embodiments. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0084] Figure 1 A schematic diagram of a driving intention recognition method provided in an embodiment of the present application Figure 1 ,like Figure 1 As shown, the method includes:

[0085] S101. Acquire historical vehicle power data and historical operation data of a real vehicle, pre-process the historical vehicle power data and historical operation data to obtain standardized data, and filter the standardized data to obtain target data.

[0086] In this embodiment, the historical vehicle power data and historical operation data of a large number of real vehicles of multiple drivers and multiple regions are collected by the vehicle-mounted sensors and recording devices installed in the vehicle and the shooting equipment carried by the drone to ensure the diversity and representativeness of the data set. Among them, the vehicle-mounted sensors and recording devices include but are not limited to vehicle-mounted units, global positioning systems, vehicle-mounted inertial measurement units, etc. The historical vehicle power data may include, for example, the historical speed signal, historical longitudinal acceleration, historical lateral acceleration, historical motor speed, and historical motor torque of the vehicle; the historical operation data includes historical accelerator pedal position, historical brake pedal status, and historical steering angle, etc.

[0087] It should be noted that due to the huge amount of historical vehicle power data and historical operation data, these data need to be uploaded to the cloud for storage so that they can be directly obtained from the cloud for subsequent use.

[0088] Furthermore, the historical vehicle power data and historical operation data are preprocessed to obtain standardized data. The preprocessing operation includes but is not limited to: screening out signal fields that are closely related to driving intention recognition; identifying and processing outliers and missing values; correcting data deviations by deletion or filling; deleting data or filling it by interpolation; and removing or weakening noise and interference in the time series through certain algorithms, such as weighted averaging or filtering technology.

[0089] Take the above-mentioned outliers, missing values, and preprocessing operations to remove or weaken noise interference in time series as examples. The possible implementation methods of its preprocessing operations are described in detail. Outliers: 1. Abnormal data caused by collection errors or information transmission errors. For example, the standard deviation method can be used to judge this type of outliers, and the box plot method can be combined to remove outliers. The box plot uses the quartiles of the data to construct an intuitive graphical representation, and regards data above or below the boundary of the data distribution as outliers. 2. Another type of outlier is a series of values ​​that have not changed. For example, the downsampling method can be used to process this outlier. Missing values: For this type of data, a variety of operations are taken. If it is determined that the data is continuously missing, this part of the data will be deleted; if a small amount of missing data is present, the forward data filling method and the average value filling method are used. Remove or weaken noise interference in time series: For example, the Kalman filter method can be used to smooth the data to remove or reduce the interference of these noises, such as Figure 2 As shown, Figure 2 A schematic diagram of the results of the Kalman velocity filter provided in an embodiment of the present application.

[0090] Furthermore, since the standardized data contains the data required for effective identification of driving intention, but also contains a large amount of invalid information, and even some information may interfere with the identification of driving intention, the standardized data is screened and processed to obtain the target data.

[0091] S102: Establish a short-time driving data prediction model based on the Mamba module, and use the target data to train the short-time driving data prediction model to obtain a target short-time driving data prediction model.

[0092] In this embodiment, the Mamba module is the basis for building a short-term driving data prediction model. When establishing the short-term driving data prediction model, the target data is input into the short-term driving data prediction model. The short-term driving data prediction model continuously adjusts the model parameters by learning and analyzing a large amount of known data, so that it can better fit the training data, thereby improving the accuracy of the prediction. The target short-term driving data prediction model finally obtained has the ability to predict driving behavior or related data in the short term in the future based on the input current driving data information, such as predicting the vehicle speed change in the next few seconds or minutes.

[0093] It should be noted that the Mamba module is a selective structural state space model, which is a model generated by introducing a selective state space mechanism based on the state space model (SSM). In addition, the Mamba module uses a novel hardware-aware parallel computing method to ensure efficient training of the model. Among them, the state space model is a mathematical model used to describe the dynamic behavior of a system over time. The state space model describes the state of the system through state variables, and establishes a mathematical model of the system through state equations and observation equations. The state space model shows the potential to optimize performance and computational complexity at the same time. It uses convolution calculations to capture sequence information and eliminate hidden states, so that it benefits from parallel computing and achieves near-linear complexity in processing speed. The state space model can represent any cyclic process with potential states. The evolution of the internal state of the system is represented by a first-order differential equation, and the relationship between the potential state and the output sequence is described by another set of differential equations. The input sequence x(t) can be mapped to the output sequence y(t) through the potential state h(t):

[0094] h(t)'=Ah(t)+Bx(t), y(t)=Ch(t)

[0095] In the formula, h(t)' represents the rate of change or derivative of the potential state at time t, indicating the change of the state over time. A represents the state transfer matrix, A∈R N×N , B represents the input matrix, B∈R D×N , C represents the output matrix, C∈R N×D , variables N and D represent the state expansion factor and dimension factor respectively. A and B can be discretized by zero-order hold and time sampling with an interval of Δ, defined as follows:

[0096]

[0097] In the formula, represents the discretized state transfer matrix, Represents the discretized input matrix.

[0098] The discretized state space model can be expressed as:

[0099]

[0100] In the formula, h t represents the potential state at discrete time t, h t-1 represents the potential state at the last discrete time t-1, y t Represents the output vector at discrete time t.

[0101] Discretized state-space models can be trained in parallel using convolution operations and analyzed efficiently using recurrent neural networks.

[0102] S103: Acquire vehicle power data and operation data of the actual vehicle, input the vehicle power data and operation data into a target short-time driving data prediction model for processing, and obtain a prediction result.

[0103] In this embodiment, the real-time vehicle power data and operation data of the real vehicle are obtained through the vehicle-mounted sensors and recording equipment and the shooting equipment carried by the drone, and these data are input into the target short-term driving data prediction model after pre-processing. After receiving the input data, the target short-term driving data prediction model will analyze and calculate these data according to its pre-learned and trained knowledge and patterns to obtain the prediction results.

[0104] The prediction results refer to the speed, longitudinal acceleration and lateral acceleration of the actual vehicle in a preset time period in the future.

[0105] For example, the vehicle power data and operation data of a real vehicle can be defined as a multivariate time series X in =[x1,x2,...,x L ]∈R L×M , and the prediction result is X out =[x L+1 ,x L+2 ,...,x L+H ]∈R H×M , where L is the length of the historical lookback window, M is the feature dimension, and H is the prediction time length, that is, the preset time period in the future.

[0106] S104, obtaining preset rules corresponding to a plurality of driving intentions, identifying the prediction results using the preset rules, and obtaining the target driving intention.

[0107] In this embodiment, preset rules corresponding to multiple driving intentions are obtained, and these rules are pre-set according to the typical characteristics and behavior patterns of different driving intentions. For the obtained prediction results, these preset rules are used to compare and identify them one by one. By analyzing whether the prediction results meet the characteristics of a preset rule, it is determined to which driving intention the prediction results belong, and the driving intention that is finally determined to best match the prediction results is the target driving intention. Among them, the preset rules are mainly set according to speed, longitudinal acceleration and lateral acceleration. Longitudinal acceleration refers to the acceleration or deceleration of the vehicle in the forward or backward direction; lateral acceleration refers to the acceleration generated by the vehicle during the turning process, which is perpendicular to the direction of travel of the vehicle.

[0108] The driving intention recognition method provided in the embodiment of the present application obtains target data by preprocessing and screening historical vehicle power data and historical operation data, reduces the influence of interference data, makes the data have a clear structure in the time series and improves the correlation between data. A short-term driving data prediction model is established based on the simple calculation structure of the Mamba module, and the short-term driving data prediction model is trained using the target data to obtain a target short-term driving data prediction model. By obtaining the current vehicle power data and operation data of the actual vehicle, these data are input into the target short-term driving data prediction model to obtain a prediction result. And by obtaining the preset rules corresponding to multiple driving intentions, the preset rules are used to identify the prediction results, and the target driving intention is obtained, thereby achieving the technical effect of reducing the calculation cost and improving the prediction accuracy and reliability.

[0109] Figure 3 A schematic diagram of a driving intention recognition method provided in an embodiment of the present application Figure 2 ,like Figure 3 As shown, in this embodiment Figure 1 Based on the embodiment, a driving intention recognition method is described in detail, and the method includes:

[0110] S301. Acquire historical vehicle power data and historical operation data of a real vehicle, analyze the historical vehicle power data and historical operation data, and determine a plurality of signal fields corresponding to driving intentions.

[0111] In this embodiment, since the acquired historical vehicle power data and historical operation data contain some data irrelevant to driving intention, the historical vehicle power data and historical operation data are analyzed to determine multiple signal fields corresponding to driving intention. For example, Table 1 is a partial driving signal field provided in an embodiment of the present application, as shown in Table 1.

[0112] Table 1

[0113] Field Name meaning unit id Unique identifier of the vehicle - terminal_time time s ESP_VehicleSpeed Vehicle speed m / s ESP_LatAccel Lateral acceleration <![CDATA[m / s 2 ]]> ESP_LongAccel Longitudinal acceleration <![CDATA[m / s 2 ]]> PCU_AccPedl Accelerator pedal position degrees

[0114] S302: Filter historical vehicle power data and historical operation data according to the signal field to obtain first driving data, and pre-process the first driving data to obtain standardized data.

[0115] In this embodiment, by screening the historical vehicle power data and the historical operation data according to the signal field, a data subset that meets the conditions, namely the first driving data, can be extracted, and the first driving data is preprocessed so that it has a unified standard and format, which is convenient for subsequent further processing and analysis. The preprocessing process is similar to the above step S101 and will not be repeated here.

[0116] S303: Determine the driving scene type corresponding to the standardized data, select the first standardized data corresponding to any segment of the driving scene type, and divide the first standardized data according to the complete start-stop rule to obtain target data.

[0117] In this embodiment, since the historical vehicle power data and historical operation data are collected from multiple regions, including different driving scenes such as urban roads, highways and rural roads, the driving scene type corresponding to the standardized data can be determined according to various characteristics and indicators, and then any driving scene type is selected, and the first standardized data corresponding to this driving scene type is obtained, and the first standardized data is divided according to the complete start-stop rule to obtain the target data. Among them, the complete start-stop rule refers to the distance from the vehicle speed starting at 0 to the vehicle speed ending at 0, and the complete start-stop rule involves the vehicle's start, stop, acceleration, deceleration and other behaviors.

[0118] For example, Figure 4 A schematic diagram of the structure of the standardized data division provided in the embodiment of the present application, such as Figure 4 As shown, first determine the driving scene type corresponding to the standardized data, and then divide the standardized data into different journey events according to different driving scene types, select the first standardized data corresponding to any journey event, divide the journey event according to the complete start-stop rule, obtain multiple journey events, select any journey event, and obtain the target data. In addition, for example, the journey event can also be divided into acceleration events and braking events. The acceleration event represents a section of the vehicle's driving journey where the throttle attribute value is not 0; the braking event represents a section of the journey where the throttle attribute value is 0.

[0119] S304: Establish a short-time driving data prediction model based on the Mamba module, divide the target data into multiple window data according to a preset time period, and use a sliding window method to input the multiple window data into the short-time driving data prediction model in sequence according to a first preset time to obtain multiple first prediction results.

[0120] In this embodiment, the target data is divided according to the preset time period to form a plurality of independent window data, and the sliding window method is adopted, that is, at a fixed time interval (first preset time), the window data is input into the short-term driving data prediction model at one time. When each window data is input, the short-term driving data prediction model will perform calculation and inference according to its internal logic and algorithm, and output a first prediction result for the window data. Through such operation, a series of a plurality of first prediction results corresponding to the window data of different time periods can be obtained, so that the driving state and trend of the vehicle in different time periods can be understood and predicted more comprehensively and meticulously.

[0121] S305. Obtain true values ​​corresponding to multiple first prediction results, use a time series loss function to respectively calculate the multiple first prediction results and the true values ​​to obtain a first calculation result, and optimize the short-time driving data prediction model based on the first calculation result to obtain a target short-time driving data prediction model.

[0122] In this embodiment, since the target data is time series driving data and the previous and next data have a strong correlation, the time series loss function is used to calculate the difference or error between each first prediction result and the corresponding true value. According to these results, the short-term driving data prediction model is adjusted and optimized so that the short-term driving data prediction model can better fit the real data and improve the accuracy of the prediction.

[0123] Among them, the time series loss function can be expressed as:

[0124]

[0125] in, Represents the mean square error between the true value and the predicted result, y (i) Indicates the true value corresponding to the first prediction result at time i, represents the first prediction result at time i, H represents the prediction time length, that is, the preset time period in the future; L D represents the nth-order difference with a step size of m. Preferably, the step size m=1 and the order n=1 are set, which can be expressed as:

[0126]

[0127] In the formula, y (t) Represents the true value corresponding to the first prediction result at time t, represents the first prediction result at time t+k, where k represents the time interval. represents the first-order difference of the first prediction result at time t+k, d (t+k) Indicates the first-order difference of the true value corresponding to the first prediction result at time t+k.

[0128] The time series loss function introduces the first-order difference on the basis of the mean square error. The mean square error measures the average error between the predicted value and the true value, focusing on the accuracy of the prediction; while the first-order difference loss function focuses on the changing trend of adjacent time steps in the time series, emphasizing the continuity and dynamics of the prediction. This combined method enables the short-term driving data prediction model to better capture and simulate the inherent change law of the time series while reducing the error, thereby improving the reliability and stability of the prediction. In addition, due to the introduction of the first-order difference, the short-term driving data prediction model can more flexibly adapt to the local fluctuations of the time series, enhance the ability to capture short-term patterns, and reduce the cumulative error in long-term predictions.

[0129] S306: Acquire vehicle power data and operation data of the actual vehicle, input the vehicle power data and operation data into the target short-time driving data prediction model for processing, and obtain a prediction result.

[0130] Step S306 is similar to the above step S103 and will not be described again.

[0131] S307, the driving intention includes acceleration driving intention, deceleration driving intention and turning intention, and the first preset rule corresponding to the acceleration driving intention, the second preset rule corresponding to the deceleration driving intention and the third preset rule corresponding to the turning intention are respectively obtained.

[0132] In this embodiment, since the prediction result is used to refer to the changes in the speed, longitudinal acceleration and lateral acceleration of the actual vehicle in the future preset time period, Figures 5 to 7 , the first preset rule, the second preset rule and the third preset rule are explained, Figure 5 A schematic diagram of a flow chart of a first preset rule provided in an embodiment of the present application, Figure 6 A schematic diagram of a flow chart of a second preset rule provided in an embodiment of the present application, Figure 7 A flowchart of the third preset rule provided in an embodiment of the present application.

[0133] like Figure 5 As shown, the first preset rule corresponding to the acceleration driving intention is: first determine whether the speed change trend increases with the increase of time in the prediction result. If so, determine whether the longitudinal acceleration changes with time is greater than the first preset value. If so, determine whether the duration of the longitudinal acceleration is greater than or equal to the first preset time. If so, identify the prediction result as the acceleration driving intention.

[0134] like Figure 6 As shown, the second preset rule corresponding to the deceleration driving intention is: first determine whether the speed change trend decreases with the increase of time in the prediction result. If so, determine whether the longitudinal acceleration change with time is less than the second preset value. If so, determine whether the duration of the longitudinal acceleration is greater than or equal to the first preset time. If so, identify the prediction result as the deceleration driving intention.

[0135] like Figure 7 As shown, the third preset rule corresponding to the turning intention is: determine whether the lateral acceleration in the prediction result is of the same sign as time increases; if so, calculate the absolute value of the lateral acceleration and determine whether the absolute value is greater than the third preset value; if so, determine whether the duration is greater than or equal to the first preset time; if so, identify the prediction result as a turning driving intention.

[0136] S308: According to the first preset rule, the second preset rule and the third preset rule, the prediction result is identified and processed to determine the target driving intention corresponding to the prediction result.

[0137] In this embodiment, the prediction result is analyzed and identified by using the first preset rule, the second preset rule and the third preset rule to determine whether the prediction result meets which preset rule, and then obtain the driving intention corresponding to the prediction result.

[0138] The driving intention recognition method provided by the embodiment of the present application, by obtaining the historical vehicle power data and historical operation data of the real vehicle and analyzing and determining the signal field, helps to accurately grasp the key elements related to the driving intention. Screening and preprocessing the historical driving data can improve the quality and availability of the data, remove irrelevant or interfering information, and make the data processed by the subsequent model more accurate and reliable. By determining the driving scene type and dividing it accordingly, the clarity and relevance of the structure of the data are improved, and the driving intention analysis under different scenes can be better adapted, thereby improving the adaptability and accuracy of the model to various complex driving situations. A short-term driving data prediction model is established based on the Mamba module, which reduces the computational cost in the training process and the prediction process. By dividing the target data into multiple window data and inputting them in a sliding window manner, it is beneficial to capture the time series characteristics of the data. And by training the model using the time series loss function, while reducing the error, it can also better capture and simulate the inherent change law of the time series, thereby improving the reliability and stability of the prediction. The real-time driving data is processed using the target short-term driving data prediction model, and the prediction results can be quickly obtained to provide support for timely driving decisions. Different preset rules are set for different driving intentions to make the recognition process more precise and targeted, and to improve the accuracy of judgments on various specific driving intentions.

[0139] Figure 8 A schematic diagram of a driving intention recognition method provided in an embodiment of the present application Figure 3 In this embodiment, Figure 3 Based on the embodiment, a possible implementation method of inputting vehicle power data and operation data into a target short-term driving data prediction model for processing to obtain a prediction result is described in detail, wherein the target short-term driving data prediction model includes an embedding layer, a random inactivation layer, a first Mamba module, a second Mamba module, a first normalization layer, a feedforward network, a second normalization layer and a mapping layer. Figure 8 As shown, the method includes:

[0140] S801. Input the vehicle power data and the operation data into an embedding layer, use a linear layer in the embedding layer to label the vehicle power data and the operation data according to a variable scale to obtain a first result, and input the first result into a random deactivation layer to obtain a second result.

[0141] In this embodiment, the vehicle power data and operation data are input into the embedding layer. The linear layer in the embedding layer will analyze and process these data according to the pre-set variable scale, and mark or label these data with specific marks, so as to better identify and distinguish the characteristics and attributes of the vehicle power data and operation data in the future. The first result obtained after the linear layer processing will be passed to the random deactivation layer. The function of the random deactivation layer is to randomly deactivate some neurons according to a certain probability during the training process, that is, temporarily not participate in the calculation. This can prevent the model from overfitting and increase the generalization ability of the model. Through the processing of the random deactivation layer, the second result is obtained. Compared with the first result, this result has been adjusted and optimized to a certain extent, which is more conducive to the further analysis and application of the subsequent model.

[0142] S802: Input the second result to the first Mamba module for processing to obtain a third result.

[0143] In this embodiment, when the second result is input into the first Mamba module, the module will perform specific calculation and conversion on the second result according to the algorithm and logic set inside the module to obtain the third result.

[0144] S803, flip the second result to obtain the first processing result, use the second Mamba module to process the first processing result to obtain the second processing result, and flip the second processing result to obtain the fourth result.

[0145] In this embodiment, the second result is flipped, that is, each element or feature thereof is reversed according to a certain rule, such as the positive and negative conversion of the numerical value, the reversal of the order, etc., so as to obtain the first processing result. The first processing result is input into the second Mamba module. The second Mamba module performs specific calculations and conversions on the first processing result according to the algorithm and logic set inside it to obtain the second processing result. The second processing result is flipped again, and a fourth result is obtained through a flipping operation similar to the above. The above flipping process can view the data from different angles, which helps to obtain the correlation between all vehicle power data and operation data.

[0146] S804. Add the third result and the fourth result to obtain a calculation result, and input the calculation result to the first normalization layer, the feedforward network, the second normalization layer and the mapping layer in sequence to obtain a prediction result.

[0147] In this embodiment, by integrating the output results of the first Mamba module and the second Mamba module, the third result and the fourth result are added to obtain the calculation result, which can better capture the information and correlation of the data in the front and back directions, so that the model has a more comprehensive understanding of the data. Further, the calculation result is input to the first normalization layer, and the calculation result is standardized to a Gaussian distribution using the first normalization layer to enhance the convergence and stability of the deep model and minimize the differences caused by inconsistent measurements. The Gaussian distributed data is then converted into a time series representation through a feedforward network (FeedforwardNetwork, FFN). The FFN layer encodes the observed time series and decodes the future sequence representation using dense nonlinear connections. The second normalization layer is used to adjust the future sequence representation, and the tokenized time information is reconstructed into the time series to be predicted through the mapping layer, and then transposed to produce the final prediction result.

[0148] The driving intention recognition method provided in the embodiment of the present application processes the vehicle power data and operation data through the embedding layer marker and random deactivation layer, which can better process and refine the data. The data is flipped and two Mamba modules are combined into a two-way Mamba model, which helps to obtain the correlation between all variables, analyze the data from different angles, enrich the data features, and improve the adaptability and generalization ability of the model. Adding different results and processing them through multiple layers helps to fully integrate information and obtain more accurate and comprehensive prediction results.

[0149] Fig. 9 This is a schematic diagram of the structure of the target short-term driving data prediction model provided in the embodiment of the present application. Fig. 9 As shown, in this embodiment Figure 8Based on the embodiment, the target short-time driving data prediction model is described. The target short-time driving data prediction model includes an embedding layer, a random deactivation layer, a first Mamba module, a second Mamba module, a first normalization layer, a feedforward network, a second normalization layer and a mapping layer. The embedding layer uses the linear layer therein to mark the vehicle power data and the operation data to obtain a first result in a standardized time series format; the first result is input into the random deactivation layer to obtain a second result, and the random deactivation layer is used to randomly discard a part of neurons during the training process to enhance the generalization ability of the model. Since the selection mechanism of the Mamba module is unidirectional and cannot process global variables, two Mamba modules are used to process the second result, that is, the second result is input into the first Mamba module for processing to obtain a third result; at the same time, the second result is flipped to obtain a first processing result, and the first processing result is processed by the second Mamba module to obtain a second processing result, and the second processing result is flipped to obtain a fourth result. The third result and the fourth result are added to obtain a calculation result. The calculation results are sequentially input into the first normalization layer, the feedforward network, the second normalization layer and the mapping layer to obtain the prediction results. Among them, the second multivariate Mamba module includes multiple linear layers, multiple convolutional layers, activation functions, and state space models.

[0150] Fig.10 A schematic diagram of the structure of the driving intention recognition device provided in the embodiment of the present application is shown in FIG. Fig.10 As shown, the driving intention recognition device 100 provided in this embodiment includes:

[0151] The processing module 1001 is used to obtain historical vehicle power data and historical operation data of the actual vehicle, pre-process the historical vehicle power data and historical operation data to obtain standardized data, and filter the standardized data to obtain target data, which is used to represent data corresponding to any driving scene of the actual vehicle from start to end;

[0152] A training module 1002 is used to establish a short-time driving data prediction model based on the Mamba module, and train the short-time driving data prediction model using target data to obtain a target short-time driving data prediction model;

[0153] The prediction module 1003 is used to obtain the vehicle power data and operation data of the actual vehicle, input the vehicle power data and operation data into the target short-term driving data prediction model for processing, and obtain the prediction result, which refers to the speed, longitudinal acceleration and lateral acceleration of the actual vehicle in the future preset time period;

[0154] The identification module 1004 is used to obtain preset rules corresponding to multiple driving intentions, identify the prediction results using the preset rules, and obtain the target driving intention.

[0155] In a possible implementation, the processing module 1001 is further configured to:

[0156] Analyze historical vehicle dynamics data and historical operation data to determine multiple signal fields corresponding to driving intent;

[0157] According to the signal field, historical vehicle power data and historical operation data are screened to obtain first driving data, and the first driving data is preprocessed to obtain standardized data;

[0158] Determine the driving scene type corresponding to the standardized data, select any segment of first standardized data corresponding to the driving scene type, and divide the first standardized data according to the complete start-stop rule to obtain target data.

[0159] In a possible implementation, the training module 1002 is further configured to:

[0160] A short-term driving data prediction model is established based on the Mamba module, the target data is divided into a plurality of window data according to a preset time period, and the plurality of window data are sequentially input into the short-term driving data prediction model according to a first preset time period by using a sliding window method to obtain a plurality of first prediction results;

[0161] Obtaining true values ​​corresponding to the plurality of first prediction results, and using a time series loss function to respectively calculate the plurality of first prediction results and the true values ​​to obtain calculation results;

[0162] According to the calculation results, the short-time driving data prediction model is optimized to obtain the target short-time driving data prediction model, in which the time series loss function introduces the first-order difference on the basis of the mean square error.

[0163] In a possible implementation, the prediction module 1003 is further configured to:

[0164] The target short-term driving data prediction model includes an embedding layer, a random inactivation layer, a first Mamba module, a second Mamba module, a first normalization layer, a feedforward network, a second normalization layer, and a mapping layer;

[0165] Inputting the vehicle power data and the operation data into the embedding layer, processing the vehicle power data and the operation data using the embedding layer to obtain a first result, and inputting the first result into the random deactivation layer to obtain a second result;

[0166] Inputting the second result into the first Mamba module for processing to obtain a third result;

[0167] Inputting the second result into the second Mamba module for processing to obtain a fourth result;

[0168] Add the third result and the fourth result to obtain a calculation result;

[0169] The calculation results are sequentially input into the first normalization layer, the feedforward network, the second normalization layer and the mapping layer to obtain the prediction results.

[0170] In a possible implementation, the prediction module 1003 is further configured to:

[0171] Flipping the second result to obtain the first processing result, and processing the first processing result using the second Mamba module to obtain the second processing result;

[0172] The second processing result is flipped to obtain a fourth result.

[0173] In a possible implementation, the prediction module 1003 is further configured to:

[0174] The vehicle power data and the operation data are labeled according to the variable scale using the linear layer in the embedding layer to obtain a first result.

[0175] In a possible implementation, the identification module 1004 is further configured to:

[0176] Respectively obtaining a first preset rule corresponding to an acceleration driving intention, a second preset rule corresponding to a deceleration driving intention, and a third preset rule corresponding to a turning intention;

[0177] According to the first preset rule, the second preset rule and the third preset rule, the prediction result is identified and processed to determine the target driving intention corresponding to the prediction result.

[0178] The driving intention recognition device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and this embodiment will not be described in detail here.

[0179] Fig.11 This is a schematic diagram of the structure of the driving intention recognition device provided in the embodiment of the present application. Fig.11 As shown, the driving intention recognition device 11 provided in this embodiment includes: at least one processor 1101 and a memory 1102. Optionally, the device 11 also includes a communication component 1103. The processor 1101, the memory 1102 and the communication component 1103 are connected via a bus 1104.

[0180] In a specific implementation process, at least one processor 1101 executes the computer-executable instructions stored in the memory 1102, so that at least one processor 1101 executes the above method.

[0181] The specific implementation process of the processor 1101 can be found in the above method embodiment, and its implementation principle and technical effect are similar, so this embodiment will not be repeated here.

[0182] In the above embodiments, it should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. A general-purpose processor may be a microprocessor or any conventional processor. The steps of the method disclosed in the invention may be directly implemented as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0183] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (Non-volatile Memory, NVM), such as at least one disk memory.

[0184] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, the bus in the drawings of the present application is not limited to only one bus or one type of bus.

[0185] An embodiment of the present application also provides a computer program product, including a computer program, which implements the above method when executed by a processor.

[0186] An embodiment of the present application also provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the above method is implemented.

[0187] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The readable storage medium can be any available medium that can be accessed by a general or special-purpose computer.

[0188] An exemplary readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (Application Specific Integrated Circuits, referred to as: ASIC). Of course, the processor and the readable storage medium can also exist in the device as discrete components.

[0189] The division of units is only a logical function division, and there may be other divisions in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interface, device or unit, which can be electrical, mechanical or other forms.

[0190] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0191] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0192] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0193] Those skilled in the art can understand that all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, disk or optical disk and other media that can store program codes.

[0194] It should be noted that the terms "first", "second", "third", "fourth", etc. in the claims, the specification and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the invention described herein, for example, can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, products or devices.

[0195] 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 laws, regulations and standards in the relevant regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0196] Finally, it should be noted that those skilled in the art will readily conceive of other embodiments of the present invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention, which follow the general principles of the present invention and include common knowledge or customary technical means in the art not disclosed by the present invention, are not limited to the precise structure described above and shown in the drawings, and may be modified and changed in various ways without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A driving intention recognition method, characterized in that: include: Acquire historical vehicle power data and historical operation data of the actual vehicle, pre-process the historical vehicle power data and the historical operation data to obtain standardized data, and filter the standardized data to obtain target data, wherein the target data is used to represent data corresponding to any driving scenario of the actual vehicle from start to end; Establishing a short-time driving data prediction model based on the Mamba module, and using the target data to train the short-time driving data prediction model to obtain a target short-time driving data prediction model; Acquiring vehicle power data and operation data of the actual vehicle, inputting the vehicle power data and the operation data into a target short-time driving data prediction model for processing, and obtaining a prediction result, wherein the prediction result refers to the speed, longitudinal acceleration, and lateral acceleration of the actual vehicle in a future preset time period; Preset rules corresponding to multiple driving intentions are obtained, and the prediction results are identified using the preset rules to obtain the target driving intention.

2. The method according to claim 1, characterized in that The target short-time driving data prediction model includes an embedding layer, a random inactivation layer, a first Mamba module, a second Mamba module, a first normalization layer, a feedforward network, a second normalization layer and a mapping layer. The vehicle power data and the operation data are input into the target short-time driving data prediction model for processing to obtain a prediction result, including: Inputting the vehicle power data and the operation data into the embedding layer, processing the vehicle power data and the operation data using the embedding layer to obtain a first result, and inputting the first result into the random deactivation layer to obtain a second result; Inputting the second result into the first Mamba module for processing to obtain a third result; Inputting the second result into the second Mamba module for processing to obtain a fourth result; Adding the third result and the fourth result to obtain a calculation result; The calculation result is sequentially input into the first normalization layer, the feedforward network, the second normalization layer and the mapping layer to obtain the prediction result.

3. The method according to claim 2, characterized in that The step of inputting the second result into the second Mamba module for processing to obtain a fourth result includes: Flipping the second result to obtain a first processing result, and using the second Mamba module to process the first processing result to obtain a second processing result; The second processing result is flipped to obtain the fourth result.

4. The method according to claim 2, characterized in that: The using the embedding layer to process the vehicle power data and the operation data to obtain a first result includes: The vehicle power data and the operation data are labeled according to a variable scale using a linear layer in the embedding layer to obtain a first result.

5. The method according to claim 1, characterized in that The acquiring of historical vehicle power data and historical operation data of the actual vehicle, preprocessing the historical vehicle power data and the historical operation data to obtain standardized data, and screening the standardized data to obtain target data includes: Analyzing the historical vehicle power data and the historical operation data to determine a plurality of signal fields corresponding to the driving intention; According to the signal field, the historical vehicle power data and the historical operation data are screened to obtain first driving data, and the first driving data is preprocessed to obtain the standardized data; Determine the driving scene type corresponding to the standardized data, select any segment of first standardized data corresponding to the driving scene type, and divide the first standardized data according to the complete start-stop rule to obtain the target data.

6. The method according to claim 1, characterized in that The multiple driving intentions include acceleration driving intention, deceleration driving intention and turning intention, and the preset rules corresponding to the multiple driving intentions are obtained, and the prediction results are identified by using the preset rules to obtain the target driving intention, including: Respectively obtaining a first preset rule corresponding to the acceleration driving intention, a second preset rule corresponding to the deceleration driving intention, and a third preset rule corresponding to the turning intention; According to the first preset rule, the second preset rule and the third preset rule, the prediction result is identified and processed to determine the target driving intention corresponding to the prediction result.

7. The method according to claim 1, characterized in that The short-time driving data prediction model is established based on the Mamba module, and the short-time driving data prediction model is trained using the target data to obtain a target short-time driving data prediction model, including: Establishing a short-time driving data prediction model based on the Mamba module, dividing the target data into a plurality of window data according to a preset time period, and sequentially inputting the plurality of window data into the short-time driving data prediction model according to a first preset time by using a sliding window method, to obtain a plurality of first prediction results; Acquire true values ​​corresponding to a plurality of first prediction results, and respectively calculate the plurality of first prediction results and the true values ​​using a time series loss function to obtain a first calculation result; According to the first calculation result, the short-time driving data prediction model is optimized to obtain the target short-time driving data prediction model, wherein the time series loss function is a first-order difference introduced on the basis of the mean square error.

8. A driving intention recognition device, characterized in that: The device comprises: a processing module, used to obtain historical vehicle power data and historical operation data of a real vehicle, pre-process the historical vehicle power data and the historical operation data to obtain standardized data, and filter the standardized data to obtain target data, wherein the target data is used to represent data corresponding to any driving scenario of the real vehicle from start to end; A training module, used for establishing a short-time driving data prediction model based on the Mamba module, and training the short-time driving data prediction model using the target data to obtain a target short-time driving data prediction model; A prediction module, used to obtain vehicle power data and operation data of a real vehicle, input the vehicle power data and the operation data into a target short-time driving data prediction model for processing, and obtain a prediction result, wherein the prediction result refers to the speed, longitudinal acceleration and lateral acceleration of the real vehicle in a preset time period in the future; The recognition module is used to obtain preset rules corresponding to multiple driving intentions, and use the preset rules to identify the prediction results to obtain the target driving intention.

9. A driving intention recognition device, characterized in that: The device comprises: a memory and a processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor executes the driving intention recognition method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the driving intention recognition method as described in any one of claims 1 to 7.

Images