Vehicle parameter prediction method and device

Abstract

The present application discloses a vehicle parameter prediction method and device. The method can be applied to the field of intelligent driving technology, and specifically may include: extracting features from the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle; extracting features from the future driving area information of the target vehicle at a future time to obtain the speed-related area features of the target vehicle at a future time; predicting the future speed of the target vehicle at a future time based on the historical speed spatial features and the speed-related area features at a future time. When predicting the future speed of the target vehicle, the present application not only considers the historical speed information, but also considers the future driving area information at a future time, thereby increasing the accuracy of the future speed prediction.

Description

Technical Field

[0001] The present application relates to the field of intelligent driving technology, and in particular to a vehicle parameter prediction method and device. Background Art

[0002] The vehicle speed information reflects the driving status of the extended-range vehicle under different working conditions, and plays a very critical role in improving the fuel economy of the extended-range vehicle. Vehicle speed prediction is of great value to vehicle parameter prediction. Through vehicle speed prediction, the extended-range vehicle can intelligently manage power output according to the characteristics of the road and the needs of the driver. For example, when it is predicted that the vehicle is about to enter a downhill section, the extended-range vehicle can use the electric motor for regenerative braking and store energy to reduce dependence on the fuel engine, thereby saving fuel. In addition, when it is predicted that the vehicle needs to speed up, the extended-range vehicle can select the best power output method based on the cooperative working principle of the electric motor and the fuel engine to provide the required power with minimum fuel consumption.

[0003] With the rapid development of artificial intelligence technology, the current vehicle speed prediction in the vehicle parameter prediction process also applies deep learning methods, such as deep belief networks, stacked autoencoder neural networks, etc. However, this vehicle speed prediction method also lacks a reasonable modeling mechanism, resulting in poor accuracy of vehicle speed prediction. Summary of the invention

[0004] Based on this, it is necessary to provide a vehicle parameter prediction method and device that can accurately predict the vehicle speed in response to the above technical problems.

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

[0006] Extract the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle;

[0007] Extract features of the future driving area information of the target vehicle at the future time, and obtain the speed-related area features of the target vehicle at the future time;

[0008] The future speed of the target vehicle at a future time is predicted based on the historical speed spatial characteristics and the speed-related regional characteristics at a future time.

[0009] In one embodiment, the historical vehicle speed information includes vehicle speed information at each historical moment in the historical period;

[0010] The historical speed information of the target vehicle is extracted to obtain the historical speed spatial features of the target vehicle, including:

[0011] Process the vehicle speed information at each historical moment to obtain the vehicle speed similarity corresponding to each historical moment;

[0012] Based on the first Chebyshev graph convolutional network, the vehicle speed information at each historical moment is feature extracted according to the vehicle speed similarity corresponding to each historical moment, and the historical vehicle speed spatial features of the target vehicle are obtained.

[0013] In one embodiment, the vehicle speed information at each historical moment is processed to obtain the vehicle speed similarity corresponding to each historical moment, including:

[0014] Determine the cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment;

[0015] The cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment is used as the vehicle speed similarity corresponding to each historical moment.

[0016] In one embodiment, based on the first Chebyshev graph convolutional network, according to the speed similarity corresponding to each historical moment, feature extraction is performed on the speed information of each historical moment to obtain the historical speed spatial feature of the target vehicle, including:

[0017] According to the similarity of vehicle speeds corresponding to each historical moment, a first adjacency matrix is ​​constructed;

[0018] Based on the first Chebyshev graph convolutional network and the first adjacency matrix, the vehicle speed information at each historical moment is extracted to obtain the historical vehicle speed spatial features of the target vehicle.

[0019] In one of the embodiments, the first Chebyshev graph convolutional network includes at least two Chebyshev graph convolutional layers connected end to end;

[0020] Based on the Chebyshev graph convolutional network, the vehicle speed information at each historical moment is extracted according to the first adjacency matrix to obtain the historical vehicle speed spatial features of the target vehicle, including:

[0021] The vehicle speed information at each historical moment is input into the first Chebyshev graph convolution layer, and the first adjacency matrix is ​​input into each Chebyshev graph convolution layer respectively, so as to obtain the vehicle speed spatial features output by each Chebyshev graph convolution layer based on the input data;

[0022] The speed spatial features output by the final Chebyshev graph convolutional layer are used as the historical speed spatial features of the target vehicle;

[0023] Among them, the input data of the first Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed information at each historical moment; the input data of each remaining Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed spatial features output by the previous Chebyshev graph convolution layer of the Chebyshev graph convolution layer.

[0024] In one embodiment, feature extraction is performed on the future driving area information of the target vehicle at the future time to obtain the speed-related area features of the target vehicle at the future time, including:

[0025] Based on the feature extraction network, feature extraction is performed on the future driving area information of the target vehicle at the future time, and a regional feature matrix of the future driving area information is obtained;

[0026] Determine a second adjacency matrix according to the regional feature matrix and a transposed matrix corresponding to the regional feature matrix;

[0027] The vehicle speed associated regional features of the target vehicle at a future time are determined according to the second adjacency matrix and the regional feature matrix.

[0028] In one embodiment, determining the second adjacency matrix according to the regional feature matrix and the transposed matrix corresponding to the regional feature matrix includes:

[0029] The regional feature matrix and the transposed matrix corresponding to the regional feature matrix are multiplied to obtain an initial adjacency matrix;

[0030] The initial adjacency matrix is ​​preprocessed to obtain a second adjacency matrix; wherein the preprocessing includes normalization processing and sorting selection processing.

[0031] In one embodiment, determining the speed-related regional features of the target vehicle at a future time according to the second adjacency matrix and the regional feature matrix includes:

[0032] The second adjacency matrix and the regional feature matrix are fused to obtain regional fusion features;

[0033] Based on the second Chebyshev graph convolutional network, feature extraction is performed on the regional fusion features to obtain the speed-related regional features of the target vehicle at the future moment.

[0034] In one embodiment, predicting the future speed of the target vehicle at a future time according to the historical speed spatial characteristics and the speed associated area characteristics at a future time includes:

[0035] Based on the recurrent neural network unit, the future speed of the target vehicle at the future moment is predicted according to the second adjacency matrix, the spatial characteristics of the historical speeds and the characteristics of the speed-related regions at the future moment;

[0036] Among them, the recurrent neural network unit includes a third Chebyshev graph convolutional network; the third Chebyshev graph convolutional network is used to fuse the second adjacency matrix and the preliminary fusion features; the preliminary fusion features are obtained by fusing the historical vehicle speed spatial features and the vehicle speed associated area features at future moments.

[0037] In one embodiment, the method further comprises:

[0038] Determine the SOC value of the target vehicle at each future time according to the future speed of the target vehicle at each future time;

[0039] Based on the state variable update function of the SOC value and the output power of the range extender, and the cost function of the SOC value and the output power of the range extender, the optimal control parameters of the range extender of the target vehicle at each future moment are determined according to the SOC value of the target vehicle at each future moment.

[0040] In a second aspect, the present application provides a vehicle parameter prediction method, the device comprising:

[0041] The first extraction module is used to extract the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle;

[0042] The second extraction module is used to extract features of the future driving area information of the target vehicle at the future time, and obtain the speed-related area features of the target vehicle at the future time;

[0043] The prediction module is used to predict the future speed of the target vehicle at a future moment according to the historical speed spatial characteristics and the speed-related regional characteristics at a future moment.

[0044] In a third aspect, the present application further provides a computer device, the computer device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the following steps when executing the computer program:

[0045] Extract the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle;

[0046] Extract features of the future driving area information of the target vehicle at the future time, and obtain the speed-related area features of the target vehicle at the future time;

[0047] The future speed of the target vehicle at a future time is predicted based on the historical speed spatial characteristics and the speed-related regional characteristics at a future time.

[0048] In a fourth aspect, the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0049] Extract the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle;

[0050] Extract features of the future driving area information of the target vehicle at the future time, and obtain the speed-related area features of the target vehicle at the future time;

[0051] The future speed of the target vehicle at a future time is predicted based on the historical speed spatial characteristics and the speed-related regional characteristics at a future time.

[0052] In a fifth aspect, the present application further provides a computer program product, the computer program product comprising a computer program, and when the computer program is executed by a processor, the following steps are implemented:

[0053] Extract the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle;

[0054] Extract features of the future driving area information of the target vehicle at the future time, and obtain the speed-related area features of the target vehicle at the future time;

[0055] The future speed of the target vehicle at a future time is predicted based on the historical speed spatial characteristics and the speed-related regional characteristics at a future time.

[0056] The above-mentioned vehicle parameter prediction method and device obtain the historical speed spatial features of the target vehicle by extracting features from the historical speed information of the target vehicle. Then, feature extraction is performed on the future driving area information of the target vehicle at a future moment to obtain the speed-related area features of the target vehicle at a future moment. Finally, based on the historical speed spatial features and the speed-related area features at a future moment, the future speed of the target vehicle at a future moment is predicted. Compared with the traditional speed prediction method, when predicting the future speed of the target vehicle, the present application not only considers the historical speed information, but also considers the future driving area information at a future moment, thereby increasing the accuracy of the future speed prediction. The present application also extracts features based on the historical speed information to obtain the historical speed spatial features of the target vehicle, and based on the feature extraction of the future driving area information at a future moment, obtains the speed-related area features of the target vehicle at a future moment associated with the speed of the target vehicle, and performs speed prediction based on the historical speed spatial features and the speed-related area features at a future moment, thereby further increasing the accuracy of the future speed prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 An application environment diagram of a vehicle parameter prediction method provided in this embodiment;

[0058] Figure 2 A schematic diagram of a flow chart of a first vehicle parameter prediction method provided in this embodiment;

[0059] Figure 3 A schematic diagram of a process for determining the historical speed spatial characteristics of a target vehicle provided in this embodiment;

[0060] Figure 4A schematic diagram of the principle of determining the historical speed spatial characteristics of a target vehicle provided in this embodiment;

[0061] Figure 5 A schematic diagram of a process for determining the speed-related area characteristics of a target vehicle at a future time provided in this embodiment;

[0062] Figure 6 A schematic diagram of the principle of determining the vehicle speed associated area characteristics of the target vehicle at a future time provided in this embodiment;

[0063] Figure 7 A schematic diagram of the internal principle of the recurrent neural network unit provided for this embodiment;

[0064] Figure 8 A schematic diagram of a flow chart of a second vehicle parameter prediction method provided in this embodiment;

[0065] Fig. 9 A schematic diagram of a flow chart of a third vehicle parameter prediction method provided in this embodiment;

[0066] Fig.10 A structural block diagram of a vehicle parameter prediction device provided in this embodiment;

[0067] Fig.11 This is a diagram of the internal structure of the computer device provided in this embodiment. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0069] The path updating method provided in the embodiment of the present application can be applied to the situation of how to predict vehicle parameters, and in particular, to the situation of how to predict the future speed of a vehicle. Optionally, the vehicle parameter prediction method can be executed by the target vehicle, or can be executed by a server. For example, the target vehicle / server can predict the future speed based on a preset vehicle parameter prediction logic, combined with the historical speed information and future driving area information of the target vehicle. Furthermore, the vehicle parameter prediction method can also be executed by the interaction between the target vehicle and the server, that is, the method can be applied to the following situations: Figure 1In the application environment shown. Wherein, when there is a need to predict vehicle parameters, the server 104 interacts with the target vehicle 102, obtains the historical speed information of the target vehicle 102 and the future driving area information of the target vehicle 102 at a future time from the target vehicle 102, and extracts features from the historical speed information of the target vehicle 102 to obtain the historical speed spatial features of the target vehicle. Then, feature extraction is performed on the future driving area information of the target vehicle 102 at a future time to obtain the speed-related area features of the target vehicle 102 at a future time. The server 104 predicts the future speed of the target vehicle 102 at a future time based on the historical speed spatial features and the speed-related area features at a future time.

[0070] The server 104 can be implemented as an independent server or a server cluster composed of multiple servers. The target vehicle can be an ordinary car, an electric car, or a vehicle with an automatic driving function. The present application does not limit the type of the target vehicle.

[0071] In one embodiment, Figure 2 is a flow chart of a vehicle parameter prediction method provided according to an embodiment of the present application, in which the method is applied to Figure 1 Taking the server in as an example, the method includes the following steps:

[0072] S201, extracting features from historical speed information of a target vehicle to obtain spatial features of historical speed of the target vehicle.

[0073] The target vehicle refers to the vehicle for which vehicle parameters prediction is required, which can be an ordinary car, an electric car, or a vehicle with an automatic driving function. The historical speed information refers to the speed information of the target vehicle before the current moment (i.e., the historical period). The historical speed spatial feature refers to the feature of spatial information feature extraction based on the historical speed, which can characterize the correlation of speed information at each historical moment.

[0074] An optional implementation of this embodiment is: based on the feature extraction network, feature extraction is performed on the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle. In other words, the historical speed information of the target vehicle is input into the feature extraction network, and the feature extraction network outputs the historical speed spatial features of the target vehicle.

[0075] Another optional implementation of this embodiment is: according to the historical speed information of the target vehicle, the speed space vector information of each historical moment is obtained, and based on the speed space vector information of each historical moment, the historical speed space characteristics of the target vehicle are determined.

[0076] S202, extracting features of the future driving area information of the target vehicle at the future time, and obtaining the speed-related area features of the target vehicle at the future time.

[0077] Among them, the future moment refers to the moment after the current moment. The future driving area information refers to the driving area information of the target vehicle at the future moment. The future driving area information includes road information, terrain information, regional weather information and driving operation information (such as the driver's operating habits), etc. The road information includes lane line information, speed limit information, traffic flow information, congestion information and obstacle information, etc. Terrain information includes slope information, surface morphology, surface coverage, etc. The speed-related regional features refer to the relevant features that have a greater impact on the speed of the target vehicle. Regional weather information includes weather type and corresponding severity, such as rainy days, and the corresponding rainfall amount on rainy days.

[0078] An optional implementation of this embodiment is: extracting features of the future driving area information of the target vehicle at the future time through a feature extraction network to obtain the speed-related area features of the target vehicle at the future time. Specifically, the future driving area information of the target vehicle at the future time is input into the feature extraction network, and the feature extraction network extracts features of the future driving area information of the target vehicle at the future time, and outputs the speed-related area features of the target vehicle at the future time.

[0079] Another optional implementation of this embodiment is: obtaining the future driving area information of the target vehicle at a future time, extracting features in the future driving area information, including extracting weather information, road information, terrain information and driving operation information, and obtaining the speed-related area features of the target vehicle at a future time. The driving operation information can be determined based on the historical operating habits of the driver in the driving area.

[0080] S203, predicting the future vehicle speed of the target vehicle at a future time according to the historical vehicle speed spatial characteristics and the vehicle speed associated area characteristics at a future time.

[0081] An optional implementation of this embodiment is: based on the neural network model, the future speed of the target vehicle at a future moment is predicted according to the historical speed spatial features and the speed-related area features at a future moment. Specifically, the historical speed spatial features and the speed-related area features at a future moment are input into the neural network model, and the neural network model outputs the future speed of the target vehicle at a future moment. The neural network model may be, but is not limited to, a recurrent neural network unit, i.e., GRU (Gate Recurrent Unit).

[0082] Another optional implementation of this embodiment is: according to the historical vehicle speed spatial characteristics, the speed change correlation of the speed information of the target vehicle at each historical moment is determined, and the future speed of the target vehicle at the future moment is predicted based on the speed change correlation and the speed correlation area characteristics at the future moment.

[0083] The above vehicle parameter prediction method obtains the historical speed spatial features of the target vehicle by extracting features from the historical speed information of the target vehicle. Then, feature extraction is performed on the future driving area information of the target vehicle at a future moment to obtain the speed-related area features of the target vehicle at a future moment. Finally, based on the historical speed spatial features and the speed-related area features at a future moment, the future speed of the target vehicle at a future moment is predicted. Compared with the traditional speed prediction method, the present application not only considers the historical speed information but also considers the future driving area information at a future moment when predicting the future speed of the target vehicle, thereby increasing the accuracy of the future speed prediction. The present application also extracts features based on the historical speed information to obtain the historical speed spatial features of the target vehicle, and based on the feature extraction of the future driving area information at a future moment, obtains the speed-related area features of the target vehicle at a future moment associated with the speed of the target vehicle, and performs speed prediction based on the historical speed spatial features and the speed-related area features at a future moment, thereby further increasing the accuracy of the future speed prediction. In addition, the future driving area information of the present application includes road information, terrain information, regional weather information and driving operation information, and more factors affecting the future vehicle speed are considered, which further increases the accuracy of future vehicle speed prediction.

[0084] In one embodiment, the historical vehicle speed information includes the vehicle speed information at each historical moment in the historical period. In order to increase the accuracy of the obtained historical vehicle speed spatial features, Figure 3 As shown, an optional implementation of S201 includes:

[0085] S301, processing the vehicle speed information at each historical moment to obtain the vehicle speed similarity corresponding to each historical moment.

[0086] The vehicle speed similarity refers to the similarity between the vehicle speed information at each historical moment.

[0087] Optionally, in this embodiment, a reference moment is first selected, and the cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment is determined based on the vehicle speed information at the selected reference moment. The cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment is used as the vehicle speed similarity corresponding to each historical moment.

[0088] In this embodiment, an optional implementation method for selecting a reference time is: selecting a historical moment in a historical period as a reference moment, or selecting a moment outside the historical period as a reference moment.

[0089] S302, based on the first Chebyshev graph convolutional network, according to the vehicle speed similarity corresponding to each historical moment, feature extraction is performed on the vehicle speed information of each historical moment to obtain the historical vehicle speed spatial feature of the target vehicle.

[0090] Among them, the first Chebyshev graph convolutional network refers to one of the Chebyshev graph convolutional networks in this application, and "first" is used to distinguish different Chebyshev graph convolutional networks. The Chebyshev graph convolutional network is a deep learning model based on a graph structure, which is used to process graph data, especially image data. The Chebyshev graph convolutional network mainly models the local structure of image data, and uses Chebyshev polynomials to approximate the convolution operation, thereby achieving effective representation and feature extraction of image data. Compared with traditional convolutional neural networks (such as CNN (Convolutional Neural Networks, convolutional neural networks)), Chebyshev graph convolutional networks are more suitable for processing non-Euclidean graph data.

[0091] Optionally, in this embodiment, a first adjacency matrix is ​​constructed according to the similarity of vehicle speeds corresponding to each historical moment, that is, the eigenvalues ​​in the first adjacency matrix are the similarity of vehicle speeds corresponding to each historical moment. Based on the first Chebyshev graph convolutional network, feature extraction is performed on the vehicle speed information at each historical moment according to the first adjacency matrix to obtain the historical vehicle speed spatial features of the target vehicle. Among them, the first adjacency matrix refers to the adjacency matrix constructed by the similarity of vehicle speeds corresponding to each historical moment.

[0092] Optionally, in order to further increase the accuracy of the historical vehicle speed spatial features, the first Chebyshev graph convolution network in this embodiment includes at least two Chebyshev graph convolution layers connected end to end, and preferably includes three Chebyshev graph convolution layers connected end to end.

[0093] In this embodiment, Figure 4As shown, based on the first Chebyshev graph convolution network, according to the first adjacency matrix, the vehicle speed information at each historical moment is feature extracted to obtain the historical vehicle speed spatial features of the target vehicle. An optional implementation method is: the vehicle speed information at each historical moment is input into the first Chebyshev graph convolution layer, and the first adjacency matrix is ​​input into each Chebyshev graph convolution layer respectively, to obtain the vehicle speed spatial features output by each Chebyshev graph convolution layer based on the input data. The vehicle speed spatial features output by the last Chebyshev graph convolution layer are used as the historical vehicle speed spatial features of the target vehicle. Among them, the input data of the first Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed information at each historical moment; the input data of each of the remaining Chebyshev graph convolution layers includes the first adjacency matrix and the vehicle speed spatial features output by the previous Chebyshev graph convolution layer of the Chebyshev graph convolution layer. In this embodiment, the historical vehicle speed spatial features can be determined by the following formula (1):

[0094] (1)

[0095] Among them, H v Represents the spatial characteristics of historical vehicle speed; A v represents the first adjacency matrix; V i Indicates the vehicle speed information at each historical moment; W （0） is the weight parameter of the first Chebyshev graph convolution layer, W （1） is the weight parameter of the second Chebyshev graph convolution layer, W （2） is the weight parameter of the third Chebyshev graph convolutional layer, and σ is the ReLU activation function.

[0096] In this embodiment, the cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment is determined. The cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment is used as the vehicle speed similarity corresponding to each historical moment. According to the vehicle speed similarity corresponding to each historical moment, a first adjacency matrix is ​​constructed. The vehicle speed information at each historical moment is input into the first Chebyshev graph convolution layer, and the first adjacency matrix is ​​input into each Chebyshev graph convolution layer respectively, to obtain the vehicle speed spatial features output by each Chebyshev graph convolution layer based on the input data, and the vehicle speed spatial features output by the last Chebyshev graph convolution layer are used as the historical vehicle speed spatial features of the target vehicle. Based on the execution method of this embodiment, the accuracy of the historical vehicle speed spatial features of the target vehicle can be effectively increased, thereby increasing the accuracy of the predicted future vehicle speed of the target vehicle.

[0097] In one embodiment, in order to increase the accuracy of the determined vehicle speed associated area features, such as Figure 5 As shown, an optional implementation of S203 includes:

[0098] S501, based on the feature extraction network, extract features of the future driving area information of the target vehicle at the future time, and obtain a regional feature matrix of the future driving area information.

[0099] The feature extraction network refers to a neural network model that can extract useful regional features (i.e., features that affect the speed of the target vehicle) from the future driving area information, including but not limited to a CNN network. The regional feature matrix refers to a feature matrix constructed by the feature extraction network based on the useful regional features extracted from the future driving area information of the target vehicle at a future time.

[0100] An optional implementation of this embodiment is: the future driving area information of the target vehicle at a future time is input into the feature extraction network, and the feature extraction network extracts and outputs useful area features based on the future driving area information. Based on the useful area features, a regional feature matrix of the future driving area information is constructed.

[0101] Another optional implementation of this embodiment is: the future driving area information of the target vehicle at a future time is input into the feature extraction network, and the feature extraction network outputs the regional feature matrix of the future driving area information based on the future driving area information. The regional feature matrix can be determined by the following formula (2):

[0102] (2)

[0103] Among them, X t Indicates the future driving area information of the target vehicle at a future time; The regional feature matrix representing the future driving area information.

[0104] S502: Determine a second adjacency matrix according to the regional feature matrix and the transposed matrix corresponding to the regional feature matrix.

[0105] The transposed matrix refers to a matrix obtained by transposing the regional feature matrix; and the second adjacency matrix refers to an adjacency matrix associated with future driving area information at a future time.

[0106] An optional implementation of this embodiment is: transpose the regional feature matrix to obtain a transposed matrix corresponding to the regional feature matrix, and use the product of the regional feature matrix and the transposed matrix as the second adjacency matrix.

[0107] Another optional implementation of this embodiment is: transpose the regional feature matrix to obtain the transposed matrix corresponding to the regional feature matrix. Multiply the regional feature matrix and the transposed matrix corresponding to the regional feature matrix to obtain an initial adjacency matrix. Preprocess the initial adjacency matrix to obtain a second adjacency matrix; wherein the preprocessing includes normalization and sorting selection. An optional implementation of preprocessing the initial adjacency matrix to obtain the second adjacency matrix is: normalize the initial adjacency matrix, and then sort and select the normalized initial adjacency matrix to obtain the second adjacency matrix. This embodiment can determine the second adjacency matrix by the following formula (3):

[0108] (3)

[0109] Among them, in formula (3) A regional feature matrix representing the future driving area information; represents the transposed matrix of the regional feature matrix, Indicates normalization processing; represents the sorting mechanism, i.e., sort selection processing; Represents the second adjacency matrix.

[0110] S503: Determine the vehicle speed-related regional features of the target vehicle at a future time according to the second adjacency matrix and the regional feature matrix.

[0111] An optional implementation of this embodiment is: fusing the second adjacency matrix and the regional feature matrix, and using the result obtained after the fusion process as the speed-related regional feature of the target vehicle at the future moment.

[0112] Another optional implementation of this embodiment is: fusing the second adjacency matrix and the regional feature matrix to obtain regional fusion features. Figure 6 As shown, based on the second Chebyshev graph convolutional network, feature extraction is performed on the regional fusion features to obtain the speed-related regional features of the target vehicle at the future time. In this embodiment, the speed-related regional features of the target vehicle at the future time can be determined by the following formula (4):

[0113] (4)

[0114] Among them, in the above formula (4) Indicates the characteristics of the vehicle speed-related area; A regional feature matrix representing the future driving area information; represents the second adjacency matrix; W （3） Represents the weight coefficients of the second Chebyshev graph convolutional network.

[0115] In this embodiment, based on the feature extraction network, feature extraction is performed on the future driving area information to obtain a regional feature matrix. Then, the regional feature matrix and the transposed matrix corresponding to the regional feature matrix are multiplied to obtain an initial adjacency matrix. And based on the preprocessing of the initial adjacency matrix, a second adjacency matrix is ​​obtained. The second adjacency matrix and the regional feature matrix are fused to obtain regional fusion features. Finally, based on the second Chebyshev graph convolution network, feature extraction is performed on the regional fusion features to obtain the speed-related regional features of the target vehicle at the future moment. In this embodiment, the regional feature matrix is ​​determined based on the future driving area information, which can make the influencing factors of the future speed of the target vehicle (i.e., the speed-related regional features) determined more accurate, and can increase the accuracy of the predicted future speed.

[0116] In one embodiment, in order to further increase the accuracy of predicting the future speed of the target vehicle at a future time, an optional implementation of S203 includes:

[0117] Based on the recurrent neural network unit, the future speed of the target vehicle at the future moment is predicted according to the second adjacency matrix, the historical speed spatial characteristics and the speed associated area characteristics at the future moment.

[0118] The recurrent neural network unit includes a third Chebyshev graph convolutional network; the third Chebyshev graph convolutional network is used to fuse the second adjacency matrix and the preliminary fusion feature; the preliminary fusion feature is obtained by fusing the historical vehicle speed spatial feature and the vehicle speed associated area feature at the future moment. The recurrent neural network unit is a neural network structure that can be used to process sequence data, such as GRU.

[0119] Optionally, in this embodiment, based on the recurrent neural network unit, according to the second adjacency matrix, the historical speed spatial features and the speed associated area features at the future moment, the optional implementation method of predicting the future speed of the target vehicle at the future moment is: Figure 7 As shown, the second adjacency matrix, the historical vehicle speed spatial features, and the vehicle speed associated area features at future moments are input into the recurrent neural network unit. Based on the recurrent neural network unit, the historical vehicle speed spatial features and the vehicle speed associated area features at future moments are fused (preferably by additive fusion processing) to obtain preliminary fusion features. Based on the third Chebyshev graph convolutional network, the target fusion features are determined according to the second adjacency matrix and the preliminary fusion features. The aggregate features are determined according to the target fusion features and the output features of the recurrent neural network unit at the previous moment. The future vehicle speed of the target vehicle at the future moment is predicted according to the aggregate features, the preliminary fusion features, the target fusion features, and the second adjacency matrix.

[0120] In this embodiment, the target fusion feature can be determined by the following formula (5):

[0121] (5)

[0122] Among them, in formula (5) represents the target fusion feature; represents the preliminary fusion features; represents the second adjacency matrix; W （4） is the weight parameter of the third Chebyshev graph convolutional network.

[0123] In this embodiment, the aggregation characteristics can be determined by the following formula (6):

[0124] (6)

[0125] Among them, in formula (6) represents the target fusion feature, It represents the output features of the recurrent neural network unit at the previous moment. That is to say, this application predicts the future speed at a future moment (i.e., moment t). When predicting the vehicle speed at a future moment, it is necessary to combine the output features of the recurrent neural network unit at moment t-1 for prediction.

[0126] Optionally, in this embodiment, according to the aggregated features, the preliminary fusion features, the target fusion features and the second adjacency matrix, an optional implementation method for predicting the future speed of the target vehicle at a future time is as follows: Figure 7 As shown, the reset gate parameters of the recurrent neural network unit are determined based on the preliminary fusion features and the output features of the recurrent neural network unit at the previous moment. The output candidate features of the recurrent neural network unit are determined based on the reset gate parameters, the target fusion features, and the output features of the recurrent neural network unit at the previous moment. The update gate parameters of the recurrent neural network unit are determined based on the aggregate features and the update gate activation function of the recurrent neural network unit. The future speed of the target vehicle at a future moment is predicted based on the update gate parameters of the recurrent neural network unit, the output candidate features, and the output features of the recurrent neural network unit at the previous moment.

[0127] In this embodiment, the reset gate parameter of the recurrent neural network unit can be determined by the following formula (7):

[0128] (7)

[0129] Among them, in formula (7) Represents the reset gate parameters of the recurrent neural network unit; represents the preliminary fusion features; Represents the output feature of the recurrent neural network unit at the previous moment; W represents the weight matrix of the reset gate; Represents the activation function of the reset gate.

[0130] In this embodiment, the output candidate features of the recurrent neural network unit can be determined by the following formula (8):

[0131] (8)

[0132] Among them, in formula (8) Represents the reset gate parameters of the recurrent neural network unit; represents the preliminary fusion features; Represents the output features of the recurrent neural network unit at the previous moment; represents the target fusion feature; Represents the output candidate features of the recurrent neural network unit.

[0133] In this embodiment, the update gate parameter of the recurrent neural network unit can be determined by the following formula (9):

[0134] (9)

[0135] Among them, in formula (9) Represents aggregate features; represents the activation function of the update gate; Represents updating gate parameters.

[0136] Optionally, in this embodiment, based on the update gate parameters of the recurrent neural network unit, the output candidate features and the output features of the recurrent neural network unit at the previous moment, an optional implementation method for predicting the future speed of the target vehicle at the future moment is as follows: Figure 7 As shown, according to the update gate parameter of the recurrent neural network unit, the first weight coefficient of the output candidate feature and the second weight coefficient of the output feature of the recurrent neural network unit at the previous moment are determined. According to the first weight coefficient and the second weight coefficient, the output candidate feature and the output feature of the recurrent neural network unit at the previous moment are weighted to obtain a weighted processing result. The weighted processing result is input into the fully connected layer of the recurrent neural network unit to obtain the future speed of the target vehicle at the future moment.

[0137] Optionally, in this embodiment, according to the update gate parameter of the recurrent neural network unit, an optional implementation method for determining the first weight coefficient of the output candidate feature and the second weight coefficient of the output feature of the recurrent neural network unit at the previous moment is: using the update gate parameter of the recurrent neural network unit as the second weight coefficient of the output feature of the recurrent neural network unit at the previous moment. The difference between 1 and the update gate parameter of the recurrent neural network unit is used as the first weight coefficient of the output candidate feature.

[0138] In this embodiment, the weighted processing result can be determined by the following formula (10):

[0139] (10)

[0140] In formula (10), Indicates updating gate parameters; Represents the output candidate features of the recurrent neural network unit; Represents the output features of the recurrent neural network unit at the previous moment.

[0141] In this embodiment, the recurrent neural network unit is improved, and a third Chebyshev graph convolution network is added to the recurrent neural network unit. The third Chebyshev graph convolution network is used to fuse the second adjacency matrix and the preliminary fusion features; the preliminary fusion features are obtained by fusing the historical vehicle speed spatial features and the vehicle speed associated area features at future times. Based on the improved recurrent neural network unit and the third Chebyshev graph convolution network inside the recurrent neural network unit, the future speed of the target vehicle at the future time is predicted according to the second adjacency matrix, the historical vehicle speed spatial features and the vehicle speed associated area features at future times, so that the future speed of the target vehicle at the future time obtained is more accurate.

[0142] In one embodiment, in order to predict the optimal control parameters of the target vehicle at each future time based on the predicted future vehicle speed at the future time, such as Figure 8 As shown, an optional implementation of a vehicle parameter prediction method includes:

[0143] S801, determining the SOC value of the target vehicle at each future moment according to the future speed of the target vehicle at each future moment.

[0144] Among them, the SOC (State of Charge, battery state of charge, also called remaining power) value refers to the predicted remaining power of the target vehicle at a future moment.

[0145] Optionally, in this embodiment, the future speed of the target vehicle at each future moment can be input into a trained SOC prediction model, and the SOC prediction model outputs the SOC value of the target vehicle at each future moment. In other words, in this application, the SOC value corresponding to each future moment can be predicted by the future speed at each future moment.

[0146] S802, based on the state variable update function of the SOC value and the output power of the range extender, and the cost function of the SOC value and the output power of the range extender, according to the SOC value of the target vehicle at each future moment, determine the optimal control parameters of the range extender of the target vehicle at each future moment.

[0147] Among them, the range extender refers to an electric vehicle component that can provide additional electrical energy, thereby enabling the electric vehicle to increase its mileage. The range extender output power refers to the output power of the range extender of the target vehicle. The state variable update equation refers to the equation of the correlation between the range extender output power and the SOC value. The optimal control parameters are only the control parameters that can achieve the minimum fuel consumption of the range extender. The cost function refers to the conversion cost function between the output power of the range extender of the target vehicle and the SOC value.

[0148] Optionally, in this embodiment, a global optimization can be performed based on a DP (Dynamic Programming) algorithm to establish a state variable update function (11) of the SOC value and the output power of the range extender:

[0149] SOC(t+1) = f(SOC(t), P(t)) (11)

[0150] In formula (11), SOC(t+1) represents the SOC value at time t+1; SOC(t) represents the SOC value at time t; and P(t) represents the output power of the range extender at time t.

[0151] In this embodiment, the cost function of the SOC value and the range extender output power can be expressed by the following formula (12):

[0152] J = m fuel (SOC,P) (12)

[0153] Where, J in formula (12) represents the cost function; m fuel is the fuel consumption per unit time; SOC represents the SOC value; P represents the output power of the range extender.

[0154] Exemplarily, in this embodiment, based on the state variable update function of the SOC value and the output power of the range extender (i.e., formula (11)), and the cost function of the SOC value and the output power of the range extender (i.e., formula (12)), according to the SOC value of the target vehicle at each future moment, the implementation method of determining the optimal control parameters of the range extender of the target vehicle at each future moment is as follows: according to the future vehicle speed at each future moment, the SOC value at each future moment can be predicted, and according to the state variable update function of the SOC value and the output power of the range extender and the cost function of the SOC value and the output power of the range extender, reverse deduction is performed to determine the optimal control parameters of the output power of the range extender at each future moment, and based on the optimal control parameters, the minimum fuel consumption of the target vehicle can be achieved, which can effectively improve the fuel economy of the target vehicle.

[0155] In one embodiment, if Fig. 9 As shown, another optional implementation of the vehicle parameter prediction method includes:

[0156] S901, determining the cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment.

[0157] S902: taking the cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment as the vehicle speed similarity corresponding to each historical moment.

[0158] S903: construct a first adjacency matrix according to the vehicle speed similarities corresponding to each historical moment.

[0159] S904: Input the first adjacency matrix to each Chebyshev graph convolution layer respectively, and input the vehicle speed information at each historical moment to the first Chebyshev graph convolution layer to obtain the vehicle speed spatial features output by each Chebyshev graph convolution layer based on the input data.

[0160] S905, using the speed spatial features output by the last Chebyshev graph convolution layer as the historical speed spatial features of the target vehicle. The input data of the first Chebyshev graph convolution layer includes the first adjacency matrix and the speed information at each historical moment; the input data of each remaining Chebyshev graph convolution layer includes the first adjacency matrix and the speed spatial features output by the previous Chebyshev graph convolution layer of the Chebyshev graph convolution layer.

[0161] S906, based on the feature extraction network, extract features of the future driving area information of the target vehicle at the future time, and obtain a regional feature matrix of the future driving area information.

[0162] S907, multiplying the regional feature matrix and the transposed matrix corresponding to the regional feature matrix to obtain an initial adjacency matrix.

[0163] S908, preprocessing the initial adjacency matrix to obtain a second adjacency matrix, wherein the preprocessing includes normalization processing and sorting and selection processing.

[0164] S909, fusing the second adjacency matrix and the regional feature matrix to obtain regional fusion features.

[0165] S9010, based on the second Chebyshev graph convolutional network, extracts the regional fusion features to obtain the speed-related regional features of the target vehicle at the future moment.

[0166] S9011, based on the recurrent neural network unit, predict the future speed of the target vehicle at the future moment according to the second adjacency matrix, the historical speed spatial features and the speed-related regional features at the future moment. The recurrent neural network unit includes a third Chebyshev graph convolutional network; the third Chebyshev graph convolutional network is used to fuse the second adjacency matrix and the preliminary fusion features; the preliminary fusion features are obtained by fusing the historical speed spatial features and the speed-related regional features at the future moment.

[0167] S9012, determining the SOC value of the target vehicle at each future moment according to the future speed of the target vehicle at each future moment.

[0168] S9013, based on the state variable update function of the SOC value and the output power of the range extender, and the cost function of the SOC value and the output power of the range extender, according to the SOC value of the target vehicle at each future moment, determine the optimal control parameters of the range extender of the target vehicle at each future moment.

[0169] The vehicle parameter prediction method of this embodiment obtains the historical speed spatial feature of the target vehicle by extracting the feature of the historical speed information of the target vehicle. Then, the feature of the future driving area information of the target vehicle at the future moment is extracted to obtain the speed-related area feature of the target vehicle at the future moment. Finally, based on the historical speed spatial feature and the speed-related area feature at the future moment, the future speed of the target vehicle at the future moment is predicted. Compared with the traditional speed prediction method, when predicting the future speed of the target vehicle, the present application not only considers the historical speed information, but also considers the future driving area information at the future moment, thereby increasing the accuracy of the future speed prediction. The present application also extracts the feature based on the historical speed information to obtain the historical speed spatial feature of the target vehicle, and extracts the feature based on the future driving area information at the future moment to obtain the speed-related area feature of the target vehicle at the future moment associated with the speed of the target vehicle, and performs speed prediction based on the historical speed spatial feature and the speed-related area feature at the future moment, thereby further increasing the accuracy of the future speed prediction.

[0170] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0171] Based on the same inventive concept, the embodiment of the present application also provides a vehicle parameter prediction device for implementing the vehicle parameter prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more vehicle parameter prediction device embodiments provided below can refer to the limitations of the vehicle parameter prediction method above, and will not be repeated here.

[0172] In one embodiment, by Fig.10 FIG. 2 shows a structural block diagram of a vehicle parameter prediction device in one embodiment. Fig.10 As shown, a vehicle parameter prediction device 1 is provided, the device comprising: a first extraction module 10, a second extraction module 20 and a prediction module 30, wherein:

[0173] The first extraction module 10 is used to extract the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle;

[0174] The second extraction module 20 is used to extract features of the future driving area information of the target vehicle at a future time, and obtain the speed-related area features of the target vehicle at a future time;

[0175] The prediction module 30 is used to predict the future speed of the target vehicle at a future time according to the historical speed spatial characteristics and the speed associated area characteristics at a future time.

[0176] In one embodiment, the historical vehicle speed information includes the vehicle speed information at each historical moment in the historical period. Fig.10 The first extraction module 10 is further specifically used for:

[0177] Process the vehicle speed information at each historical moment to obtain the vehicle speed similarity corresponding to each historical moment;

[0178] Based on the first Chebyshev graph convolutional network, the vehicle speed information at each historical moment is feature extracted according to the vehicle speed similarity corresponding to each historical moment, and the historical vehicle speed spatial features of the target vehicle are obtained.

[0179] In one embodiment, the Fig.10 The first extraction module 10 is further specifically used for:

[0180] Determine the cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment;

[0181] The cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment is used as the vehicle speed similarity corresponding to each historical moment.

[0182] In one embodiment, the Fig.10 The first extraction module 10 is further specifically used for:

[0183] According to the similarity of vehicle speeds corresponding to each historical moment, a first adjacency matrix is ​​constructed;

[0184] Based on the first Chebyshev graph convolutional network and the first adjacency matrix, the vehicle speed information at each historical moment is extracted to obtain the historical vehicle speed spatial features of the target vehicle.

[0185] In one embodiment, the first Chebyshev graph convolutional network includes at least two Chebyshev graph convolutional layers connected end to end, on this basis, Fig.10 The first extraction module 10 is further specifically used for:

[0186] The vehicle speed information at each historical moment is input into the first Chebyshev graph convolution layer, and the first adjacency matrix is ​​input into each Chebyshev graph convolution layer respectively, so as to obtain the vehicle speed spatial features output by each Chebyshev graph convolution layer based on the input data;

[0187] The speed spatial features output by the final Chebyshev graph convolutional layer are used as the historical speed spatial features of the target vehicle;

[0188] Among them, the input data of the first Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed information at each historical moment; the input data of each remaining Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed spatial features output by the previous Chebyshev graph convolution layer of the Chebyshev graph convolution layer.

[0189] In one embodiment, the Fig.10 The second extraction module 20 is further specifically used for:

[0190] Based on the feature extraction network, feature extraction is performed on the future driving area information of the target vehicle at the future time, and a regional feature matrix of the future driving area information is obtained;

[0191] Determine a second adjacency matrix according to the regional feature matrix and a transposed matrix corresponding to the regional feature matrix;

[0192] The vehicle speed associated regional features of the target vehicle at a future time are determined according to the second adjacency matrix and the regional feature matrix.

[0193] In one embodiment, the Fig.10 The second extraction module 20 is further specifically used for:

[0194] The regional feature matrix and the transposed matrix corresponding to the regional feature matrix are multiplied to obtain an initial adjacency matrix;

[0195] The initial adjacency matrix is ​​preprocessed to obtain a second adjacency matrix; wherein the preprocessing includes normalization processing and sorting selection processing.

[0196] In one embodiment, the Fig.10 The second extraction module 20 is further specifically used for:

[0197] The second adjacency matrix and the regional feature matrix are fused to obtain regional fusion features;

[0198] Based on the second Chebyshev graph convolutional network, feature extraction is performed on the regional fusion features to obtain the speed-related regional features of the target vehicle at the future moment.

[0199] In one embodiment, the Fig.10 The prediction module 30 is also specifically used for:

[0200] Based on the recurrent neural network unit, the future speed of the target vehicle at the future moment is predicted according to the second adjacency matrix, the spatial characteristics of the historical speeds and the characteristics of the speed-related regions at the future moment;

[0201] Among them, the recurrent neural network unit includes a third Chebyshev graph convolutional network; the third Chebyshev graph convolutional network is used to fuse the second adjacency matrix and the preliminary fusion features; the preliminary fusion features are obtained by fusing the historical vehicle speed spatial features and the vehicle speed associated area features at future moments.

[0202] In one embodiment, the Fig.10 The vehicle parameter prediction device further comprises:

[0203] A first determination module is used to determine the SOC value of the target vehicle at each future moment according to the future speed of the target vehicle at each future moment;

[0204] The second determination module is used to determine the optimal control parameters of the range extender of the target vehicle at each future moment according to the SOC value of the target vehicle at each future moment based on the state variable update function of the SOC value and the output power of the range extender and the cost function of the SOC value and the output power of the range extender.

[0205] Each module in the above-mentioned vehicle parameter prediction device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0206] In one embodiment, a computer device is provided. The computer device may be a platform side, and its internal structure diagram may be as shown in FIG. Fig.11 As shown. The computer device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store vehicle parameter information. The network interface of the computer device is used to communicate with an external user side through a network connection. When the computer program is executed by the processor, a vehicle parameter prediction method is implemented.

[0207] Those skilled in the art will understand that Figure 1 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. Specifically, the computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0208] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0209] Extract the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle;

[0210] Extract features of the future driving area information of the target vehicle at the future time, and obtain the speed-related area features of the target vehicle at the future time;

[0211] The future speed of the target vehicle at a future time is predicted based on the historical speed spatial characteristics and the speed-related regional characteristics at a future time.

[0212] In one embodiment, when the processor executes the computer program, the following steps are further implemented: the historical vehicle speed information includes vehicle speed information at each historical moment in the historical period;

[0213] Accordingly, feature extraction is performed on the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle, including:

[0214] Process the vehicle speed information at each historical moment to obtain the vehicle speed similarity corresponding to each historical moment;

[0215] Based on the first Chebyshev graph convolutional network, the vehicle speed information at each historical moment is feature extracted according to the vehicle speed similarity corresponding to each historical moment, and the historical vehicle speed spatial features of the target vehicle are obtained.

[0216] In one embodiment, when the processor executes the computer program, the following steps are further implemented: processing the vehicle speed information at each historical moment to obtain the vehicle speed similarity corresponding to each historical moment, including:

[0217] Determine the cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment;

[0218] The cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment is used as the vehicle speed similarity corresponding to each historical moment.

[0219] In one embodiment, when the processor executes the computer program, the following steps are further implemented: based on the first Chebyshev graph convolutional network, according to the vehicle speed similarity corresponding to each historical moment, feature extraction is performed on the vehicle speed information at each historical moment to obtain the historical vehicle speed spatial feature of the target vehicle, including:

[0220] According to the similarity of vehicle speeds corresponding to each historical moment, a first adjacency matrix is ​​constructed;

[0221] Based on the first Chebyshev graph convolutional network and the first adjacency matrix, the vehicle speed information at each historical moment is extracted to obtain the historical vehicle speed spatial features of the target vehicle.

[0222] In one embodiment, when the processor executes the computer program, the following steps are further implemented: the first Chebyshev graph convolution network includes at least two Chebyshev graph convolution layers connected end to end;

[0223] Accordingly, based on the Chebyshev graph convolutional network, the vehicle speed information at each historical moment is feature extracted according to the first adjacency matrix to obtain the historical vehicle speed spatial features of the target vehicle, including:

[0224] The vehicle speed information at each historical moment is input into the first Chebyshev graph convolution layer, and the first adjacency matrix is ​​input into each Chebyshev graph convolution layer respectively, so as to obtain the vehicle speed spatial features output by each Chebyshev graph convolution layer based on the input data;

[0225] The speed spatial features output by the final Chebyshev graph convolutional layer are used as the historical speed spatial features of the target vehicle;

[0226] Among them, the input data of the first Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed information at each historical moment; the input data of each remaining Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed spatial features output by the previous Chebyshev graph convolution layer of the Chebyshev graph convolution layer.

[0227] In one embodiment, when the processor executes the computer program, the following steps are further implemented: feature extraction is performed on the future driving area information of the target vehicle at the future time to obtain the speed-related area features of the target vehicle at the future time, including:

[0228] Based on the feature extraction network, feature extraction is performed on the future driving area information of the target vehicle at the future time, and a regional feature matrix of the future driving area information is obtained;

[0229] Determine a second adjacency matrix according to the regional feature matrix and a transposed matrix corresponding to the regional feature matrix;

[0230] The vehicle speed associated regional features of the target vehicle at a future time are determined according to the second adjacency matrix and the regional feature matrix.

[0231] In one embodiment, when the processor executes the computer program, the following steps are further implemented: determining a second adjacency matrix according to the regional feature matrix and the transposed matrix corresponding to the regional feature matrix, including:

[0232] The regional feature matrix and the transposed matrix corresponding to the regional feature matrix are multiplied to obtain an initial adjacency matrix;

[0233] The initial adjacency matrix is ​​preprocessed to obtain a second adjacency matrix; wherein the preprocessing includes normalization processing and sorting selection processing.

[0234] In one embodiment, when the processor executes the computer program, the following steps are further implemented: determining the speed-related regional features of the target vehicle at a future time according to the second adjacency matrix and the regional feature matrix, including:

[0235] The second adjacency matrix and the regional feature matrix are fused to obtain regional fusion features;

[0236] Based on the second Chebyshev graph convolutional network, feature extraction is performed on the regional fusion features to obtain the speed-related regional features of the target vehicle at the future moment.

[0237] In one embodiment, when the processor executes the computer program, the following steps are further implemented: predicting the future speed of the target vehicle at a future time according to the historical speed spatial characteristics and the speed associated area characteristics at a future time, including:

[0238] Based on the recurrent neural network unit, the future speed of the target vehicle at the future moment is predicted according to the second adjacency matrix, the spatial characteristics of the historical speeds and the characteristics of the speed-related regions at the future moment;

[0239] Among them, the recurrent neural network unit includes a third Chebyshev graph convolutional network; the third Chebyshev graph convolutional network is used to fuse the second adjacency matrix and the preliminary fusion features; the preliminary fusion features are obtained by fusing the historical vehicle speed spatial features and the vehicle speed associated area features at future moments.

[0240] In one embodiment, when the processor executes the computer program, the processor further implements the following steps:

[0241] Determine the SOC value of the target vehicle at each future time according to the future speed of the target vehicle at each future time;

[0242] Based on the state variable update function of the SOC value and the output power of the range extender, and the cost function of the SOC value and the output power of the range extender, the optimal control parameters of the range extender of the target vehicle at each future moment are determined according to the SOC value of the target vehicle at each future moment.

[0243] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0244] Extract the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle;

[0245] Extract features of the future driving area information of the target vehicle at the future time, and obtain the speed-related area features of the target vehicle at the future time;

[0246] The future speed of the target vehicle at a future time is predicted based on the historical speed spatial characteristics and the speed-related regional characteristics at a future time.

[0247] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: the historical vehicle speed information includes vehicle speed information at each historical moment in the historical period;

[0248] Accordingly, feature extraction is performed on the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle, including:

[0249] Process the vehicle speed information at each historical moment to obtain the vehicle speed similarity corresponding to each historical moment;

[0250] Based on the first Chebyshev graph convolutional network, the vehicle speed information at each historical moment is feature extracted according to the vehicle speed similarity corresponding to each historical moment, and the historical vehicle speed spatial features of the target vehicle are obtained.

[0251] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: processing the vehicle speed information at each historical moment to obtain the vehicle speed similarity corresponding to each historical moment, including:

[0252] Determine the cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment;

[0253] The cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment is used as the vehicle speed similarity corresponding to each historical moment.

[0254] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: based on the first Chebyshev graph convolutional network, according to the vehicle speed similarity corresponding to each historical moment, feature extraction is performed on the vehicle speed information at each historical moment to obtain the historical vehicle speed spatial feature of the target vehicle, including:

[0255] According to the similarity of vehicle speeds corresponding to each historical moment, a first adjacency matrix is ​​constructed;

[0256] Based on the first Chebyshev graph convolutional network and the first adjacency matrix, the vehicle speed information at each historical moment is extracted to obtain the historical vehicle speed spatial features of the target vehicle.

[0257] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented: the first Chebyshev graph convolutional network includes at least two Chebyshev graph convolutional layers connected end to end;

[0258] Accordingly, based on the Chebyshev graph convolutional network, the vehicle speed information at each historical moment is feature extracted according to the first adjacency matrix to obtain the historical vehicle speed spatial features of the target vehicle, including:

[0259] The vehicle speed information at each historical moment is input into the first Chebyshev graph convolution layer, and the first adjacency matrix is ​​input into each Chebyshev graph convolution layer respectively, so as to obtain the vehicle speed spatial features output by each Chebyshev graph convolution layer based on the input data;

[0260] The speed spatial features output by the final Chebyshev graph convolutional layer are used as the historical speed spatial features of the target vehicle;

[0261] Among them, the input data of the first Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed information at each historical moment; the input data of each remaining Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed spatial features output by the previous Chebyshev graph convolution layer of the Chebyshev graph convolution layer.

[0262] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: feature extraction is performed on the future driving area information of the target vehicle at the future time to obtain the speed-related area features of the target vehicle at the future time, including:

[0263] Based on the feature extraction network, feature extraction is performed on the future driving area information of the target vehicle at the future time, and a regional feature matrix of the future driving area information is obtained;

[0264] Determine a second adjacency matrix according to the regional feature matrix and a transposed matrix corresponding to the regional feature matrix;

[0265] The vehicle speed associated regional features of the target vehicle at a future time are determined according to the second adjacency matrix and the regional feature matrix.

[0266] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: determining a second adjacency matrix according to the regional feature matrix and the transposed matrix corresponding to the regional feature matrix, including:

[0267] The regional feature matrix and the transposed matrix corresponding to the regional feature matrix are multiplied to obtain an initial adjacency matrix;

[0268] The initial adjacency matrix is ​​preprocessed to obtain a second adjacency matrix; wherein the preprocessing includes normalization processing and sorting selection processing.

[0269] In one embodiment, determining the speed-related regional features of the target vehicle at a future time according to the second adjacency matrix and the regional feature matrix includes:

[0270] The second adjacency matrix and the regional feature matrix are fused to obtain regional fusion features;

[0271] Based on the second Chebyshev graph convolutional network, feature extraction is performed on the regional fusion features to obtain the speed-related regional features of the target vehicle at the future moment.

[0272] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: predicting the future speed of the target vehicle at a future time according to the historical speed spatial characteristics and the speed associated area characteristics at a future time, including:

[0273] Based on the recurrent neural network unit, the future speed of the target vehicle at the future moment is predicted according to the second adjacency matrix, the spatial characteristics of the historical speeds and the characteristics of the speed-related regions at the future moment;

[0274] Among them, the recurrent neural network unit includes a third Chebyshev graph convolutional network; the third Chebyshev graph convolutional network is used to fuse the second adjacency matrix and the preliminary fusion features; the preliminary fusion features are obtained by fusing the historical vehicle speed spatial features and the vehicle speed associated area features at future moments.

[0275] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0276] Determine the SOC value of the target vehicle at each future time according to the future speed of the target vehicle at each future time;

[0277] Based on the state variable update function of the SOC value and the output power of the range extender, and the cost function of the SOC value and the output power of the range extender, the optimal control parameters of the range extender of the target vehicle at each future moment are determined according to the SOC value of the target vehicle at each future moment.

[0278] In one embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the following steps:

[0279] Extract the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle;

[0280] Extract features of the future driving area information of the target vehicle at the future time, and obtain the speed-related area features of the target vehicle at the future time;

[0281] The future speed of the target vehicle at a future time is predicted based on the historical speed spatial characteristics and the speed-related regional characteristics at a future time.

[0282] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: the historical vehicle speed information includes vehicle speed information at each historical moment in the historical period;

[0283] Accordingly, feature extraction is performed on the historical speed information of the target vehicle to obtain the historical speed spatial features of the target vehicle, including:

[0284] Process the vehicle speed information at each historical moment to obtain the vehicle speed similarity corresponding to each historical moment;

[0285] Based on the first Chebyshev graph convolutional network, the vehicle speed information at each historical moment is feature extracted according to the vehicle speed similarity corresponding to each historical moment, and the historical vehicle speed spatial features of the target vehicle are obtained.

[0286] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: processing the vehicle speed information at each historical moment to obtain the vehicle speed similarity corresponding to each historical moment, including:

[0287] Determine the cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment;

[0288] The cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment is used as the vehicle speed similarity corresponding to each historical moment.

[0289] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: based on the first Chebyshev graph convolutional network, according to the vehicle speed similarity corresponding to each historical moment, feature extraction is performed on the vehicle speed information at each historical moment to obtain the historical vehicle speed spatial feature of the target vehicle, including:

[0290] According to the similarity of vehicle speeds corresponding to each historical moment, a first adjacency matrix is ​​constructed;

[0291] Based on the first Chebyshev graph convolutional network and the first adjacency matrix, the vehicle speed information at each historical moment is extracted to obtain the historical vehicle speed spatial features of the target vehicle.

[0292] In one embodiment, when the computer program is executed by a processor, the following steps are further implemented: the first Chebyshev graph convolutional network includes at least two Chebyshev graph convolutional layers connected end to end;

[0293] Accordingly, based on the Chebyshev graph convolutional network, the vehicle speed information at each historical moment is feature extracted according to the first adjacency matrix to obtain the historical vehicle speed spatial features of the target vehicle, including:

[0294] The vehicle speed information at each historical moment is input into the first Chebyshev graph convolution layer, and the first adjacency matrix is ​​input into each Chebyshev graph convolution layer respectively, so as to obtain the vehicle speed spatial features output by each Chebyshev graph convolution layer based on the input data;

[0295] The speed spatial features output by the final Chebyshev graph convolutional layer are used as the historical speed spatial features of the target vehicle;

[0296] Among them, the input data of the first Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed information at each historical moment; the input data of each remaining Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed spatial features output by the previous Chebyshev graph convolution layer of the Chebyshev graph convolution layer.

[0297] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: feature extraction is performed on the future driving area information of the target vehicle at the future time to obtain the speed-related area features of the target vehicle at the future time, including:

[0298] Based on the feature extraction network, feature extraction is performed on the future driving area information of the target vehicle at the future time, and a regional feature matrix of the future driving area information is obtained;

[0299] Determine a second adjacency matrix according to the regional feature matrix and a transposed matrix corresponding to the regional feature matrix;

[0300] The vehicle speed associated regional features of the target vehicle at a future time are determined according to the second adjacency matrix and the regional feature matrix.

[0301] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: determining a second adjacency matrix according to the regional feature matrix and the transposed matrix corresponding to the regional feature matrix, including:

[0302] The regional feature matrix and the transposed matrix corresponding to the regional feature matrix are multiplied to obtain an initial adjacency matrix;

[0303] The initial adjacency matrix is ​​preprocessed to obtain a second adjacency matrix; wherein the preprocessing includes normalization processing and sorting selection processing.

[0304] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: determining the speed-related regional features of the target vehicle at a future time according to the second adjacency matrix and the regional feature matrix, including:

[0305] The second adjacency matrix and the regional feature matrix are fused to obtain regional fusion features;

[0306] Based on the second Chebyshev graph convolutional network, feature extraction is performed on the regional fusion features to obtain the speed-related regional features of the target vehicle at the future moment.

[0307] In one embodiment, when the computer program is executed by the processor, the following steps are further implemented: predicting the future speed of the target vehicle at a future time according to the historical speed spatial characteristics and the speed associated area characteristics at a future time, including:

[0308] Based on the recurrent neural network unit, the future speed of the target vehicle at the future moment is predicted according to the second adjacency matrix, the spatial characteristics of the historical speeds and the characteristics of the speed-related regions at the future moment;

[0309] Among them, the recurrent neural network unit includes a third Chebyshev graph convolutional network; the third Chebyshev graph convolutional network is used to fuse the second adjacency matrix and the preliminary fusion features; the preliminary fusion features are obtained by fusing the historical vehicle speed spatial features and the vehicle speed associated area features at future moments.

[0310] In one embodiment, when the computer program is executed by a processor, the following steps are also implemented:

[0311] Determine the SOC value of the target vehicle at each future time according to the future speed of the target vehicle at each future time;

[0312] Based on the state variable update function of the SOC value and the output power of the range extender, and the cost function of the SOC value and the output power of the range extender, the optimal control parameters of the range extender of the target vehicle at each future moment are determined according to the SOC value of the target vehicle at each future moment.

[0313] It should be noted that the data involved in this application (including but not limited to messages and information in the vehicle parameter prediction process) are all data that have been fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.

[0314] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0315] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0316] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A vehicle parameter prediction method, characterized in that: The method comprises: The vehicle speed information at each historical moment is processed to obtain the vehicle speed similarity corresponding to each historical moment; based on the first Chebyshev graph convolutional network, according to the vehicle speed similarity corresponding to each historical moment, the vehicle speed information at each historical moment is feature extracted to obtain the historical vehicle speed spatial feature of the target vehicle; Based on the feature extraction network, feature extraction is performed on the future driving area information of the target vehicle at the future time to obtain a regional feature matrix of the future driving area information; a second adjacency matrix is ​​determined according to the regional feature matrix and a transposed matrix corresponding to the regional feature matrix; and a vehicle speed-related regional feature of the target vehicle at the future time is determined according to the second adjacency matrix and the regional feature matrix; Based on the recurrent neural network unit, the future vehicle speed of the target vehicle at the future moment is predicted according to the second adjacency matrix, the historical vehicle speed spatial characteristics and the vehicle speed associated area characteristics at the future moment.

2. The method according to claim 1, characterized in that The processing of the vehicle speed information at each historical moment to obtain the vehicle speed similarity corresponding to each historical moment includes: Determine the cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment; The cosine similarity between the vehicle speed information at the reference moment and the vehicle speed information at each historical moment is used as the vehicle speed similarity corresponding to each historical moment.

3. The method according to claim 2, characterized in that The method further comprises: Select a reference moment from various historical moments.

4. The method according to claim 1, characterized in that: Based on the first Chebyshev graph convolutional network, according to the speed similarity corresponding to each historical moment, feature extraction is performed on the speed information of each historical moment to obtain the historical speed spatial feature of the target vehicle, including: According to the similarity of vehicle speeds corresponding to each historical moment, a first adjacency matrix is ​​constructed; Based on the first Chebyshev graph convolutional network and according to the first adjacency matrix, feature extraction is performed on the vehicle speed information at each historical moment to obtain the historical vehicle speed spatial feature of the target vehicle.

5. The method according to claim 4, characterized in that The first Chebyshev graph convolutional network includes at least two Chebyshev graph convolutional layers connected end to end; Based on the first Chebyshev graph convolutional network and according to the first adjacency matrix, feature extraction is performed on the vehicle speed information at each historical moment to obtain the historical vehicle speed spatial feature of the target vehicle, including: Inputting the vehicle speed information at each historical moment into the first Chebyshev graph convolution layer, and inputting the first adjacency matrix into each Chebyshev graph convolution layer respectively, to obtain the vehicle speed spatial features output by each Chebyshev graph convolution layer based on the input data; The vehicle speed spatial feature output by the final Chebyshev graph convolutional layer is used as the historical vehicle speed spatial feature of the target vehicle; Among them, the input data of the first Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed information at each historical moment; the input data of each remaining Chebyshev graph convolution layer includes the first adjacency matrix and the vehicle speed spatial features output by the previous Chebyshev graph convolution layer of the Chebyshev graph convolution layer.

6. The method according to claim 1, characterized in that The determining of a second adjacency matrix according to the regional feature matrix and a transposed matrix corresponding to the regional feature matrix comprises: Multiplying the regional feature matrix and the transposed matrix corresponding to the regional feature matrix to obtain an initial adjacency matrix; The initial adjacency matrix is ​​preprocessed to obtain the second adjacency matrix; wherein the preprocessing includes normalization processing and sorting selection processing.

7. The method according to claim 1, characterized in that The determining, based on the second adjacency matrix and the regional feature matrix, the vehicle speed associated regional feature of the target vehicle at the future moment includes: Fusing the second adjacency matrix and the regional feature matrix to obtain regional fusion features; Based on the second Chebyshev graph convolutional network, feature extraction is performed on the regional fusion features to obtain the speed-related regional features of the target vehicle at the future moment.

8. The method according to claim 1, characterized in that The recurrent neural network unit includes a third Chebyshev graph convolutional network; the third Chebyshev graph convolutional network is used to fuse the second adjacency matrix and the preliminary fusion features; the preliminary fusion features are obtained by fusing the historical vehicle speed spatial features and the vehicle speed associated area features at the future moment.

9. The method according to claim 1, characterized in that: The method further comprises: Determining the remaining power SOC value of the target vehicle at each future moment according to the future vehicle speed of the target vehicle at each future moment; Based on the state variable update function of the SOC value and the output power of the range extender, and the cost function of the SOC value and the output power of the range extender, the optimal control parameters of the range extender of the target vehicle at each future moment are determined according to the SOC value of the target vehicle at each future moment.

10. A vehicle parameter prediction device, characterized in that: The device comprises: The first extraction module is used to process the vehicle speed information at each historical moment to obtain the vehicle speed similarity corresponding to each historical moment; based on the first Chebyshev graph convolutional network, according to the vehicle speed similarity corresponding to each historical moment, feature extraction is performed on the vehicle speed information at each historical moment to obtain the historical vehicle speed spatial feature of the target vehicle; A second extraction module is used to extract features of the future driving area information of the target vehicle at the future time based on the feature extraction network to obtain a regional feature matrix of the future driving area information; determine a second adjacency matrix according to the regional feature matrix and a transposed matrix corresponding to the regional feature matrix; determine the vehicle speed-related regional features of the target vehicle at the future time according to the second adjacency matrix and the regional feature matrix; A prediction module is used to predict the future speed of the target vehicle at the future moment based on a recurrent neural network unit, according to the second adjacency matrix, the historical vehicle speed spatial characteristics and the vehicle speed associated area characteristics at the future moment.

Images