Dynamic path recommendation method and device, in-vehicle infotainment system and storage medium

Abstract

The invention provides a dynamic path recommendation method and device, an in-vehicle infotainment system and a storage medium. The method comprises the following steps: extracting the time and position of a trajectory data point in a historical driving trajectory of a vehicle, and generating an interest point time domain vector for the point; extracting the destination type, position and arrival time in the historical navigation record of the user, and generating a point-of-interest airspace vector; matching the interest point space domain vectors of the same and similar points with the interest point time domain vectors to obtain the residence time length of the interest points, and generating a user behavior habit space-time feature map; performing multi-layer space-time convolution processing on the user behavior habit space-time feature map in sequence, performing normalization processing on features obtained by processing the previous layer of space-time convolution network, and inputting the normalized features into the next layer of space-time convolution network to obtain a user behavior habit thermodynamic map; and determining a recommended path according to the current position of the vehicle, the current real-time road condition and the user behavior habit thermodynamic diagram. According to the method, customized dynamic path recommendation conforming to user behavior habits can be obtained.

Description

Technical Field

[0001] Embodiments of the present invention generally relate to the field of vehicle navigation technology, and more specifically, to a dynamic route recommendation method, device, vehicle system and storage medium. Background Art

[0002] With the rapid development of the domestic economy and the increase in population density in large cities, the number of cars in cities has continued to increase and urban traffic problems have become increasingly serious.

[0003] An efficient dynamic route guidance algorithm is one of the effective ways to improve the utilization rate of road resources and traffic speed and alleviate traffic congestion. However, according to research results at home and abroad, most of the existing dynamic route guidance algorithms only calculate routes dynamically based on real-time traffic conditions, and only use traffic information as a prerequisite for route calculation, without considering the behavior of travelers, which limits the user's choice space. When encountering unavoidable congestion, no other solutions can be given, which brings a bad driving experience to the car owner and cannot bring satisfactory emotional value to the car owner. Summary of the invention

[0004] In order to solve the above problems in the prior art, in a first aspect, an embodiment of the present invention provides a dynamic path recommendation method, the method comprising: obtaining a vehicle historical driving trajectory and a user historical navigation record; extracting the time and position of a trajectory data point in the vehicle historical driving trajectory; generating a point of interest time domain vector according to the time and position of the trajectory data point; extracting the destination type, destination location and arrival time of each record in the user historical navigation record; generating a point of interest space domain vector according to the destination type, destination location and arrival time extracted in the user historical navigation record; matching each point of interest in the point of interest space domain vector with a point in the same or adjacent position in the point of interest time domain vector to obtain the length of time spent at each point of interest; According to the length of time stayed at each point of interest, the location of the point of interest, and the type of the point of interest, a user behavior habit spatiotemporal feature map is generated; the user behavior habit spatiotemporal feature map is converted into a graph data format; multiple spatiotemporal convolution processes are performed on the user behavior habit spatiotemporal feature map in the graph data format in sequence through a multi-layer spatiotemporal convolutional network, wherein each layer in the multi-layer spatiotemporal convolutional network includes a temporal graph convolutional network and a spatial graph convolutional network, and wherein the features obtained by the previous layer of spatiotemporal convolutional network processing are normalized and then input into the next layer of spatiotemporal convolutional network; the features obtained by the multi-layer spatiotemporal convolutional network are normalized and up-sampled to obtain a user behavior habit heat map; and a recommended path is determined according to the current position of the vehicle, the current real-time road conditions, and the user behavior habit heat map.

[0005] In some embodiments, after generating the spatial-temporal feature map of user behavior habits, the method further includes: assigning a weight to each point of interest through an attention mechanism according to the priority of each point of interest; and updating the spatial-temporal feature map of user behavior habits according to the weight assigned to each point of interest.

[0006] In some implementations, the method further includes: receiving a user operation on the recommended route; and updating the weight of the point of interest that is the destination of the recommended route according to the user operation.

[0007] In some embodiments, performing multiple spatiotemporal convolution processes includes: inputting the unnormalized features output by the first spatiotemporal convolutional network layer in the multi-layer spatiotemporal convolutional network into each subsequent spatiotemporal convolutional network layer.

[0008] In some embodiments, performing multiple spatiotemporal convolution processes includes: in each layer of the spatiotemporal convolution network, first processing using a temporal graph convolution network, and then processing using a spatial graph convolution network.

[0009] In some embodiments, the method further includes: after all points of interest in the spatial domain vector of the points of interest have been matched, if there are still points of interest that have stayed for a long time and have not been matched in the temporal domain vector of the points of interest, extracting the positions, arrival times and stay times of these points of interest; determining the types of the points of interest based on the positions of the points of interest; and filling the extracted positions, arrival times, stay times and types of the points of interest into the spatial-temporal feature map of the user behavior habits.

[0010] In some embodiments, the multi-layer spatiotemporal convolutional network includes a three-layer spatiotemporal convolutional network.

[0011] In some embodiments, the method further includes: sending a prompt message through a human-computer interface indicating whether to update the model; receiving a user's choice of the prompt message; and if the user chooses to update the model, upgrading the model of the multi-layer spatiotemporal convolutional network using a recent spatiotemporal feature map of user behavior habits.

[0012] In a second aspect, an embodiment of the present invention proposes a dynamic path recommendation device, the device comprising: a historical data acquisition module, configured to acquire a vehicle historical driving trajectory and a user historical navigation record; a trajectory data extraction module, configured to extract the time and position of the trajectory data points in the vehicle historical driving trajectory; a time domain vector generation module, configured to generate a time domain vector of a point of interest according to the time and position of the trajectory data points; a navigation data extraction module, configured to extract the destination type, destination location and arrival time of each record in the user historical navigation record; a spatial domain vector generation module, configured to generate a spatial domain vector of a point of interest according to the destination type, destination location and arrival time extracted from the user historical navigation record; a matching module, configured to match each point of interest in the spatial domain vector of the point of interest with a point in the time domain vector of the point of interest with the same or adjacent position, and obtain the length of time spent at each point of interest; A space-time feature map generation module is configured to generate a space-time feature map of user behavior habits based on the length of time spent at each point of interest, the location of the point of interest, and the type of the point of interest; a format conversion module is configured to convert the space-time feature map of user behavior habits into a graph data format; a space-time convolution module is configured to sequentially perform multiple space-time convolution processes on the space-time feature map of user behavior habits in the graph data format through a multi-layer space-time convolution network, wherein each layer in the multi-layer space-time convolution network includes a temporal graph convolution network and a spatial graph convolution network, and wherein the features obtained by the previous layer of space-time convolution network processing are normalized and then input into the next layer of space-time convolution network; a heat map acquisition module is configured to normalize and upsample the features obtained by the multi-layer space-time convolution network to obtain a user behavior habit heat map; a path determination module is configured to determine a recommended path based on the current position of the vehicle, the current real-time road conditions, and the user behavior habit heat map.

[0013] In some embodiments, the device also includes an attention mechanism module, which is configured to: after generating a spatial-temporal feature map of user behavior habits, assign a weight to each point of interest through an attention mechanism according to the priority of each point of interest; and update the spatial-temporal feature map of user behavior habits according to the weight assigned to each point of interest.

[0014] In some embodiments, the device further includes a weight updating module configured to: receive a user operation on the recommended path; and update the weight of the point of interest that is the destination of the recommended path according to the user operation.

[0015] In some embodiments, the spatiotemporal convolution module is further configured to: input the unnormalized features output by the first layer of the spatiotemporal convolution network in the multi-layer spatiotemporal convolution network into each subsequent layer of the spatiotemporal convolution network.

[0016] In a third aspect, an embodiment of the present invention proposes a vehicle system, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the dynamic path recommendation method described in any of the above embodiments is implemented.

[0017] In a fourth aspect, an embodiment of the present invention provides a storage medium storing computer-readable instructions, which, when executed by a processor, executes the dynamic path recommendation method described in any of the above embodiments.

[0018] An embodiment of the present invention provides a dynamic route guidance solution, which not only recommends routes based on the real-time traffic conditions in the city, but also guides users to appropriate points of interest to carry out corresponding activities according to their behavior habits when congestion occurs, and waits for the congestion to ease or end.

[0019] The technical solution adopted by the implementation mode of the present invention to solve its technical problems is: establishing a heat map of urban points of interest based on user behavior habits. In order to establish the heat map of points of interest, it is first necessary to obtain the vehicle driving trajectory and the user navigation record, and then fuse the two types of data to obtain the user behavior habit space-time feature map, and then establish a graph neural network model for the data, and finally send the user behavior habit space-time feature map into the model for training to obtain a heat map of urban points of interest for guiding users.

[0020] Whenever a user encounters unavoidable congestion while driving, the car system will make congestion avoidance recommendations to the user based on the user's city interest point heat map, guiding the user to the interest points that match their behavioral habits, thereby improving user experience and satisfaction. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and other objects, features and advantages of the embodiments of the present invention will become readily understood by reading the following detailed description with reference to the accompanying drawings. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, in which:

[0022] Figure 1 A flow chart of a dynamic path recommendation method according to an embodiment of the present invention is shown;

[0023] Figure 2 A schematic diagram of generating a space-time feature map according to an embodiment of the present invention is shown;

[0024] Figure 3 A schematic diagram showing a first example of spatiotemporal convolution processing according to an embodiment of the present invention;

[0025] Figure 4 A schematic diagram showing a second example of spatiotemporal convolution processing according to an embodiment of the present invention;

[0026] Figure 5 A schematic diagram showing a third example of spatiotemporal convolution processing according to an embodiment of the present invention;

[0027] Figure 6 A block diagram of a dynamic path recommendation device according to an embodiment of the present invention is shown.

[0028] In the drawings, the same or corresponding reference numerals represent the same or corresponding parts. DETAILED DESCRIPTION

[0029] The principle and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided only to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way.

[0030] In one aspect, an embodiment of the present invention provides a dynamic path recommendation method. Figure 1 , which shows a flow chart of a dynamic route recommendation method 100 according to an embodiment of the present invention. As an example, the method can be executed on a vehicle system, or can be executed on a server (e.g., a cloud server), or some steps are executed on the vehicle system and some steps are executed on the server, and the present invention is not limited in this respect.

[0031] like Figure 1 As shown, the method 100 includes steps S101-S111.

[0032] In step S101, the vehicle's historical driving trajectory and the user's historical navigation record are obtained.

[0033] In step S102, the time and position (eg, longitude and latitude) of the trajectory data points in the historical driving trajectory of the vehicle are extracted.

[0034] In step S103, a time domain vector of the point of interest is generated according to the time and position of the trajectory data point.

[0035] In step S104, the destination type, destination location and arrival time of each record in the user's historical navigation record are extracted.

[0036] In step S105, a spatial vector of the point of interest is generated according to the destination type, destination location and arrival time extracted from the user's historical navigation record.

[0037] In step S106, each point of interest in the spatial domain vector of the point of interest is matched with a point in the temporal domain vector of the point of interest that is at the same position or adjacent to the point of interest, and the length of time spent at each point of interest is obtained.

[0038] In step S107, a spatial-temporal characteristic map of user behavior habits is generated based on the length of time spent at each point of interest, the location of the point of interest, and the type of the point of interest.

[0039] refer to Figure 2 , which shows a schematic diagram of generating a space-time feature map according to an embodiment of the present invention. Through the above steps, in the process of generating the space-time feature map of user behavior habits, the vehicle driving trajectory and the user navigation record are data fused. In order to obtain fused data, the vehicle driving trajectory and the user navigation record are firstly refined. For the vehicle driving trajectory, it is necessary to extract the time and longitude and latitude of the trajectory data point, and establish the time domain vector of the point of interest for the longitude and latitude. The vector is used to record the location of the vehicle at each moment in the time dimension; and for each navigation record of the user, it is necessary to extract the longitude and latitude of the destination, the type of the destination, such as a restaurant, a shopping mall, etc., each type is converted into a corresponding enumeration value for easy calculation, and the arrival time of the destination is also recorded. After all the navigation records are extracted, the spatial vector of the point of interest is obtained, which is used to record the points of interest that the user has navigated to. After the time domain vector of the point of interest and the spatial vector of the point of interest are all obtained, each point of interest in the spatial vector is matched with the point with the same longitude and latitude or in a similar range in the time domain vector, and the time the user stays at each point of interest is obtained, so that the space-time feature map representing the user's behavior habits is obtained.

[0040] As an embodiment of the present invention, the method may also include: after all points of interest in the spatial domain vector of the points of interest have been matched, if there are still points of interest that have stayed for a long time and have not been matched in the temporal domain vector of the points of interest, extracting the positions (such as longitude and latitude), arrival time and stay time of these points of interest; determining the type of the points of interest according to the positions of the points of interest; filling the extracted positions, arrival time, stay time and type of the points of interest into the spatial-temporal feature map of user behavior habits, thereby obtaining a complete spatial-temporal feature map of user behavior habits.

[0041] In step S108, the spatial-temporal characteristic graph of user behavior habits is converted into a graph data format.

[0042] In step S109, multiple spatiotemporal convolution processes are sequentially performed on the spatiotemporal feature graph of user behavior habits in the graph data format through a multi-layer spatiotemporal convolution network, wherein each layer in the multi-layer spatiotemporal convolution network includes a temporal graph convolution network and a spatial graph convolution network, and wherein the features obtained by the previous layer of spatiotemporal convolution network processing are normalized and then input into the next layer of spatiotemporal convolution network. The multi-layer spatiotemporal convolution network and the normalization processing therebetween constitute a graph neural network model for analyzing the spatiotemporal feature graph of user behavior habits. As one embodiment of the present invention, the multi-layer spatiotemporal convolution network can be a three-layer spatiotemporal convolution network.

[0043] Since the spatial-temporal feature map of user behavior habits belongs to non-Euclidean spatial data, conventional convolutional neural networks cannot process this type of data well. Therefore, an embodiment of the present invention adopts a graph convolutional neural network to analyze and process the user behavior habit feature data. At the same time, due to the spatial-temporal characteristics of the behavior habit data, an embodiment of the present invention adopts a spatial-temporal convolution method to extract features, and finally obtains a probability distribution map of a user appearing at a certain point of interest at a certain time, that is, a user behavior habit heat map.

[0044] refer to Figure 3 , which shows a schematic diagram of a first example of spatiotemporal convolution processing according to an embodiment of the present invention.

[0045] In this first example, the spatial-temporal feature map of user behavior habits is first converted into the format of graph data to facilitate the calculation of the graph neural convolutional network. Optionally, data enhancement can also be performed. The data is then sent to a three-layer continuous spatial-temporal convolutional network model for training, where temporal convolution is used to extract temporal features and spatial convolution is used to extract spatial features. As an embodiment of the present invention, performing multiple spatial-temporal convolution processing includes: in each layer of the spatial-temporal convolutional network, first processing using a temporal graph convolutional network, and then processing using a spatial graph convolutional network. After each spatial-temporal convolution, the output data is normalized to ensure that the data distribution is as consistent with the normal distribution as possible, to ensure the rationality of the output results, and finally upsample the final data to obtain a user behavior habit heat map.

[0046] In step S110, the features obtained through the multi-layer spatiotemporal convolutional network are normalized and upsampled to obtain a user behavior habit heat map.

[0047] In step S111, the recommended route is determined based on the vehicle's current location, current real-time traffic conditions, and the user behavior habit heat map. When the user encounters congestion, the vehicle system will combine the real-time traffic conditions and the user behavior habit heat map to provide the user with a more humane and customized route guidance service.

[0048] The dynamic route recommendation method proposed in the embodiment of the present invention is not limited to real-time traffic information when performing route guidance, but also uses user behavior habit data to provide users with customized services with higher emotional value.

[0049] As an embodiment of the present invention, after generating the spatial-temporal feature map of user behavior habits, the method may further include: assigning weights to each point of interest through an attention mechanism according to the priority of each point of interest; and updating the spatial-temporal feature map of user behavior habits according to the weight assigned to each point of interest.

[0050] refer to Figure 4, which shows a schematic diagram of a second example of spatiotemporal convolution processing according to an embodiment of the present invention. In this second example, the spatiotemporal feature map of user behavior habits is first converted into the format of graph data to facilitate the calculation of the graph neural convolution network. Optionally, data enhancement can also be performed. If there are many points of interest, a corresponding priority can be assigned to each point of interest, and an attention mechanism can be used to distinguish each point of interest and give different weights. The priority can be determined based on one or more of the factors such as the dwell time of the point of interest, the type or attribute of the point of interest, the number of times visited, the location of the point of interest, the distance from home or company, etc. The data is then sent to a continuous three-layer spatiotemporal convolutional network model for training, and the temporal convolution is used to extract temporal features, and the spatial convolution is used to extract spatial features. As an embodiment of the present invention, performing multiple spatiotemporal convolution processing includes: in each layer of the spatiotemporal convolutional network, first using the temporal graph convolutional network for processing, and then using the spatial graph convolutional network for processing. After each spatiotemporal convolution, the output data is normalized to ensure that the data distribution conforms to the normal distribution as much as possible and the rationality of the output results. Finally, the final data is upsampled to obtain the user behavior habit heat map.

[0051] In the implementation method that introduces the attention mechanism, as a further implementation method, the method may also include: receiving the user's operation on the recommended path; and updating the weight of the point of interest as the destination of the recommended path according to the user's operation. After obtaining the user's permission, after the path is recommended to the user, the weight of the point of interest can be adjusted according to the user's choice of the recommended path, and the weight of the point of interest in the attention mechanism will affect the generation of the heat map, so as to adjust the recommendation strategy in time, ensure that the strategy has a certain timeliness, and bring a better user experience.

[0052] As an embodiment of the present invention, performing multiple spatiotemporal convolution processes includes: inputting the unnormalized features output by the first layer of the spatiotemporal convolution network in the multi-layer spatiotemporal convolution network into each subsequent layer of the spatiotemporal convolution network.

[0053] refer to Figure 5, which shows a schematic diagram of a third example of spatiotemporal convolution processing according to an embodiment of the present invention. In this third example, the spatiotemporal feature map of user behavior habits is first converted into a graph data format to facilitate the calculation of the graph neural convolution network. Optionally, data enhancement can also be performed. The data is then sent to a three-layer continuous spatiotemporal convolutional network model for training, where temporal convolution is used to extract temporal features and spatial convolution is used to extract spatial features. As an embodiment of the present invention, performing multiple spatiotemporal convolution processing includes: in each layer of the spatiotemporal convolutional network, first processing is performed using a temporal graph convolutional network, and then processing is performed using a spatial graph convolutional network. After each spatiotemporal convolution, the output data is normalized to ensure that the data distribution conforms to the normal distribution as much as possible and to ensure the rationality of the output results. Finally, the final data is upsampled to obtain a user behavior habit heat map. In addition, considering the sparsity of feature data, the unnormalized features output by the first layer of the spatiotemporal convolutional network in the multi-layer spatiotemporal convolutional network can be input into each subsequent layer of the spatiotemporal convolutional network, that is, a residual structure is introduced. As Figure 5 As shown, the unnormalized output of the first-layer spatiotemporal convolutional network is input into the second-layer spatiotemporal convolutional network and the third-layer spatiotemporal convolutional network. Thus, the second-layer spatiotemporal convolutional network receives the unnormalized output and the normalized output of the first-layer spatiotemporal convolutional network as input, and the third-layer spatiotemporal convolutional network receives the unnormalized output of the first-layer spatiotemporal convolutional network and the normalized output of the second-layer spatiotemporal convolutional network as input. In this example, by introducing a residual structure in the three-layer spatiotemporal convolution, the phenomenon of gradient explosion or disappearance is avoided, and the robustness of the model is improved.

[0054] As an embodiment of the present invention, the method may also include: sending a prompt message through a human-computer interface to indicate whether to update the model; receiving the user's choice of the prompt message; and when the user chooses to update the model, upgrading the model of the multi-layer spatiotemporal convolutional network using a recent spatiotemporal feature map of user behavior habits.

[0055] For example, the vehicle system can periodically ask the user whether to update the model. After obtaining the user's consent, for example, the recent spatial-temporal feature map of the user's behavior habits can be uploaded to the cloud to fine-tune the model and upgrade the model online to provide users with a better user experience, thereby meeting the user's ever-changing needs.

[0056] On the other hand, an embodiment of the present invention provides a dynamic path recommendation device, referring to Figure 6 , which shows a block diagram of a dynamic path recommendation device according to an embodiment of the present invention. The device includes modules 601-611.

[0057] The historical data acquisition module 601 can be configured to acquire the vehicle's historical driving trajectory and the user's historical navigation record.

[0058] The trajectory data extraction module 602 may be configured to extract the time and location of trajectory data points in the historical driving trajectory of the vehicle.

[0059] The time domain vector generation module 603 may be configured to generate a time domain vector of an interest point according to the time and position of the trajectory data point.

[0060] The navigation data extraction module 604 may be configured to extract the destination type, destination location, and arrival time of each record in the user's historical navigation records.

[0061] The airspace vector generation module 605 may be configured to generate an airspace vector of the point of interest according to the destination type, destination location and arrival time extracted from the user's historical navigation record.

[0062] The matching module 606 may be configured to match each interest point in the interest point spatial domain vector with a point in the interest point temporal domain vector that is located at the same or adjacent position, and obtain the length of time spent at each interest point.

[0063] The space-time feature map generation module 607 can be configured to generate a space-time feature map of user behavior habits based on the length of time spent at each point of interest, the location of the point of interest, and the type of the point of interest.

[0064] The format conversion module 608 may be configured to convert the spatial-temporal feature graph of user behavior habits into a graph data format.

[0065] The spatiotemporal convolution module 609 can be configured to perform multiple spatiotemporal convolution processing on the spatiotemporal feature graph of user behavior habits in graph data format in sequence through a multi-layer spatiotemporal convolution network, wherein each layer in the multi-layer spatiotemporal convolution network includes a temporal graph convolution network and a spatial graph convolution network, and wherein the features obtained by the previous layer of spatiotemporal convolution network processing are normalized and then input into the next layer of spatiotemporal convolution network.

[0066] The heat map acquisition module 610 can be configured to normalize and upsample the features obtained through the multi-layer spatiotemporal convolutional network to obtain a user behavior habit heat map.

[0067] The route determination module 611 may be configured to determine a recommended route based on the vehicle's current location, current real-time road conditions, and a user behavior habit heat map.

[0068] As an embodiment of the present invention, the device may also include an attention mechanism module, which is configured to: after generating a spatial-temporal feature map of user behavior habits, assign a weight to each point of interest through an attention mechanism according to the priority of each point of interest; and update the spatial-temporal feature map of user behavior habits according to the weight assigned to each point of interest.

[0069] As an embodiment of the present invention, the device may further include a weight updating module configured to: receive a user's operation on the recommended path; and update the weight of the point of interest as the destination of the recommended path according to the user's operation.

[0070] As an embodiment of the present invention, the spatiotemporal convolution module can be further configured to: input the unnormalized features output by the first layer of the spatiotemporal convolution network in the multi-layer spatiotemporal convolution network into each subsequent layer of the spatiotemporal convolution network.

[0071] It should be noted that the functions implemented by each module in the dynamic path recommendation device proposed in the embodiment of the present invention correspond one-to-one to each step of the dynamic path recommendation method described above. For other specific implementations, examples and beneficial effects of the device, please refer to the description of the method above.

[0072] On the other hand, an embodiment of the present invention provides a vehicle system, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the dynamic route recommendation method described in any of the above embodiments is implemented.

[0073] In yet another aspect, an embodiment of the present invention provides a storage medium storing computer-readable instructions, which, when executed by a processor, executes the dynamic path recommendation method described in any of the above embodiments.

[0074] An embodiment of the present invention provides a dynamic route guidance solution, which not only recommends routes based on the real-time traffic conditions in the city, but also guides users to appropriate points of interest to carry out corresponding activities according to their behavior habits when congestion occurs, and waits for the congestion to ease or end.

[0075] The technical solution adopted by the implementation mode of the present invention to solve its technical problems is: establishing a heat map of urban points of interest based on user behavior habits. In order to establish the heat map of points of interest, it is first necessary to obtain the vehicle driving trajectory and the user navigation record, and then fuse the two types of data to obtain the user behavior habit space-time feature map, and then establish a graph neural network model for the data, and finally send the user behavior habit space-time feature map into the model for training to obtain a heat map of urban points of interest for guiding users.

[0076] Whenever a user encounters unavoidable congestion while driving, the car system will make congestion avoidance recommendations to the user based on the user's city interest point heat map, guiding the user to the interest points that match his or her behavioral habits to prevent the user from being trapped in the vast traffic flow.

[0077] In addition, by regularly asking users whether to update the model and obtaining their consent, we upload recent spatial-temporal feature maps of user behavior habits to the cloud for model fine-tuning and online model upgrade to provide users with a better user experience, thus meeting their ever-changing needs.

[0078] For the purpose of illustration, the foregoing description of the embodiments of the present invention has been given, which is not exhaustive nor intended to limit the present invention to the disclosed exact form. It will be appreciated by those skilled in the art that various changes can be made without departing from the scope of the present invention, and that the elements therein can be replaced with equivalents. In addition, without departing from the basic scope of the present invention, many modifications can be made so that specific situations or materials are adapted to the teachings of the present invention. Therefore, the present invention is not intended to be limited to the specific embodiments disclosed as the best mode for realizing the present invention, and the present invention will include all embodiments falling within the scope of the appended claims.

Claims

1. A dynamic path recommendation method, characterized in that: The method comprises: Obtain vehicle historical driving trajectory and user historical navigation records; Extracting the time and location of trajectory data points in the historical driving trajectory of the vehicle; Generate a time domain vector of a point of interest according to the time and position of the trajectory data point; Extracting the destination type, destination location and arrival time of each record in the user's historical navigation record; Generate a point of interest airspace vector according to the destination type, destination location and arrival time extracted from the user's historical navigation record; Matching each point of interest in the spatial domain vector of the point of interest with a point in the same or adjacent position in the temporal domain vector of the point of interest to obtain the length of time spent at each point of interest; Generate a spatial-temporal feature map of user behavior habits based on the length of time spent at each point of interest, the location of the point of interest, and the type of the point of interest; Converting the user behavior habit space-time feature graph into a graph data format; The user behavior habit spatiotemporal feature graph in the graph data format is sequentially subjected to multiple spatiotemporal convolution processing by a multi-layer spatiotemporal convolution network, wherein each layer of the multi-layer spatiotemporal convolution network includes a temporal graph convolution network and a spatial graph convolution network, and wherein the features obtained by the previous layer of spatiotemporal convolution network processing are normalized and then input into the next layer of spatiotemporal convolution network; Normalizing and upsampling the features obtained through the multi-layer spatiotemporal convolutional network to obtain a user behavior habit heat map; The recommended route is determined based on the vehicle's current location, current real-time traffic conditions and the user's behavior habit heat map.

2. The method according to claim 1, characterized in that After generating the spatial-temporal characteristic graph of user behavior habits, the method further includes: According to the priority of each point of interest, a weight is assigned to each point of interest through the attention mechanism; According to the weight assigned to each point of interest, the spatial-temporal feature map of user behavior habits is updated.

3. The method according to claim 2, characterized in that The method further comprises: receiving a user's operation on the recommended path; The weight of the point of interest that is the destination of the recommended route is updated according to the user's operation.

4. The method according to claim 1, characterized in that: Performing multiple spatiotemporal convolutions includes: The unnormalized features output by the first layer of the spatiotemporal convolutional network in the multi-layer spatiotemporal convolutional network are input into each subsequent layer of the spatiotemporal convolutional network.

5. The method according to claim 1, characterized in that Performing multiple spatiotemporal convolutions includes: In each layer of the spatiotemporal convolutional network, the temporal graph convolutional network is first used for processing, and then the spatial graph convolutional network is used for processing.

6. The method according to claim 1, characterized in that The method further comprises: After all the interest points in the interest point spatial domain vector have been matched, if there are still interest points that have stayed for a long time and have not been matched in the interest point temporal domain vector, the positions, arrival times and stay times of these interest points are extracted; Determine the type of the point of interest based on its location; The location, arrival time, stay time and type of the extracted points of interest are filled into the user behavior habit space-time feature map.

7. The method according to claim 1, characterized in that The multi-layer spatiotemporal convolutional network includes a three-layer spatiotemporal convolutional network.

8. The method according to claim 1, characterized in that The method further comprises: Send a prompt message whether to update the model through the human-machine interface; receiving a user's selection of the prompt message; When the user chooses to update the model, the multi-layer spatiotemporal convolutional network model is upgraded using the recent spatiotemporal feature graph of user behavior habits.

9. A dynamic path recommendation device, characterized in that: The device comprises: A historical data acquisition module configured to acquire the vehicle's historical driving trajectory and the user's historical navigation records; A trajectory data extraction module configured to extract the time and position of trajectory data points in the historical driving trajectory of the vehicle; A time domain vector generation module, configured to generate a time domain vector of a point of interest according to the time and position of the trajectory data point; A navigation data extraction module configured to extract the destination type, destination location and arrival time of each record in the user's historical navigation record; an airspace vector generating module configured to generate an airspace vector of a point of interest according to the destination type, destination location and arrival time extracted from the user's historical navigation record; A matching module, configured to match each point of interest in the spatial domain vector of the point of interest with a point in the same or adjacent position in the temporal domain vector of the point of interest, to obtain a length of time spent at each point of interest; A space-time feature map generation module is configured to generate a space-time feature map of user behavior habits according to the length of time spent at each point of interest, the location of the point of interest, and the type of the point of interest; A format conversion module configured to convert the user behavior habit space-time feature graph into a graph data format; A spatiotemporal convolution module, configured to sequentially perform multiple spatiotemporal convolution processes on the user behavior habit spatiotemporal feature graph in graph data format through a multi-layer spatiotemporal convolution network, wherein each layer of the multi-layer spatiotemporal convolution network includes a temporal graph convolution network and a spatial graph convolution network, and wherein the features obtained by the previous layer of spatiotemporal convolution network processing are normalized and then input into the next layer of spatiotemporal convolution network; A heat map acquisition module, configured to perform normalization and upsampling on the features obtained through the multi-layer spatiotemporal convolutional network to obtain a user behavior habit heat map; The path determination module is configured to determine a recommended path based on the current location of the vehicle, the current real-time road conditions and the user behavior habit heat map.

10. The device according to claim 9, characterized in that The device also includes an attention mechanism module, which is configured to: after generating a user behavior habit space-time feature map, assign a weight to each interest point through an attention mechanism according to the priority of each interest point; and update the user behavior habit space-time feature map according to the weight assigned to each interest point.

11. The device according to claim 10, characterized in that The device also includes a weight updating module configured to: receive a user's operation on the recommended path; and update the weight of the point of interest as the destination of the recommended path according to the user's operation.

12. The device according to claim 9, characterized in that The spatiotemporal convolution module is further configured to: input the unnormalized features output by the first spatiotemporal convolutional network layer in the multi-layer spatiotemporal convolutional network into each subsequent spatiotemporal convolutional network layer.

13. A vehicle computer system, characterized in that: The vehicle system includes a memory and a processor, the memory stores a computer program, and when the computer program is executed by the processor, the dynamic path recommendation method according to any one of claims 1-8 is implemented.

14. A storage medium storing computer-readable instructions, wherein when the instructions are executed by a processor, the dynamic path recommendation method according to any one of claims 1 to 8 is executed.

Images