Vehicle trajectory prediction method and system, computer equipment and medium

Abstract

The invention provides a vehicle track prediction method and system, computer equipment and a medium, and the method comprises the steps: obtaining a track data set, extracting vehicle lane changing data points from the track data set, carrying out the prediction of a driving behavior intention according to the vehicle lane changing data points and a vehicle track data coding result, and obtaining a driving behavior intention prediction result; and correcting the vehicle trajectory data coding result, and performing vehicle trajectory prediction based on the corrected vehicle trajectory data coding result and the driving behavior intention prediction result. According to the method, the vehicle trajectory prediction is performed based on the vehicle trajectory data coding result and the driving behavior intention prediction result, the historical vehicle trajectory data can be combined, and the driving behavior intention of the driver is fully considered, so that the precision of vehicle trajectory prediction is improved, and the method can adapt to more complex traffic scenes. The method can respond to the change of the traffic environment in real time, can adjust the prediction result in time, and guarantees the safety and smoothness of vehicle driving.

Description

Technical Field

[0001] The present application relates to the field of intelligent driving technology, and particularly to a vehicle trajectory prediction method and system, a computer device, and a medium. Background Art

[0002] In the context of the rapid development of vehicle intelligent driving and autonomous driving technologies, accurately predicting the driving trajectory of a vehicle is of great significance for improving road safety, optimizing traffic flow, and enhancing the driving experience. However, when predicting the vehicle trajectory, related technologies often ignore the driver's intention and the influence of surrounding vehicles, resulting in limited prediction accuracy and being unable to meet the requirements in complex traffic scenarios. At the same time, when predicting the vehicle trajectory, related technologies may not consider the sufficiency of historical trajectory data, nor the impact of driving behavior on vehicle trajectory prediction. Summary of the Invention

[0003] In view of the above-mentioned disadvantages of the prior art, the purpose of the present application is to provide a vehicle trajectory prediction method and system, a computer device, and a medium, which are used to solve the technical problems existing in the prior art.

[0004] To achieve the above object and other related objects, the present application provides a vehicle trajectory prediction method, including the following steps:

[0005] Obtain a trajectory data set, where the trajectory data set is obtained based on vehicle trajectory data, and the vehicle trajectory data includes vehicle positions at multiple moments;

[0006] Extract vehicle lane change data points from the trajectory data set, where the vehicle lane change data points include vehicle lane change starting points and / or vehicle lane change ending points;

[0007] Predict the driving behavior intention according to the vehicle lane change data points and the encoding result of the vehicle trajectory data to obtain a driving behavior intention prediction result, where the encoding result of the vehicle trajectory data is obtained by encoding the vehicle trajectory data;

[0008] Correct the encoding result of the vehicle trajectory data, and perform vehicle trajectory prediction based on the corrected encoding result of the vehicle trajectory data and the driving behavior intention prediction result.

[0009] In an embodiment of the present application, the process of predicting the driving behavior intention according to the vehicle lane change data points and the encoding result of the vehicle trajectory data to obtain a driving behavior intention prediction result includes:

[0010] Input the vehicle lane change data points and the encoded result of the vehicle trajectory data into a fully connected layer composed of multiple neurons, and add an activation function in the fully connected layer to perform a non-linear mapping on the vehicle lane change data points and the encoded result of the vehicle trajectory data, so as to obtain the driving behavior intention prediction probability;

[0011] Perform normalization processing on the driving behavior intention prediction probability, and determine the corresponding driving behavior intention prediction result according to the normalization processing result of the driving behavior intention prediction probability; wherein, the driving behavior intention prediction result includes a left lane change from the current lane to the left lane, a right lane change from the current lane to the right lane, and / or lane keeping without changing the current lane.

[0012] In an embodiment of the present application, the process of correcting the encoded result of the vehicle trajectory data includes:

[0013] Obtain a matrix determined in advance or in real time, and calculate a correction coefficient based on the matrix, some or all of the encoded results in the encoded result of the vehicle trajectory data, and the vehicle position corresponding to the last moment in the vehicle trajectory data;

[0014] Perform normalization processing on the correction coefficient, and correct some or all of the encoded results in the encoded result of the vehicle trajectory data through the normalization processing result of the correction coefficient to obtain the corrected encoded result of the vehicle trajectory data.

[0015] In an embodiment of the present application, the process of predicting the vehicle trajectory based on the corrected encoded result of the vehicle trajectory data and the driving behavior intention prediction result includes:

[0016] Obtain the decoded result of the vehicle trajectory data corresponding to the encoded result of the vehicle trajectory data, and select the decoded result at the target moment from the decoded result of the vehicle trajectory data; wherein, the target moment is one of the moments in the set of corresponding moments of the vehicle trajectory data;

[0017] Use the decoded result at the target moment as the initial state, and output the decoded result at the first moment based on the corrected encoded result of the vehicle trajectory data and the driving behavior intention prediction result, and predict the vehicle position at the second moment through the decoded result at the first moment; wherein, the first moment is after the target moment and differs from the target moment by one moment, and the second moment is after the first moment and differs from the first moment by one moment;

[0018] Continue iterative prediction according to the predicted vehicle position at the second moment, and obtain the vehicle trajectory prediction result according to the vehicle positions at multiple predicted moments.

[0019] In one embodiment of the present application, the process of obtaining the vehicle trajectory data coding result by coding the vehicle trajectory data includes:

[0020] Screen the vehicle position at the target moment from the vehicle trajectory data, where the target moment is one of the moments in the moment set corresponding to the vehicle trajectory data;

[0021] Encode the vehicle position at the target moment by using a pre-determined or real-time target neural network to obtain the coding result at the target moment; and,

[0022] Take the coding result at the target moment and the vehicle position at the next moment after the target moment as the input of the target neural network, encode the vehicle position at the next moment through the target neural network, output the coding result at the next moment, and continue to input the coding result at the next moment into the target neural network for encoding until the vehicle trajectory data coding result corresponding to the vehicle trajectory data is obtained; where, if the target moment is the last moment in the moment set corresponding to the vehicle trajectory data, the next moment after the target moment is the current moment.

[0023] In one embodiment of the present application, the process of extracting vehicle lane change data points from the trajectory data set includes:

[0024] Extract the vehicle position at the first moment and the vehicle position at the second moment from the trajectory data set, where the first moment and the second moment are separated by a preset number of moments;

[0025] Based on the vehicle position at the first moment and the vehicle position at the second moment, calculate the position deviation of the vehicle in a continuous preset number of moments; and when the position deviation is less than the preset deviation threshold, take the vehicle position at the first moment or the vehicle position at the second moment as the vehicle lane change data point.

[0026] In one embodiment of the present application, the process of obtaining the trajectory data set based on the vehicle trajectory data includes:

[0027] Obtain the vehicle trajectory data obtained in multiple traffic scenarios in advance, denoted as the original vehicle trajectory data;

[0028] Preprocess the original vehicle trajectory data, and obtain the trajectory data set based on the preprocessed vehicle trajectory data; where the preprocessing includes denoising, interpolation and / or coordinate transformation.

[0029] The present application also provides a vehicle trajectory prediction system, and the system includes:

[0030] A data acquisition module for obtaining a trajectory dataset, which is obtained based on vehicle trajectory data, and the vehicle trajectory data includes vehicle positions at multiple moments;

[0031] A lane-changing data point module for extracting vehicle lane-changing data points from the trajectory dataset, and the vehicle lane-changing data points include vehicle lane-changing starting points and / or vehicle lane-changing ending points;

[0032] A driving behavior intention prediction module for predicting a driving behavior intention according to the vehicle lane-changing data points and the encoding result of the vehicle trajectory data to obtain a driving behavior intention prediction result, and the encoding result of the vehicle trajectory data is obtained by encoding the vehicle trajectory data;

[0033] A vehicle trajectory prediction module for correcting the encoding result of the vehicle trajectory data and performing vehicle trajectory prediction based on the corrected encoding result of the vehicle trajectory data and the driving behavior intention prediction result.

[0034] This application also provides a computer device, including a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to implement the steps of the vehicle trajectory prediction method described in any one of the above.

[0035] This application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the vehicle trajectory prediction method described in any one of the above are implemented.

[0036] As described above, this application provides a vehicle trajectory prediction method and system, a computer device and a medium, which have the following beneficial effects: by extracting vehicle lane-changing data points from the trajectory dataset, and then predicting the driving behavior intention according to the vehicle lane-changing data points and the encoding result of the vehicle trajectory data to obtain a driving behavior intention prediction result; then correcting the encoding result of the vehicle trajectory data, and performing vehicle trajectory prediction based on the corrected encoding result of the vehicle trajectory data and the driving behavior intention prediction result; wherein, the trajectory dataset is obtained based on vehicle trajectory data, the vehicle trajectory data includes vehicle positions at multiple moments, the vehicle lane-changing data points include vehicle lane-changing starting points and / or vehicle lane-changing ending points, and the encoding result of the vehicle trajectory data is obtained by encoding the vehicle trajectory data. It can be seen that this application performs vehicle trajectory prediction based on the encoding result of the vehicle trajectory data and the driving behavior intention prediction result, which can not only combine the vehicle historical trajectory data, but also fully consider the driving behavior intention of the driver, thereby improving the accuracy of vehicle trajectory prediction and being able to adapt to more complex traffic scenarios. Description of the Drawings

[0037] Figure 1Schematic flowchart of a vehicle trajectory prediction method provided by an embodiment of the present application;

[0038] Figure 2 Schematic flowchart of a vehicle trajectory prediction method provided by another embodiment of the present application;

[0039] Figure 3 Schematic diagram of the driver intention recognition model structure provided by an embodiment of the present application;

[0040] Figure 4 Schematic diagram of the vehicle trajectory prediction model structure provided by an embodiment of the present application;

[0041] Figure 5 Schematic diagram of the hardware structure of a vehicle trajectory prediction system provided by an embodiment of the present application;

[0042] Figure 6 Schematic diagram of the hardware structure of a computer device suitable for implementing one or more embodiments of the present application. Detailed implementation manners

[0043] The following uses specific specific examples to illustrate the implementation manners of the present application. Those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in this specification. The present application can also be implemented or applied through other different specific implementation manners. Various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present application. It can be understood that, without conflict, the following embodiments and the features in the embodiments can be combined with each other. In addition, it can be understood that the drawings provided in the following embodiments only illustrate the basic concept of the present application schematically, so only the components related to the present application are shown in the drawings, rather than being drawn according to the number, shape and size of the components in actual implementation. The type, quantity and ratio of each component in actual implementation can be an arbitrary change, and the component layout type may also be more complex.

[0044] Figure 1 Shows a schematic flowchart of a vehicle trajectory prediction method. Specifically, in an exemplary embodiment, as Figure 1 shown, this embodiment provides a vehicle trajectory prediction method, and the method includes the following steps:

[0045] S110, obtain a trajectory data set, the trajectory data set is obtained based on vehicle trajectory data, and the vehicle trajectory data includes vehicle positions at multiple moments;

[0046] S120, extract vehicle lane change data points from the trajectory data set, and the vehicle lane change data points include vehicle lane change start points and / or vehicle lane change end points;

[0047] S130. Predict the driving behavior intention based on the vehicle lane change data points and the encoding result of the vehicle trajectory data to obtain the driving behavior intention prediction result, where the encoding result of the vehicle trajectory data is obtained by encoding the vehicle trajectory data;

[0048] S140. Correct the encoding result of the vehicle trajectory data, and perform vehicle trajectory prediction based on the corrected encoding result of the vehicle trajectory data and the driving behavior intention prediction result.

[0049] In some exemplary embodiments, the trajectory data set can be directly a trajectory data set generated by other persons or institutions, or a data set generated based on real-time collected vehicle trajectory data, or a data set generated based on vehicle trajectory data collected at historical moments, and no specific limitation is made here. Among them, the vehicle trajectory data can be collected in one or more traffic scenarios. For example, the vehicle trajectory data can be mainly composed of data such as multiple moments and the vehicle positions corresponding to each moment collected in a highway scenario by using a high-resolution traffic camera and a radar sensor; another example is that the vehicle trajectory data can be mainly composed of data such as multiple moments and the vehicle positions corresponding to each moment collected in an urban road scenario by using a high-resolution traffic camera and a radar sensor. The specific pixel of the high-resolution traffic camera can be selected or set according to the actual situation, and no specific numerical limit is made on the camera here. For example, a camera with a pixel greater than or equal to 2 million pixels can be used as the high-resolution traffic camera.

[0050] In some exemplary embodiments, the process of obtaining a trajectory data set based on vehicle trajectory data includes: acquiring vehicle trajectory data obtained in advance in multiple traffic scenarios, denoted as original vehicle trajectory data; preprocessing the original vehicle trajectory data, and obtaining a trajectory data set based on the preprocessed vehicle trajectory data; wherein, the preprocessing includes but is not limited to denoising, interpolation, and / or coordinate transformation. As an example, the vehicle trajectory data collected in a highway scenario and / or an urban road scenario can be denoted as the original vehicle trajectory data, and then the original vehicle trajectory data is preprocessed such as denoising, interpolation, and / or coordinate transformation. At the same time, a trajectory data set including information such as vehicle speed, acceleration, position, and timestamp is obtained based on the preprocessed vehicle trajectory data. Among them, the preprocessing process of denoising, interpolation, and / or coordinate transformation of the original vehicle trajectory data can refer to the related art and will not be elaborated here. For example, when the sampling frequency of the vehicle trajectory data is 10 Hz, in order to improve the continuity and authenticity of the vehicle trajectory data, resampling can be performed at a frequency of 5 Hz. Another example is that in order to reduce noise and abnormal vehicle trajectory data, the SEMA (Symmetric Exponential Moving Average) method can be used to smooth the vehicle trajectory data collected in time series, so as to perform denoising. Specifically, the smoothing process can be: wherein, is the estimated density at the x sampling point, n is the number of sampling points, i ∈ x, h is the bandwidth parameter, is the kernel function.

[0051] In some exemplary embodiments, the process of extracting vehicle lane change data points from the trajectory data set includes: extracting the vehicle position at the first moment and the vehicle position at the second moment from the trajectory data set, with the first moment and the second moment differing by a preset number of moments; calculating the position deviation of the vehicle under consecutive preset number of moments based on the vehicle position at the first moment and the vehicle position at the second moment; and when the position deviation is less than a preset deviation threshold, taking the vehicle position at the first moment or the vehicle position at the second moment as the vehicle lane change data point. Among them, the preset number of moments can be specifically selected or set according to the actual situation, and no specific numerical limit is provided here. For example, it can be recorded that the first moment and the second moment differ by 3 moments as the first moment and the second moment differ by a preset number of moments. At the same time, the preset deviation threshold can also be specifically selected or set according to the actual situation, and no specific numerical limit is provided here. As an example, for example, O k-m:k represents the trajectory data set from the k - m moment to the k moment, then there is O k-m:k =[P k-m ,P k-m+1 ,…,P k, where k and m are positive integers, k > m, and k - m > 0; P k represents the vehicle position (x k , y k ) at the k-th moment, where x k represents the lateral position of the vehicle at the k-th moment, and y k represents the longitudinal position of the vehicle at the k-th moment; similarly, P k-m represents the vehicle position (x k-m , y k-m ) at the (k - m)-th moment, and P k-m+1 represents the vehicle position (x k-m+1 , y k-m+1 ) at the (k - m + 1)-th moment. Extract the vehicle position (x k-m:k , y k ) at the k-th moment and the vehicle position (x k , y k-3 ) at the (k - 3)-th moment from the trajectory data set O k-3 ), and then based on the vehicle position (x k , y k ) at the k-th moment and the vehicle position (x k-3 , y k-3 ) at the (k - 3)-th moment, calculate the position deviation ψ of the vehicle at three consecutive preset moments, and there is: When the position deviation ψ is less than the preset deviation threshold ψ s , take the vehicle position (x k , y k ) at the k-th moment as the starting point or the ending point of the vehicle lane change. Among them, if the k-th moment represents the current moment, then the (k - 3)-th moment represents a historical moment that is 3 moments different from the current moment and before the current moment.

[0052] In some exemplary embodiments, the process of predicting the driving behavior intention based on the vehicle lane change data points and the encoding result of the vehicle trajectory data includes: inputting the vehicle lane change data points and the encoding result of the vehicle trajectory data into a fully connected layer composed of multiple neurons, and adding an activation function in the fully connected layer to perform a non-linear mapping on the vehicle lane change data points and the encoding result of the vehicle trajectory data to obtain the prediction probability of the driving behavior intention; performing a normalization process on the prediction probability of the driving behavior intention, and determining the corresponding prediction result of the driving behavior intention according to the normalization result of the prediction probability of the driving behavior intention; where the prediction result of the driving behavior intention includes a left lane change from the current lane to the left lane, a right lane change from the current lane to the right lane, and / or a lane keeping without changing the current lane. As an example, the fully connected layer can be three layers, each layer having 128 neurons, and the activation function can adopt the ReLU function.

[0053] In some exemplary embodiments, the process of obtaining the encoded result of vehicle trajectory data by encoding the vehicle trajectory data includes: screening out the vehicle position at the target moment from the vehicle trajectory data, where the target moment is one of the moments in the set of moments corresponding to the vehicle trajectory data; using a pre-determined or real-time target neural network to encode the vehicle position at the target moment to obtain the encoded result at the target moment; and using the encoded result at the target moment and the vehicle position at the next moment after the target moment as the input of the target neural network, encoding the vehicle position at the next moment through the target neural network, outputting the encoded result at the next moment, and continuing to input the encoded result at the next moment into the target neural network for encoding until the encoded result of the vehicle trajectory data corresponding to the vehicle trajectory data is obtained; where if the target moment is the last moment in the set of moments corresponding to the vehicle trajectory data, the next moment after the target moment is the current moment. As an example, the pre-determined or real-time target neural network includes, but is not limited to, a Gated Recurrent Unit (GRU). For example, the GRU hidden layer can be used to encode the vehicle position to obtain the corresponding encoded result. Among them, the hidden units in the GRU hidden layer can be selected or set according to the actual situation. For example, it can be set that the GRU hidden layer contains 128 hidden units. Specifically, the vehicle position (x k-1 , y k-1 ) at the (k - 1)-th moment is screened out from the vehicle trajectory data; the GRU is used to encode the vehicle position (x k-1 , y k-1 ) at the (k - 1)-th moment to obtain the encoded result at the (k - 1)-th moment And, the encoded result at the (k - 1)-th moment and the vehicle position (x k , y k ) at the k-th moment are used as the input of the GRU hidden layer, and the vehicle position (x k , y k ) at the k-th moment is encoded through the GRU hidden layer, and the encoded result at the k-th moment There is: In the formula, P k represents the vehicle position (x k , y k ) at the k-th moment, and fGRU() represents the hidden function in the GRU hidden layer. Then, the encoded result at the k-th moment Continue to input it into the GRU hidden layer for encoding. The encoding ends until the vehicle position at each moment in the vehicle trajectory data obtains the corresponding encoding result. Then, taking the encoding results of all moments as a set, the vehicle trajectory data encoding result corresponding to the vehicle trajectory data can be obtained. Among them, if the k-th moment represents the current moment, the (k - 1)-th moment represents the target moment, and the k-th moment is the next moment after the (k - 1)-th moment. For the trajectory data set O from the (k - m)-th moment to the k-th moment k-m:k , its corresponding vehicle trajectory data encoding result is Among them, represents the vehicle trajectory data encoding result from the (k - m)-th moment to the k-th moment, represents the encoding result of the (k - m)-th moment, represents the encoding result of the (k - m + 1)-th moment. Thus, it can be seen that by using GRU to encode the vehicle trajectory data, since there are update gates and reset gates in GRU, GRU can control the flow of information, so that the gradient will not rapidly decrease or increase during the backpropagation process, thereby solving the gradient vanishing and gradient explosion problems of traditional RNN (Recurrent Neural Networks) when processing long sequence data.

[0054] In some exemplary embodiments, the process of correcting the vehicle trajectory data encoding result includes: obtaining a matrix determined in advance or in real time, and calculating a correction coefficient based on the matrix, some or all of the encoding results in the vehicle trajectory data encoding result, and the vehicle position corresponding to the last moment in the vehicle trajectory data; performing normalization processing on the correction coefficient, and correcting some or all of the encoding results in the vehicle trajectory data encoding result through the normalization processing result of the correction coefficient to obtain the corrected vehicle trajectory data encoding result. As an example, specifically, the matrix determined in advance or in real time can be denoted as matrix C, and then calculate the correction coefficient based on matrix C, some or all of the encoding results in the vehicle trajectory data encoding result, and the vehicle position corresponding to the last moment in the vehicle trajectory data, as follows: In the formula, cor represents the correction coefficient, P k represents the vehicle position at the k-th moment, C represents the matrix determined in advance or in real time, and matrix C can be a coefficient, represents the vehicle trajectory data encoding result from the (k - m)-th moment to the k-th moment, represents the transpose of. Perform normalization processing on the correction coefficient, as follows: Among them, represents the normalization processing result of the correction coefficient cor, and softmax() represents the normalization function; through the normalization processing result of the correction coefficient The encoding result of the vehicle trajectory data from the (k - m)-th moment to the k-th moment is corrected to obtain the corrected encoding result cor of the vehicle trajectory data f , and there is:

[0055] In some exemplary embodiments, the process of predicting the vehicle trajectory based on the corrected encoding result of the vehicle trajectory data and the driving behavior intention prediction result includes: obtaining the decoding result of the vehicle trajectory data corresponding to the encoding result of the vehicle trajectory data, and selecting the decoding result at the target moment from the decoding result of the vehicle trajectory data; wherein, the target moment is one of the moments in the set of moments corresponding to the vehicle trajectory data; taking the decoding result at the target moment as the initial state, and outputting the decoding result at the first moment based on the corrected encoding result of the vehicle trajectory data and the driving behavior intention prediction result, and predicting the vehicle position at the second moment through the decoding result at the first moment; wherein, the first moment is after the target moment and differs from the target moment by one moment, and the second moment is after the first moment and differs from the first moment by one moment; if the target moment is the last moment in the set of moments corresponding to the vehicle trajectory data, then the first moment is the current moment, and the second moment is the next moment after the current moment; continue iterative prediction according to the predicted vehicle position at the second moment, and obtain the vehicle trajectory prediction result according to the vehicle positions at the predicted multiple moments. As an example, specifically, if the encoding result of the vehicle trajectory data is output through the GRU hidden layer, the decoding result of the vehicle trajectory data corresponding to the encoding result of the vehicle trajectory data can also be output through the GRU hidden layer, and the specific process will not be elaborated here. If the target moment is the (k - 1)-th moment, then the first moment is the k-th moment, and the second moment is the (k + 1)-th moment. At this time, by obtaining the encoding result corresponding decoding result and taking the decoding result at the (k - 1)-th moment as the initial state, and then outputting the decoding result at the k-th moment based on the corrected encoding result cor of the vehicle trajectory data f and the driving behavior intention prediction result There is: wherein, X k = [P k , cor f ; then predict the vehicle position at the (k + 1)-th moment through the decoding result at the k-th moment Then continue to predict the vehicle position at the (k + 2)-th moment according to the vehicle position at the (k + 1)-th moment and the decoding result at the k-th moment And so on for iterative prediction, and then correlate the vehicle positions at the predicted multiple moments to obtain the vehicle trajectory prediction result.​​

[0056] In an exemplary embodiment, as Figure 2 shown, this embodiment provides a vehicle trajectory prediction method, which includes the following steps:

[0057] Preprocess the real vehicle trajectory data collected in the highway scenario and / or urban road scenario to obtain the corresponding vehicle trajectory dataset. The preprocessing process can refer to some of the above embodiments and will not be elaborated here.

[0058] Encode the vehicle historical trajectory information based on the recurrent neural network of the gated recurrent unit. Among them, the vehicle historical trajectory information can also be referred to as vehicle historical position data. Specifically, taking the current moment as an example, the vehicle trajectory data or vehicle position data before the current moment can be recorded as the vehicle historical trajectory information. At the same time, when encoding the vehicle historical trajectory information, the GRU hidden layer can be used to encode the vehicle trajectory data or vehicle position data before the current moment. The specific process of encoding using the GRU hidden layer can refer to some of the above embodiments and will not be elaborated here. At the same time, the hidden state output by the GRU hidden layer includes, but is not limited to, the encoding result and / or decoding result corresponding to the vehicle position.

[0059] Construct a driver intention recognition model, and obtain the driver intention from the vehicle trajectory dataset through this driver intention recognition model. Among them, the structural schematic diagram of the constructed driver intention recognition model is as Figure 3 shown. As Figure 3As shown in the figure, when building a driver intention recognition model, the encoded result output by the GRU hidden layer can be used as the input of the fully connected layer. The fully connected layer can be set to have three layers, with 128 neurons in each layer. Then, an activation function is added to the fully connected layer to perform a non-linear mapping on the encoded results of vehicle lane change data points and vehicle trajectory data, accelerating the training and reducing gradient disappearance. When the predicted probability of driving behavior intention is obtained, the cross-entropy error function is used to calculate the difference between the probability distribution predicted by the driver intention recognition model and the probability distribution of the true label. Then, the output of the fully connected layer is connected and normalized through the normalization function softmax(), and the output of the driver intention recognition model is respectively converted into the probability distributions of left lane change, right lane change, and lane keeping, and the corresponding driving behavior intention recognition result can be obtained, that is: Behavior = {LL, LR, LK}, where Behavior represents the driver's lane change behavior, LL represents left lane change, LR represents right lane change, and LK represents lane keeping. Specifically, when obtaining the driver's intention from the vehicle trajectory dataset through the driver intention recognition model, the vehicle lane change data points can be first obtained from the vehicle trajectory dataset, and then the driving behavior intention prediction can be performed according to the vehicle lane change data points and the encoded results of the vehicle trajectory data, and the driving behavior intention prediction result can be obtained. Among them, the specific process of performing the driving behavior intention prediction according to the vehicle lane change data points and the encoded results of the vehicle trajectory data can refer to some of the above embodiments and will not be elaborated here.

[0060] Then, an attention mechanism module is introduced to correct the encoded result of the vehicle trajectory data to ensure the effectiveness of the historical trajectory information and reduce the prediction time. Since the recording of historical vehicle trajectory information in the RNN is decaying, as the prediction time length increases, the decoder will forget more historical vehicle trajectory data. Therefore, by introducing the attention mechanism to correct the problem of loss of historical vehicle position data, not only can more historical information be provided to improve the prediction accuracy, but also the vehicle trajectory expected by the driver can be predicted within a longer prediction range. Among them, the process of the attention mechanism module correcting the encoded result of the vehicle trajectory data can refer to some of the above embodiments and will not be elaborated here.

[0061] Build a vehicle trajectory prediction model and perform vehicle trajectory prediction through the vehicle trajectory prediction model. Among them, the structural schematic diagram of the built vehicle trajectory prediction model is as Figure 4 shown. As Figure 4 shown, on the one hand, the vehicle trajectory prediction model receives the output containing the GRU hidden layer and the correction of the attention mechanism, and then can also combine the driver's intention to complete the trajectory prediction of different driving behaviors and perform iteration according to the prediction results to achieve long-term vehicle trajectory prediction. Among them, Figure 4The GRU hidden layer and the GRU historical hidden layer in it are only used to distinguish the GRU hidden layers at different times. That is, the GRU hidden layer can correspond to the current time, and the GRU historical hidden layer can correspond to the historical times before the current time. Specifically, the process of the vehicle trajectory prediction model performing vehicle trajectory prediction based on the corrected vehicle trajectory data encoding result and the driving behavior intention prediction result can be referred to some of the above embodiments, and will not be elaborated here. As an example, the vehicle trajectory prediction model and the driver intention recognition model can share an encoder, thereby reducing the algorithm complexity.

[0062] Therefore, in this embodiment, the recurrent neural network based on GRU encodes the vehicle historical trajectory data, and the encoding result is output to the fully connected layer and the SoftMax function to give the probability distribution that conforms to the driver's behavior intention. Then, based on the encoding result and the driver behavior intention distribution data, the predicted trajectory that meets the requirements can be given.

[0063] In summary, the present application provides a vehicle trajectory prediction method. By extracting vehicle lane-changing data points from a trajectory dataset, and then predicting the driving behavior intention based on the vehicle lane-changing data points and the encoding result of the vehicle trajectory data to obtain the prediction result of the driving behavior intention; then correcting the encoding result of the vehicle trajectory data, and predicting the vehicle trajectory based on the corrected encoding result of the vehicle trajectory data and the prediction result of the driving behavior intention; wherein, the trajectory dataset is obtained based on the vehicle trajectory data, the vehicle trajectory data includes the vehicle positions at multiple moments, the vehicle lane-changing data points include the vehicle lane-changing starting point and / or the vehicle lane-changing ending point, and the encoding result of the vehicle trajectory data is obtained by encoding the vehicle trajectory data. It can be seen from this that this method predicts the vehicle trajectory based on the encoding result of the vehicle trajectory data and the prediction result of the driving behavior intention, which can not only combine the vehicle historical trajectory data, but also fully consider the driving behavior intention of the driver, thereby improving the accuracy of vehicle trajectory prediction and being able to adapt to more complex traffic scenarios. At the same time, by predicting the vehicle trajectory, it can also assist the intelligent driving system in reasonably planning the path, selecting a safe and efficient driving route, and improving the driving stability and reliability. Moreover, by predicting the vehicle trajectory, it can help the intelligent driving system to control the vehicle more smoothly, reduce frequent acceleration and deceleration, improve the riding comfort, and based on the potential dangerous situations identified in advance, it can give early warnings, allowing the driver to have more time to take preventive measures, thereby reducing the driver's anxiety and stress and enhancing the driving experience. Therefore, this method can not only more accurately predict the future trajectory by combining the vehicle historical trajectory information and the driver's intention, but also has higher predictability and accuracy, and the trajectory prediction based on intention recognition can significantly reduce the root mean square error between the predicted trajectory and the real trajectory, thereby improving the trajectory prediction accuracy. At the same time, this method can also respond to changes in the traffic environment in real time, such as road construction, traffic accidents, etc., and can adjust the prediction result in time to ensure the safety and smoothness of vehicle driving.

[0064] In another exemplary embodiment of the present application, as Figure 5 shown, this embodiment also provides a vehicle trajectory prediction system, including:

[0065] A data acquisition module 510, configured to obtain a trajectory dataset, the trajectory dataset is obtained based on vehicle trajectory data, and the vehicle trajectory data includes vehicle positions at multiple moments;

[0066] A lane-changing data point module 520, configured to extract vehicle lane-changing data points from the trajectory dataset, and the vehicle lane-changing data points include the vehicle lane-changing starting point and / or the vehicle lane-changing ending point;

[0067] A driving behavior intention prediction module 530 is configured to perform driving behavior intention prediction based on vehicle lane change data points and the encoding result of vehicle trajectory data to obtain a driving behavior intention prediction result, where the encoding result of vehicle trajectory data is obtained by encoding vehicle trajectory data;

[0068] A vehicle trajectory prediction module 540 is configured to correct the encoding result of vehicle trajectory data and perform vehicle trajectory prediction based on the corrected encoding result of vehicle trajectory data and the driving behavior intention prediction result.

[0069] It can be understood that the vehicle trajectory prediction system provided in the above embodiment and the vehicle trajectory prediction method provided in the above embodiment belong to the same concept. The specific manner of performing operations in the vehicle trajectory prediction method has been described in detail in the above embodiment and will not be elaborated here. In practical applications, the vehicle trajectory prediction system provided in the above embodiment can, as needed, allocate the above functions to different functional modules, that is, divide the internal structure of the vehicle trajectory prediction system into different functional modules, and then implement all or part of the functions of the corresponding functional modules through the vehicle trajectory prediction method described in the above embodiment. For example, all or part of the functions of the data acquisition module 510 can be implemented through the relevant execution process of step S110, all or part of the functions of the lane change data point module 520 can be implemented through the relevant execution process of step S120, all or part of the functions of the driving behavior intention prediction module 530 can be implemented through the relevant execution process of step S130, and all or part of the functions of the vehicle trajectory prediction module 540 can be implemented through the relevant execution process of step S140. No specific limitation is imposed here either.

[0070] In summary, the present application provides a vehicle trajectory prediction system. By extracting vehicle lane-changing data points from a trajectory dataset, and then predicting the driving behavior intention based on the vehicle lane-changing data points and the encoding result of the vehicle trajectory data to obtain a prediction result of the driving behavior intention; then correcting the encoding result of the vehicle trajectory data, and performing vehicle trajectory prediction based on the corrected encoding result of the vehicle trajectory data and the prediction result of the driving behavior intention; wherein, the trajectory dataset is obtained based on the vehicle trajectory data, the vehicle trajectory data includes vehicle positions at multiple moments, the vehicle lane-changing data points include vehicle lane-changing starting points and / or vehicle lane-changing ending points, and the encoding result of the vehicle trajectory data is obtained by encoding the vehicle trajectory data. It can be seen from this that this system performs vehicle trajectory prediction based on the encoding result of the vehicle trajectory data and the prediction result of the driving behavior intention, which can not only combine the vehicle historical trajectory data, but also fully consider the driving behavior intention of the driver, thereby improving the accuracy of vehicle trajectory prediction and being able to adapt to more complex traffic scenarios. At the same time, by performing vehicle trajectory prediction, it can also assist the intelligent driving system in reasonable path planning, selecting a safe and efficient driving route, and improving the stability and reliability of driving. Moreover, by performing vehicle trajectory prediction, it can help the intelligent driving system control the vehicle more smoothly, reduce frequent acceleration and deceleration, improve the riding comfort, and based on the potentially dangerous situations identified in advance, early warnings can be given, allowing the driver to have more time to take preventive measures, thereby reducing the driver's anxiety and stress and enhancing the driving experience. Therefore, by combining the vehicle historical trajectory information and the driver's intention, this system can not only predict the future trajectory more accurately, but also has higher predictability and accuracy, and the trajectory prediction based on intention recognition can significantly reduce the root mean square error between the predicted trajectory and the true trajectory, thereby improving the trajectory prediction accuracy. At the same time, this system can also respond to changes in the traffic environment in real time, such as road construction, traffic accidents, etc., and can adjust the prediction result in time to ensure the safety and smoothness of vehicle driving.

[0071] An embodiment of the present application further provides a computer device, which may include a memory, a processor, and a computer program stored on the memory. The processor executes the computer program to cause the computer device to execute Figure 1 or Figure 2 the steps of the vehicle trajectory prediction method described above. Figure 6 The structural schematic diagram of a computer device 1000 is shown. Refer to Figure 6 As shown, the computer device 1000 includes: a processor 1010, a memory 1020, a power supply 1030, a display unit 1040, and an input unit 1060.

[0072] The processor 1010 is the control center of the computer device 1000, connecting each component through various interfaces and circuits. By running or executing the computer programs / instructions stored in the memory 1020, it performs various functions of the computer device 1000, thereby monitoring the computer device 1000 as a whole. In the embodiments of the present application, when the processor 1010 calls the computer program stored in the memory 1020, it executes the steps of the vehicle trajectory prediction method as described in Figure 1 or Figure 2 the above. Optionally, the processor 1010 may include one or more processing units; preferably, the processor 1010 may integrate an application processor and a modem processor. Among them, the application processor mainly processes the operating system, user interface, applications, etc., and the modem processor mainly processes wireless communication. In some embodiments, the processor and the memory may be implemented on a single chip, and in some embodiments, they may also be implemented separately on independent chips.

[0073] The memory 1020 may mainly include a program storage area and a data storage area. Among them, the program storage area may store the operating system, various applications, etc.; the data storage area may store instruction data created according to the use of the computer device 1000. In addition, the memory 1020 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0074] The computer device 1000 further includes a power supply 1030 (such as a battery) for powering each component. The power supply can be logically connected to the processor 1010 through a power management system, thereby realizing functions such as management of charging, discharging, and power consumption through the power management system.

[0075] The display unit 1040 can be used to display the information input by the user or the information provided to the user, as well as various menus of the computer device 1000. In the embodiments of the present application, it is mainly used to display the display interfaces of various applications in the computer device 1000 and the objects such as text and pictures displayed in the display interfaces. The display unit 1040 may include a display panel 1050. The display panel 1050 may be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED), etc.

[0076] The input unit 1060 can be used to receive information such as numbers or characters input by the user. The input unit 1060 can include a touch panel 1070 and other input devices 1080. Among them, the touch panel 1070, also known as a touch screen, can collect touch operations of the user on or near it (such as operations of the user using any suitable object or accessory such as a finger or a stylus on or near the touch panel 1070).

[0077] Specifically, the touch panel 1070 can detect the touch operation of the user, detect the signals brought by the touch operation, convert these signals into contact coordinates, send them to the processor 1010, and receive and execute the commands sent by the processor 1010. In addition, the touch panel 1070 can be implemented in multiple types such as resistive, capacitive, infrared, and surface acoustic wave. The other input devices 1080 can include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control keys, power on / off keys, etc.), a trackball, a mouse, a joystick, etc.

[0078] Of course, the touch panel 1070 can cover the display panel 1050. After the touch panel 1070 detects a touch operation on or near it, it transmits the operation to the processor 1010 to determine the type of touch event. Subsequently, the processor 1010 provides a corresponding visual output on the display panel 1050 according to the type of touch event. Although in Figure 6 , the touch panel 1070 and the display panel 1050 are implemented as two independent components to realize the input and output functions of the computer device 1000, but in some embodiments, the touch panel 1070 and the display panel 1050 can be integrated to realize the input and output functions of the computer device 1000.

[0079] The computer device 1000 can also include one or more sensors, such as a pressure sensor, a gravitational acceleration sensor, a proximity light sensor, etc. Of course, according to the needs in specific applications, the above computer device 1000 can also include other components such as a camera.

[0080] The embodiment of the present application also provides a computer-readable storage medium. The computer program / instructions are stored in the storage medium. When the computer program / instructions are executed by the processor, the above device can execute the steps of the vehicle trajectory prediction method as described in Figure 1 or Figure 2 of the present application.

[0081] Those skilled in the art can understand that Figure 6The above are only examples of computer devices and do not constitute limitations on such devices. Such devices may include more or fewer components than shown, or combine certain components, or have different components. For convenience of description, the above parts are divided into various modules (or units) according to functions and described separately. Of course, when implementing this application, the functions of the various modules (or units) may be implemented in the same or multiple software or hardware components.

[0082] Those skilled in the art can understand that this application may be implemented in the form of a computer program product on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes. This application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of this application. It should be understood that each flow and / or block in the flowchart and / or block diagram, and the combination of flows and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be applied to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices generate means for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or combinations of blocks. These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that implement the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or combinations of blocks. These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in Figure 1 one or more of the flows Figure 1 or blocks or combinations of blocks.

[0083] In another exemplary embodiment of the present application, this embodiment further provides a vehicle, which includes a vehicle trajectory prediction system as described in the above embodiment, or applies the vehicle trajectory prediction method as described in the above embodiment. It can be understood that since the specific ways of operating the vehicle trajectory prediction system and the vehicle trajectory prediction method have been described in detail in the embodiment, the technical functions and effects of the vehicle provided in this embodiment can be referred to the above embodiment, and will not be elaborated here. As an example, for instance, the vehicle may be a vehicle with an intelligent driving system. As another example, for instance, the vehicle may be a vehicle with an autonomous driving system. As yet another example, for instance, the vehicle may be a vehicle with both an intelligent driving system and an autonomous driving system.

[0084] It can be understood that when the above embodiment processes relevant data (such as vehicle trajectory data, etc.) in terms of collection, storage, use, processing, transmission, provision, disclosure, deletion, etc., it is completed with the consent of the user or after obtaining the consent. For example, the vehicle trajectory data is authorized with the user's knowledge and consent; or it is actively provided by the user after reading the relevant instructions, or when the user uses some or all of the functions described in the above embodiment, actively authorizes / provides / uploads it, or is obtained through other means / ways with the consent of the user or after obtaining the consent.

[0085] It can be understood that although terms such as first and second may be used in the embodiments of the present application to describe moments, etc., these terms are only used to distinguish moments from each other. For example, without departing from the scope of the embodiments of the present application, the first moment can also be referred to as the second moment, and similarly, the second moment can also be referred to as the first moment.

[0086] The above embodiments only illustratively explain the principles and effects of the present application, rather than limiting the present application. Any person familiar with this technology can modify or change the above embodiments without departing from the spirit and scope of the present application. Therefore, all equivalent modifications or changes completed by those with ordinary knowledge in the technical field without departing from the spirit and technical idea disclosed in the present application should still be covered by the claims of the present application.

Claims

1. A vehicle trajectory prediction method, characterized in that, The method includes the following steps: Obtain a trajectory dataset, which is obtained based on vehicle trajectory data, and the vehicle trajectory data includes vehicle positions at multiple moments; Extract vehicle lane-changing data points from the trajectory dataset, and the vehicle lane-changing data points include vehicle lane-changing starting points and / or vehicle lane-changing ending points; Perform driving behavior intention prediction according to the vehicle lane-changing data points and the encoding result of the vehicle trajectory data to obtain a driving behavior intention prediction result, and the encoding result of the vehicle trajectory data is obtained by encoding the vehicle trajectory data; Correct the encoding result of the vehicle trajectory data, and perform vehicle trajectory prediction based on the corrected encoding result of the vehicle trajectory data and the driving behavior intention prediction result.

2. The vehicle trajectory prediction method according to claim 1, wherein The process of performing driving behavior intention prediction according to the vehicle lane-changing data points and the encoding result of the vehicle trajectory data to obtain a driving behavior intention prediction result includes: Input the vehicle lane-changing data points and the encoding result of the vehicle trajectory data into a fully connected layer composed of multiple neurons, and add an activation function in the fully connected layer to perform non-linear mapping on the vehicle lane-changing data points and the encoding result of the vehicle trajectory data to obtain a driving behavior intention prediction probability; Perform normalization processing on the driving behavior intention prediction probability, and determine the corresponding driving behavior intention prediction result according to the normalization processing result of the driving behavior intention prediction probability; wherein, the driving behavior intention prediction result includes a left lane change from the current lane to the left lane, a right lane change from the current lane to the right lane, and / or lane keeping without changing the current lane.

3. The vehicle trajectory prediction method according to claim 1, wherein, The process of correcting the encoding result of the vehicle trajectory data includes: Obtain a matrix determined in advance or in real time, and calculate a correction coefficient based on the matrix, some or all of the encoding results in the encoding result of the vehicle trajectory data, and the vehicle position corresponding to the last moment in the vehicle trajectory data; Perform normalization processing on the correction coefficient, and correct some or all of the encoding results in the encoding result of the vehicle trajectory data through the normalization processing result of the correction coefficient to obtain a corrected encoding result of the vehicle trajectory data.

4. The vehicle trajectory prediction method according to claim 1 or 3, characterized in that The process of performing vehicle trajectory prediction based on the corrected encoding result of the vehicle trajectory data and the driving behavior intention prediction result includes: Obtain a decoding result of the vehicle trajectory data corresponding to the encoding result of the vehicle trajectory data, and select the decoding result at the target moment from the decoding result of the vehicle trajectory data; wherein, the target moment is one of the moment sets corresponding to the vehicle trajectory data; Use the decoding result at the target moment as the initial state, and output the decoding result at the first moment based on the corrected encoding result of the vehicle trajectory data and the driving behavior intention prediction result, and predict the vehicle position at the second moment through the decoding result at the first moment; wherein, the first moment is after the target moment and differs from the target moment by one moment, and the second moment is after the first moment and differs from the first moment by one moment; Continue the iterative prediction based on the predicted vehicle position at the second moment, and obtain the vehicle trajectory prediction result according to the vehicle positions at multiple predicted moments.

5. The vehicle trajectory prediction method according to any one of claims 1 to 3, characterized in that The process of obtaining the encoded result of the vehicle trajectory data by encoding the vehicle trajectory data includes: Filter out the vehicle position at the target moment from the vehicle trajectory data, where the target moment is one of the moments in the set of moments corresponding to the vehicle trajectory data; Encode the vehicle position at the target moment using a pre-determined or real-time target neural network to obtain the encoded result at the target moment; and, Use the encoded result at the target moment and the vehicle position at the next moment after the target moment as the input of the target neural network, encode the vehicle position at the next moment through the target neural network, output the encoded result at the next moment, and continue to input the encoded result at the next moment into the target neural network for encoding until the encoded result of the vehicle trajectory data corresponding to the vehicle trajectory data is obtained; where, if the target moment is the last moment in the set of moments corresponding to the vehicle trajectory data, the next moment after the target moment is the current moment.

6. The vehicle trajectory prediction method according to claim 1, wherein, The process of extracting vehicle lane change data points from the trajectory data set includes: Extract the vehicle position at the first moment and the vehicle position at the second moment from the trajectory data set, where the first moment and the second moment are separated by a preset number of moments; Based on the vehicle position at the first moment and the vehicle position at the second moment, calculate the position deviation of the vehicle over a continuous preset number of moments; and when the position deviation is less than the preset deviation threshold, use the vehicle position at the first moment or the vehicle position at the second moment as the vehicle lane change data point.

7. The vehicle trajectory prediction method according to claim 1, wherein The process of obtaining the trajectory data set based on the vehicle trajectory data includes: Obtain the vehicle trajectory data obtained in multiple traffic scenarios in advance, denoted as the original vehicle trajectory data; Preprocess the original vehicle trajectory data, and obtain the trajectory data set based on the preprocessed vehicle trajectory data; where the preprocessing includes denoising, interpolation, and / or coordinate transformation.

8. A vehicle trajectory prediction system, characterized in that, The system includes: A data acquisition module for obtaining a trajectory data set, the trajectory data set being obtained based on vehicle trajectory data, the vehicle trajectory data including vehicle positions at multiple moments; A lane change data point module for extracting vehicle lane change data points from the trajectory data set, the vehicle lane change data points including vehicle lane change start points and / or vehicle lane change end points; A driving behavior intention prediction module for predicting the driving behavior intention according to the vehicle lane change data points and the encoded result of the vehicle trajectory data to obtain the driving behavior intention prediction result, the encoded result of the vehicle trajectory data being obtained by encoding the vehicle trajectory data; A vehicle trajectory prediction module for correcting the encoded result of the vehicle trajectory data, and performing vehicle trajectory prediction based on the corrected encoded result of the vehicle trajectory data and the driving behavior intention prediction result.

9. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the steps of the vehicle trajectory prediction method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored thereon. When the computer program is executed by a processor, it implements the steps of the vehicle trajectory prediction method described in any one of claims 1 to 7.

Images