Destination prediction model training method, destination prediction method, destination prediction device and vehicle

Abstract

The invention discloses a destination prediction model training method and device, a destination prediction method and device and a vehicle, and relates to the technical field of machine learning. The method comprises the steps that a training sample set is acquired, the training sample set comprises a plurality of training samples and destination labels corresponding to the samples, and each training sample comprises position information of a departure place of travel data of a regular travel vehicle, a driving energy residual state of the vehicle during departure and time information during departure; the travel rule vehicle is determined based on daily travel times and daily travel accumulated time consumption of the vehicle; and training the initial model of the destination prediction model based on the training sample set to obtain the destination prediction model. According to the method, the initial model of the destination prediction model can be trained based on the historical travel data of the vehicle, the destination prediction model capable of identifying the travel rule of the vehicle is obtained, and the accuracy of predicting the travel destination of the vehicle can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of machine learning, and particularly to a method for training a destination prediction model, a destination prediction method, an apparatus, and a vehicle. Background Art

[0002] In a conventional navigation scenario, the vehicle owner needs to input a destination in the navigation application, and then the navigation application will plan a route according to the destination input by the vehicle owner. However, during driving, if the vehicle owner needs to change the destination, they also need to stop at a safe place and then input a new destination in the navigation application.

[0003] With the development of machine learning technology, the intelligence level of vehicles has also increased accordingly. In this trend, if the vehicle can actively predict the destination that the vehicle owner may go to and push it, and let the vehicle owner select the pushed destination, the convenience of navigation can be improved.

[0004] Therefore, how to use a machine learning model to accurately predict the destination that the vehicle owner may go to has become a necessity for us. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a method for training a destination prediction model, a destination prediction method, an apparatus, and a vehicle to improve the accuracy of predicting the destination of a vehicle trip.

[0006] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0007] In a first aspect, the present application provides a method for training a destination prediction model, the method comprising:

[0008] Obtain a training sample set, the training sample set includes a plurality of training samples and a destination label corresponding to each sample. Each training sample includes the location information of the departure place of the travel data of a vehicle with travel rules, the remaining state of the driving energy of the vehicle at the time of departure, and the time information at the time of departure. The vehicle with travel rules is determined based on the daily travel times and the cumulative travel time of the vehicle per day. Train an initial model of the destination prediction model based on the training sample set to obtain a destination prediction model.

[0009] According to the above technical solution, the initial model of the destination prediction model constructs a training sample set through the trip data of regularly traveling vehicles, which can reduce the noise interference of the trip data of irregular vehicles, enabling the initial model to better learn the general laws of vehicle trips and master the associations between the vehicle position information, the remaining state of the driving energy, and the time information in the training samples and the labels of the destinations in the training samples. This enables the model to, in practical applications, generate a relatively accurate predicted destination based on the current position information of the vehicle, the remaining state of the driving energy of the vehicle at the time of departure, and the time information at the time of departure, combined with the learned laws.

[0010] Further, obtaining the training sample set includes: obtaining multiple pieces of original trip data within a preset area. Determining valid trip data that meets the valid screening conditions from the multiple pieces of original trip data. Determining the trip data of regularly traveling vehicles from the valid trip data based on the daily trip frequency and the cumulative trip duration of the vehicles corresponding to the valid trip data. Extracting training samples from the trip data of regularly traveling vehicles to obtain the training sample set.

[0011] According to the above technical solution, by screening the original trip data according to the valid screening conditions to obtain valid trip data, it can ensure the amount of information of each training sample in the training sample set and reduce the interference of low-information trip data on model training. Further, by screening out regularly traveling vehicles based on the daily trip frequency and the cumulative trip duration of the vehicles corresponding to the valid trip data, the model can focus on the training samples with stable trips, enhance the model's understanding of common trip scenarios, and improve the efficiency and accuracy of model training.

[0012] Further, the valid screening conditions include at least one of the following: the duration of the trip data exceeds a preset duration threshold. The mileage of the trip data exceeds a preset mileage threshold. The distance between the starting point and the ending point of the trip data exceeds a preset distance threshold.

[0013] According to the above technical solution, by considering the validity of the duration, mileage, and the distance between the starting point and the ending point of each trip in the original trip data, it is possible to consider the amount of information of valid trips from different dimensions.

[0014] Further, determining regularly traveling vehicles from the vehicles corresponding to the valid trip data based on the daily trip frequency and the cumulative trip duration of the vehicles corresponding to the valid trip data includes: determining irregularly traveling vehicles, low-regularity traveling vehicles, and professional operation vehicles from the vehicles corresponding to the valid trip data based on the daily trip frequency and the cumulative trip duration of the vehicles corresponding to the valid trip data. Determining the vehicles other than the irregularly traveling vehicles, low-regularity traveling vehicles, and professional operation vehicles among the vehicles corresponding to the valid trip data as regularly traveling vehicles.

[0015] Further, the vehicles with irregular travel patterns are vehicles that meet at least one of the following: The vehicles with irregular travel patterns are vehicles that meet at least one of the following:

[0016] The standard deviation of the daily travel frequency of a vehicle is greater than a first reference standard deviation, where the first reference standard deviation is equal to the product of a first standard deviation and a first coefficient plus a first average value. The first standard deviation is obtained by calculating the standard deviation of the daily travel frequency for each vehicle in a preset area and then calculating the standard deviation of the standard deviations of the daily travel frequencies of all vehicles in the preset area. The first average value is the average of the standard deviations of the daily travel frequencies of all vehicles in the preset area.

[0017] The standard deviation of the cumulative travel time of a vehicle per day is greater than a second reference standard deviation, where the second reference standard deviation is based on the product of a second standard deviation and a first coefficient plus a second average value. The second standard deviation is obtained by calculating the standard deviation of the cumulative travel time per day for each vehicle in the preset area and then calculating the standard deviation of the standard deviations of the cumulative travel times per day of all vehicles in the preset area. The second average value is the average of the standard deviations of the cumulative travel times per day of all vehicles in the preset area.

[0018] Low-regularity travel vehicles are vehicles that meet at least one of the following:

[0019] The average value of the daily travel frequency of a vehicle is less than a first reference average value, where the first reference average value is equal to a third average value minus the product of a third standard deviation and a first coefficient. The third average value is the average of the daily travel frequencies of all vehicles in the preset area, and the third standard deviation is the standard deviation of the daily travel frequencies of all vehicles in the preset area.

[0020] The travel frequency of a vehicle within a preset recent duration exceeds a frequency threshold, and the number of trips per day is less than a first number threshold, where the first number threshold is equal to the third average value minus the product of the third standard deviation and a second coefficient.

[0021] The average value of the cumulative travel time of a vehicle is less than a second reference average value, where the second reference average value is equal to a fourth average value minus the product of a fourth standard deviation and a first coefficient. The fourth average value is the average of the cumulative travel times per day of all vehicles in the preset area, and the fourth standard deviation is the standard deviation of the cumulative travel times per day of all vehicles in the preset area.

[0022] The travel frequency of a vehicle within a preset recent duration exceeds a frequency threshold, and the cumulative travel time per day is less than a first duration threshold, where the first duration threshold is equal to the fourth average value minus the product of the fourth standard deviation and a second coefficient.

[0023] Professional operation vehicles are vehicles that meet at least one of the following:

[0024] The average number of daily trips of the vehicle is greater than the third reference average value, and the third reference average value is equal to the product of the third standard deviation and the first coefficient plus the third average value.

[0025] The number of trips of the vehicle within a recent preset time period exceeds the trip number threshold, and the number of trips per day is greater than the second trip number threshold, where the second trip number threshold is equal to the third average value plus the product of the third standard deviation and the second coefficient.

[0026] The average value of the cumulative travel time of the vehicle is greater than the fourth reference average value, and the fourth reference average value is equal to the fourth average value plus the product of the fourth standard deviation and the first coefficient.

[0027] The number of trips of the vehicle within a recent preset time period exceeds the trip number threshold, and the cumulative travel time per day is greater than the second time threshold, where the second time threshold is equal to the fourth average value plus the product of the fourth standard deviation and the second coefficient.

[0028] According to the above solution, the method can scientifically quantify the vehicle travel pattern, effectively distinguish irregular travelers, low-regular travelers, and professional operation vehicles, and improve the classification accuracy through multi-dimensional statistical indicators (such as standard deviation, average value, etc.) and preset thresholds. Moreover, through the preset thresholds and different coefficients, it can combine the overall data characteristics of the region and enhance the adaptability of calculating regular travel vehicles.

[0029] Furthermore, training samples are extracted from the travel data of regular travel vehicles to obtain the training sample set, including: clustering the destinations in the travel data of regular travel vehicles to obtain multiple destination clustering clusters. Determine the first clustering cluster, the second clustering cluster, and the third clustering cluster from the multiple destination clustering clusters. Among them, the first clustering cluster is the destination clustering cluster with the most destinations falling into the target departure place clustering cluster, and the target departure place clustering cluster is the departure place clustering cluster with the most departure places among the multiple departure place clustering clusters, and the multiple departure place clustering clusters are obtained by clustering the departure places of the first trips of each day in the travel data of regular travel vehicles; the second clustering cluster is the destination clustering cluster including the most target destinations, and the target destination is the destination of the previous trip in two trips with an interval time greater than the time threshold; the third clustering cluster is the destination clustering cluster including the most driving energy replenishment places. Training samples are extracted based on the travel data corresponding to the destinations in the first clustering cluster, the second clustering cluster, and the third clustering cluster.

[0030] According to the above technical solution, by clustering the trip data of regular travel vehicles, multiple clustering clusters are obtained, and each clustering cluster represents a similar type of destination, thereby reducing the data complexity. And by dividing the first clustering cluster, the second clustering cluster, and the third clustering cluster, the daily essential destinations of regular travel vehicles can be identified, and based on the daily essential destinations, training samples are extracted from the trips, which can improve the quality of the model training data, and further improve the accuracy of the destination prediction model.

[0031] Further, the method further includes: setting the destination label of the training samples extracted based on the first clustering cluster as home. Setting the destination label of the training samples extracted based on the second clustering cluster as company. Setting the destination label of the training samples extracted based on the third clustering cluster as the frequently visited driving energy replenishment place.

[0032] According to the above technical solution, setting the labels of the training samples corresponding to the trip data in the first clustering cluster, the second clustering cluster, and the third clustering cluster as home, company, and energy replenishment place can extract the daily essential destinations of the vehicle from multiple clustering clusters. The daily essential destinations can represent the destinations with a relatively high probability when regular travel vehicles travel. By enabling the destination prediction model to learn the training samples corresponding to these labels, the training efficiency of the destination prediction model can be improved.

[0033] Further, the method further includes: selecting the top M target clustering clusters with the most destinations from the destination clustering clusters other than the first clustering cluster, the second clustering cluster, and the third clustering cluster, where M is a positive integer. Extracting training samples based on the trip data corresponding to the destinations in the top M target clustering clusters.

[0034] According to the above technical solution, in addition to the daily essential trips, regular travel vehicles may also have other high-frequency destinations, such as schools, libraries, gyms, etc. In this case, other destination clustering clusters other than the first clustering cluster, the second clustering cluster, and the third clustering cluster can be considered to provide more abundant training data for the destination prediction model.

[0035] In a second aspect, the present application provides a destination prediction method, which includes: obtaining the current location information of the vehicle, the remaining state of the driving energy of the vehicle, and the current time information. Predicting the trip destination based on the location information, the remaining state of the driving energy, the time information, and the destination prediction model. Wherein, the destination prediction model is trained based on a training sample set, the training sample set includes multiple training samples and the destination label corresponding to each sample, and each training sample includes the location information of the departure place of the trip data of the regular travel vehicle, the remaining state of the driving energy of the vehicle at the time of departure, and the time information at the time of departure.

[0036] According to the above technical solution, by inputting the vehicle's current location information, the vehicle's current remaining driving energy status and the current time information into a trained destination prediction model, the destination prediction model that has learned the vehicle's travel patterns can output a predicted destination that is more in line with the vehicle's current information, thereby improving the speed and accuracy of predicting the destination.

[0037] Furthermore, the destination prediction model includes a Bayesian probability model, and predicts the trip destination based on the location information, the remaining state of the driving energy, the time information, and the destination prediction model, including: predicting the travel probability of each of multiple candidate destinations based on the location information, the remaining state of the driving energy, the time information, and the Bayesian probability model. The first N candidate destinations with the highest travel probability are determined as the trip destination; N is a positive integer.

[0038] According to the above technical solution, multiple candidate destinations with the highest travel probability are obtained through the destination prediction model and pushed to the car owner. Based on the current information of the vehicle, multiple candidate destinations that meet the current actual itinerary can be given, which can enrich the car owner's travel options and effectively respond to complex and changing travel needs. When the car owner faces various emergencies or suddenly decides to change the itinerary, these candidate destinations can provide him with a flexible decision-making basis.

[0039] Furthermore, the obtaining of the vehicle's current position information, the vehicle's current remaining driving energy status, and the current time information includes: when it is detected that the entire vehicle is powered on, obtaining the vehicle's current position information, the vehicle's current remaining driving energy status, and the current time information.

[0040] According to the above technical solution, by detecting the power-on status of the vehicle, the prediction timing of the destination prediction method is determined. The full power-on of the vehicle means that the owner's journey is about to begin. Obtaining relevant information at this time can ensure the real-time and effectiveness of the data, allowing the destination prediction model to perform calculations based on the latest conditions, thereby improving the accuracy of the destination prediction.

[0041] In a third aspect, the present application provides a destination prediction model training device, which includes various functional modules used in the method described in the first aspect above.

[0042] In a fourth aspect, the present application provides a destination prediction device, which includes various functional modules used in the method described in the second aspect above.

[0043] In a fifth aspect, the present application provides a vehicle, the vehicle comprising: a processor and a memory; the memory stores instructions executable by the processor. When the processor is configured to execute the instructions, the electronic device implements the method described in the first aspect above. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the accompanying drawings required for the description of the specific embodiments or the prior art. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can also be obtained based on these drawings.

[0045] Figure 1 Flow diagram of a destination prediction method provided by an embodiment of this application;

[0046] Figure 2 Flow diagram of a method for generating multiple candidate destinations provided by an embodiment of this application;

[0047] Figure 3 Flow diagram of a method for training a destination prediction model provided by an embodiment of this application;

[0048] Figure 4 Flow diagram of a method for obtaining a training sample set for a destination prediction model provided by an embodiment of this application;

[0049] Figure 5 Flow diagram of a method for extracting training samples based on clustering clusters provided by an embodiment of this application;

[0050] Figure 6 Flow diagram of another method for extracting training samples based on clustering clusters provided by an embodiment of this application;

[0051] Figure 7 Flow diagram of a clustering method provided by an embodiment of this application;

[0052] Figure 8 Flow diagram of a method for merging clustering clusters provided by an embodiment of this application;

[0053] Figure 9 Block diagram of a destination prediction model training device provided according to an exemplary embodiment;

[0054] Figure 10 Block diagram of a destination prediction device provided according to an exemplary embodiment;

[0055] Figure 11 Block diagram of a vehicle provided according to an exemplary embodiment. Specific embodiments

[0056] The embodiments of the present invention will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and 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 invention. It should be understood that the preferred embodiments are only for illustrating the present invention, rather than for limiting the protection scope of the present invention.

[0057] It should be noted that in the embodiments of the present application, words such as "exemplarily" or "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as "exemplarily" or "for example" in the embodiments of the present application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, the use of words such as "exemplarily" or "for example" is intended to present relevant concepts in a specific manner.

[0058] In order to clearly describe the technical solutions of the embodiments of the present application, in the embodiments of the present application, words such as "first" and "second" are used to distinguish the same items or similar items with basically the same functions and roles. Those skilled in the art can understand that words such as "first" and "second" are not used to limit the quantity and execution order.

[0059] In a conventional navigation scenario, when the user is in a driving state and has a navigation need or needs to change the destination, a new destination needs to be input in the navigation application. At this time, if the user directly performs an input operation during driving, their attention is likely to be distracted, thereby triggering a driving risk and endangering driving safety. If the user parks the vehicle in a safe position before performing the input operation, although safety can be ensured, it will cause a waste of time, disrupt the travel rhythm, and affect the overall travel efficiency.

[0060] Currently, with the development of machine learning technology and the improvement of vehicle intelligence, if the vehicle can predict the destination that the user may go to, the user's input operation can be reduced, and the navigation convenience and the user experience can be improved.

[0061] Therefore, how to accurately predict the destination that the user may go to during the user's navigation process has become an urgent problem to be solved currently.

[0062] In view of this, the present application provides a method for training a destination prediction model and a destination prediction method. First, the method for training a destination prediction model can, based on a machine learning model, learn the historical formation data of a vehicle to master the potential rules between the vehicle's position information, the remaining state of the driving energy, and the time information and the destination. Then, the destination prediction method can generate a relatively accurate destination prediction result by obtaining the current information of the predicted vehicle and inputting it into the trained destination prediction model.

[0063] In some embodiments, the execution subject of the destination prediction method provided by the embodiments of the present application is a destination prediction device. The destination prediction device can be an in-vehicle computing device or a vehicle including an in-vehicle computing device; or, the destination prediction device can also be a processor (such as a central processing unit) in the in-vehicle computing device or the vehicle, or the in-vehicle computing device can also be a functional module or functional unit in the in-vehicle computing device or the vehicle for executing the destination prediction method. The embodiments of the present application do not limit the specific form of the destination prediction device.

[0064] The following specifically introduces the destination prediction method provided by the embodiments of the present application with reference to the accompanying drawings.

[0065] The destination prediction method provided by the embodiments of the present application, the destination prediction device, this method is as Figure 1 shown, and specifically includes the following steps:

[0066] S101. Obtain the current position information of the vehicle, the remaining state of the vehicle's current driving energy, and the current time information.

[0067] Specifically, the current position information of the vehicle can also be understood as the departure location of the vehicle. The current time information of the vehicle includes date information and time information. Among them, the date information at least includes: whether the day of the trip is a working day, and whether the next day of the trip is a working day. The time information at least includes: the departure (start) time period of the trip, and the current time period of the trip.

[0068] In some embodiments, the driving energy can include at least one of the following: electric quantity, fuel quantity, compressed natural gas, liquefied natural gas, hydrogen, bioethanol, biodiesel. The embodiments of the present application do not limit the specific form of the driving energy.

[0069] It should be noted that for the current location information of the vehicle, the GPS positioning of the vehicle can be obtained through the destination prediction device of the vehicle, or the longitude and latitude position of the vehicle can also be obtained through the base station signal of the mobile communication network passed by the vehicle. For the remaining state of the driving energy of the vehicle currently, the remaining state of the driving energy can be obtained through the built-in energy monitoring device of the vehicle. For example, a fuel vehicle can detect the remaining amount of fuel in the fuel tank through a fuel level sensor. For the current time information of the vehicle, the current time information can be obtained through the on-vehicle clock, or the current time information can also be obtained through a computing device (such as the owner's mobile phone) connected to the vehicle.

[0070] In some embodiments, the destination prediction device can, when detecting that the vehicle is fully powered on, obtain the current location information of the vehicle, the remaining state of the current driving energy of the vehicle, and the current time information, and predict the destination of the vehicle.

[0071] A possible implementation is that the power management module of the vehicle is responsible for monitoring the current and voltage states of the main power line. When the vehicle starts, the key is turned or the start button is pressed, and the power management module detects that the voltage of the main power line jumps from the sleep state to the rated working voltage, and the current changes significantly. This signal change is immediately sent to the in-vehicle electronic control unit (ECU). After receiving the signal, the ECU determines that the vehicle has completed the power-on operation according to the preset program, and then sends a trigger instruction to the destination prediction device. After receiving the instruction, the destination prediction device executes the destination prediction method and starts to obtain information such as the current location of the vehicle, the remaining state of the driving energy, and the time.

[0072] It can be seen that by detecting the power-on situation of the vehicle to judge the prediction timing of the destination prediction method, the vehicle being fully powered on means that the owner's journey is about to start. Obtaining relevant information at this time can ensure the real-time and effectiveness of the data, enabling the destination prediction model to operate based on the latest situation, thereby improving the accuracy of destination prediction.

[0073] In some other embodiments, the destination prediction device can also, when detecting that the vehicle is not driving according to the navigation route, obtain the current location information of the vehicle, the remaining state of the current driving energy of the vehicle, and the current time information, and predict the destination of the vehicle.

[0074] It can be seen that the vehicle not driving according to the navigation route may be due to complex road conditions of the current navigation route, or there are sudden accidents, temporary road controls, etc., resulting in the owner temporarily changing the travel plan. At this time, the owner may need to enter the destination again in the navigation application according to the new route. Therefore, in this case, the destination prediction device can re-execute the destination prediction method.

[0075] S102. Predict the trip destination based on the location information, remaining state of the driving energy, time information, and destination prediction model.

[0076] Among them, the destination prediction model is trained based on a training sample set. The training sample set includes multiple training samples and the destination labels corresponding to each sample. Each training sample includes the location information of the departure place of the trip data of the travel-regular vehicle, the remaining state of the driving energy of the vehicle at the departure time, and the time information at the departure time.

[0077] Specifically, for the training method of the destination prediction model, reference can be made to the following text Figure 3 , which will not be elaborated here.

[0078] It should be noted that the location information can clarify the location of the vehicle. Based on the historical travel trajectory of the vehicle owner, the range of the predicted destination can be determined. The remaining state of the driving energy can reflect the cruising range of the vehicle. Combining with the current location information of the vehicle can further narrow the range of the predicted destination and improve the prediction efficiency. The time information includes the current time period and whether it is a working day, which can, to a certain extent, limit the type of the destination (for example, when it is a working day, the possible destination is an office, and when it is a non-working day, the possible destination is a leisure place). By combining these three types of information, the destination prediction model can fully understand the current information of the vehicle and can perform destination prediction well.

[0079] In some embodiments, before inputting the above information into the destination prediction model, the destination prediction device can preprocess the above information, such as operations like classification encoding and data binning, perform feature processing on the information, and obtain the feature combination corresponding to the above information to highlight the key features in the information.

[0080] A possible implementation manner, the preprocessing operations on the above information can be as shown in Table 1.

[0081] Table 1

[0082]

[0083]

[0084] It can be seen from the above steps S101 - S102 that by inputting the current location information of the vehicle, the remaining state of the current driving energy of the vehicle, and the current time information into the trained destination prediction model, the destination prediction model that has learned the vehicle travel pattern can be used to output a predicted destination that is more in line with the current information of the vehicle, which can improve the speed and accuracy of predicting the destination.

[0085] In some embodiments, the destination prediction model used in step S102 includes a Bayesian probability model, which can predict multiple trip destinations based on the current vehicle information. In this case, as Figure 2 shown, S102 specifically includes the following steps:

[0086] S201. Predict the travel probability of each of multiple candidate destinations based on the location information, the remaining state of the driving energy, the time information, and the Bayesian probability model.

[0087] Specifically, the Bayesian probability model is a machine learning method based on Bayes' theorem, and its core idea is to use prior knowledge (historical experience or assumptions) combined with new data to update probability predictions.

[0088] Exemplarily, the Bayesian probability model can specifically be a Naive Bayes model or a Bayesian network model.

[0089] S202. Determine the top N candidate destinations with the highest travel probabilities as the trip destinations.

[0090] Wherein, N is a positive integer.

[0091] Specifically, the prediction process can be expressed as:

[0092]

[0093] Wherein, (Y = y i |f) represents the probability of destination y for the feature combination f i of, p(Y = y i ) represents the prior probability of the i-th destination, and p(f|Y = y i )×p(Y = y i ) represents the probability of the feature combination f appearing under the condition of the i-th destination. ∑ i p(f|Y = y i )×p(Y = y i ) represents the total probability for all possible destinations corresponding to the feature combination f.

[0094] In some embodiments, the specific value of N can be 3, 5, or 7, and the embodiments of the present application do not limit the specific value of N.

[0095] For example, assume there are two destinations y1 and y2, and a feature combination f (for example, the training sample corresponding to this feature combination is: the departure location in the location information of the trip is home, the remaining state of the vehicle driving energy is 20%, the time information of the trip is that the current day is a non-working day, the next day is a working day, and the time period corresponding to the trip time is from 8:00 to 12:00). The prior probabilities of the two destinations are 0.4 (i.e., p(Y = y1)) and 0.6 (i.e., p(Y = y2)) respectively. Among the multiple feature combinations corresponding to destination y1, the probability of feature combination f appearing is 0.7 (i.e., p(f|Y = y1) × p(Y = y1)). Among the multiple feature combinations corresponding to destination y2, the probability of feature combination f appearing is 0.5 (i.e., p(f|Y = y2) × p(Y = y2)). Substituting this data into formula (1), the denominator in the formula can be obtained as 0.58 (i.e., ∑ i p(f|Y = y i ) × p(Y = y i ) = p(f∣Y = y1) × p(Y = y1) + p(f∣Y = y2) × p(Y = y2)), and the travel probability (posterior probability) for destination y1 can be obtained as The travel probability for Then, destinations y1 and y2 are sorted in descending order of travel probability to construct trip destinations, and the trip destinations are pushed to the user for selection.

[0096] For another example, through the above prediction process, the travel probabilities of 7 destinations are obtained as [0.12, 0.08, 0.25, 0.05, 0.18, 0.2, 0.12]. After sorting these probability values from largest to smallest, the sorting result is [0.25, 0.2, 0.18, 0.12, 0.12, 0.08, 0.05]. Thus, the top 3 destinations with the highest travel probabilities, that is, the destinations corresponding to the probabilities of 0.25, 0.2, and 0.18, are selected as trip destinations.

[0097] As can be seen from steps S201 - S202, by using the destination prediction model to obtain multiple candidate destinations with the highest travel probabilities and pushing them to the vehicle owner, it is possible to give multiple candidate destinations that meet the current actual trip based on the current information of the vehicle, which can enrich the vehicle owner's travel choices and effectively respond to complex and changeable travel demands. When the vehicle owner faces various emergencies or changes the trip on a whim, these candidate destinations can provide a flexible decision-making basis for him.

[0098] In some embodiments, before step S102, the destination prediction device may also obtain the destination prediction model.

[0099] In a possible implementation, after the destination prediction model is pre-trained by other computing devices (collectively referred to as the destination prediction model training device hereinafter), it can be deployed in the destination prediction device.

[0100] The process of the destination prediction model training device training the destination prediction model is introduced below.

[0101] Figure 3 As shown in the flowchart of the destination prediction model training method provided in the embodiments of the present application, Figure 3 the prediction model training method includes the following steps:

[0102] S301. Obtain a training sample set.

[0103] Among them, the training sample set includes multiple training samples and the destination label corresponding to each sample. Each training sample includes the location information of the departure place of the trip data of the regularly traveling vehicle, the remaining state of the driving energy of the vehicle at the time of departure, and the time information at the time of departure. The regularly traveling vehicle is determined based on the daily travel times of the vehicle and the cumulative travel time per day.

[0104] It should be noted that the destination prediction model training device can obtain the trip data of multiple vehicles from other computing devices to construct a training sample set, so as to reduce the problem of model overfitting caused by a single training sample.

[0105] S302. Train the initial model of the destination prediction model based on the training sample set to obtain the destination prediction model.

[0106] In some embodiments, the destination prediction model training device can perform feature processing on each training sample in the training sample set before training (specifically, refer to Table 1 in step S102 above), obtain the feature combination corresponding to the training sample, and assign corresponding weights (influence factors) to the features in the feature combination, so that the initial model can learn the key features affecting destination prediction during the training process.

[0107] Specifically, for the weight assignment of the feature factors, different weights can be assigned to the same feature according to whether the destination label is a frequently visited place. The specific weight assignment can refer to Table 2:

[0108] Table 2

[0109] It should be noted that the weight assignment to the feature combination in Table 2 is not fixed, but can be adjusted according to the actual situation, and new features (such as the departure place location, whether the next day is a working day, etc.) can be added for consideration. The embodiments of the present application do not limit the assignment of the weight factors.

[0110] In a possible implementation manner, the process of the destination prediction model training device training the initial model of the destination prediction model based on the training sample set is as follows: First, perform feature extraction based on the location information, remaining driving energy state, and time information of each training sample to obtain the feature combination of each training sample. Then, count the occurrence frequency of each destination in the training sample set as the prior probability. Calculate the occurrence frequency of each destination corresponding to different feature combinations as the conditional probability (also called joint probability) of the corresponding destination. Finally, combine the prior probability and the conditional probability to calculate the posterior probability. Through the posterior probability, the model can indicate that when receiving the feature combination, it gives the travel destination with a relatively large posterior probability corresponding to this type of feature combination.

[0111] As can be seen from the above steps S301 - S302, the initial model of the destination prediction model constructs a training sample set through the travel data of regular travel vehicles, which can reduce the noise interference of irregular vehicle travel data, enabling the initial model to better learn the general laws of vehicle travel and master the association between the vehicle location information, remaining driving energy state, and time information in the training sample and the label of the destination of the training sample. This enables the model to, in practical applications, generate a relatively accurate predicted destination based on the vehicle's current location information, remaining driving energy state at the time of departure, and time information at the time of departure, combined with the learned laws.

[0112] In some embodiments, for the specific process of obtaining the training sample set in step S301, as Figure 4 shown, S301 specifically includes the following steps:

[0113] S401. Obtain multiple pieces of original travel data within a preset area.

[0114] Among them, the preset area can be an administrative region, an urban area, or a provincial area, etc.

[0115] Specifically, the process of a vehicle from one start state to one shutdown state can be regarded as one piece of original travel data.

[0116] S402. Determine the valid travel data that meets the valid screening conditions from the multiple pieces of original travel data.

[0117] It should be noted that in the original travel data, there may be trips with relatively short travel times or travel distances. Such trips may lack sufficient location changes or time spans, resulting in less information contained in them and being difficult to use for subsequent analysis. In this case, valid screening conditions can be set to eliminate such trips.

[0118] In some embodiments, the effective screening conditions include at least one of the following: The time taken for the trip data exceeds a preset duration threshold. The mileage of the trip data exceeds a preset mileage threshold. The distance between the start point and the end point of the trip data exceeds a preset distance threshold.

[0119] It should also be noted that in the effective screening conditions, "exceeds" in "exceeds the preset duration threshold", "exceeds the preset mileage threshold", and "exceeds the preset distance threshold" can be understood as greater than the threshold or greater than or equal to the threshold. The embodiments of the present application do not limit the specific numerical relationship of "exceeds".

[0120] For example, the preset duration threshold can be set to 2 minutes, 5 minutes, or 10 minutes. The preset mileage threshold is set to 700 meters, 1000 meters, or 1300 meters. The distance threshold can be set to 300 meters, 500 meters, or 800 meters. The embodiments of the present application do not limit the specific values of the preset duration threshold, the preset mileage threshold, and the preset distance threshold.

[0121] In a possible implementation, the destination prediction device may record the original trip data that meets any one of the above screening conditions as valid trip data.

[0122] For example, taking meeting any one of the effective screening conditions as the implementation method, assuming that the preset duration threshold is 5 minutes, the preset mileage threshold is 1000 meters, and the preset distance threshold is 500 meters. The time taken for a piece of original trip data is 3 minutes, the mileage is 800 meters, and the distance between the start point and the end point is 800 meters. According to the judgment, the time taken for this original trip is 3 minutes < 5 minutes (not exceeding the preset duration threshold), the mileage is 800 meters < 1000 meters (not exceeding the preset mileage threshold), and the distance between the start point and the end point is 800 meters > 500 meters (exceeding the preset distance threshold). Since the distance between the start point and the end point of this original trip exceeds the preset distance threshold, it can be regarded as valid trip data.

[0123] In another possible implementation, the destination prediction device may record the original trip data that meets multiple of the above screening conditions as valid trip data. Among them, the number of multiple screening conditions can be two or three.

[0124] For example, taking the fulfillment of at least two valid screening conditions as the implementation method, assume that the preset duration threshold is 5 minutes, the preset mileage threshold is 1000 meters, and the preset distance threshold is 500 meters. The travel time of an original trip data is 3 minutes, the travel mileage is 800 meters, and the distance between the starting point and the ending point is 800 meters. According to the judgment, the travel time of this original trip is 3 minutes < 5 minutes (not exceeding the preset duration threshold), the inspected mileage is 800 meters < 1000 meters (not exceeding the preset mileage threshold), and the inspected distance between the starting point and the ending point is 800 meters > 500 meters (exceeding the preset distance threshold). Since this original trip only meets the condition that the distance between the starting point and the ending point exceeds the preset distance threshold and does not meet other valid screening conditions, it is regarded as not meeting the valid screening conditions.

[0125] S403. Determine the trip data of the vehicles with regular travel patterns from the valid trip data based on the daily travel times and the cumulative daily travel durations of the vehicles corresponding to the valid trip data.

[0126] In some embodiments, statistical metrics of the daily travel times and the cumulative daily travel durations of the vehicles corresponding to the valid trip data can be calculated, and the vehicle types can be determined by preset thresholds to screen out the vehicles with regular travel patterns and obtain the trip data of the vehicles with regular travel patterns.

[0127] A possible implementation method is to classify the vehicle types into vehicles with irregular travel patterns, vehicles with low regular travel patterns, professional operation vehicles, and vehicles with regular travel patterns by calculating the statistical metrics of the daily travel times and the cumulative daily travel durations of the vehicles corresponding to the valid trip data. Among the vehicles within the preset area, identify and eliminate the above-mentioned vehicles with irregular travel patterns, vehicles with low regular travel patterns, and professional operation vehicles, and regard the remaining vehicles as vehicles with regular travel patterns. Specifically, this process can refer to S403a - S403b below and will not be elaborated here.

[0128] S404. Extract training samples based on the trip data of the vehicles with regular travel patterns to obtain a training sample set.

[0129] In some embodiments, although the trips of some vehicles with regular travel patterns have regularity, there may also be cases where the trip data is sparse and does not have sufficient information. The training samples extracted from such data may cause overfitting or underfitting during the training process of the model. In this case, according to the preset data qualification conditions, the qualification of the trip data of each vehicle with regular travel patterns can be screened to obtain the screened trip data for subsequent steps.

[0130] A possible implementation method is that the preset data qualification conditions may include at least one of the following: the number of trip records in the trip data of the vehicle with regular travel patterns exceeds the trip quantity threshold, and the time span of the trip data of the vehicle with regular travel patterns exceeds the trip time span threshold.

[0131] Exemplarily, the above-mentioned trip quantity threshold may be 150 trips, 200 trips, or 300 trips, and the above-mentioned trip time span threshold may be 2 months, 3 months, or 5 months. The embodiments of the present application do not limit the specific values of the trip quantity threshold and the time span threshold.

[0132] In some embodiments, because the trip data of vehicles with travel patterns has high regularity and reliability, clustering processing can be performed on the trip data of vehicles with travel patterns to obtain clustering clusters, and common locations can be screened out from the clustering clusters. And training samples are obtained based on the trip data of the common locations. Specifically, this process can refer to the following text Figure 5 , which will not be elaborated here.

[0133] It can be seen from the above steps S401 - S404 that by screening the original trip data according to the effective screening conditions to obtain effective trip data, the amount of information of each training sample in the training sample set can be ensured, and the interference of low-information trip data on model training can be reduced. Further, based on the daily trip times and the daily cumulative trip time of the vehicles corresponding to the effective trip data, vehicles with travel patterns are screened out, enabling the model to focus on the training samples with stable travel, enhancing the model's understanding of common trip scenarios, and improving the efficiency and accuracy of model training.

[0134] In some embodiments, for step S403, the specific method for obtaining vehicles with travel patterns based on the vehicles corresponding to the effective data may include the following steps S403a - S403b:

[0135] S403a. Based on the daily trip times and the daily cumulative trip time of the vehicles corresponding to the effective trip data, determine irregular travel vehicles, low-regular travel vehicles, and professional operation vehicles from the vehicles corresponding to the effective trip data.

[0136] S403b. Determine that the vehicles other than the irregular travel vehicles, low-regular travel vehicles, and professional operation vehicles among the vehicles corresponding to the effective trip data are vehicles with travel patterns.

[0137] A possible implementation manner is that the irregular travel vehicles are vehicles that meet at least one of the following:

[0138] The standard deviation of the daily trip times of the vehicle is greater than the first reference standard deviation, and the first reference standard deviation is equal to the product of the first standard deviation and the first coefficient plus the first average value. The first standard deviation is obtained by calculating the standard deviation of the daily trip times of each vehicle in the preset area and then calculating the standard deviation of the daily trip times of all vehicles in the preset area again. The first average value is the average value of the daily trip times of all vehicles in the preset area.

[0139] The standard deviation of the cumulative daily travel time of the vehicle is greater than the second reference standard deviation, which is based on the product of the second standard deviation and the first coefficient plus the second average value. The second standard deviation is obtained by calculating the standard deviation of the cumulative daily travel time standard deviations of all vehicles in the preset area after calculating the cumulative daily travel time standard deviation for each vehicle in the preset area. The second average value is the average value of the cumulative daily travel time standard deviations of all vehicles in the preset area.

[0140] In a possible implementation, the vehicles with low travel regularity are vehicles that meet at least one of the following:

[0141] The average value of the daily travel times of the vehicle is less than the first reference average value, which is equal to the third average value minus the product of the third standard deviation and the first coefficient. The third average value is the average value of the daily travel times of all vehicles in the preset area, and the third standard deviation is the standard deviation of the daily travel times of all vehicles in the preset area.

[0142] It should be noted that the third average value can be understood as the average value of the travel times of each vehicle per day within a certain time range for all vehicles, that is, after accumulating the travel times of all vehicles in the preset area, the average value is obtained according to all vehicles in the preset area and the time range. For example: the total number of trips of 100 vehicles in 30 days is 21,000, then the daily travel times of each vehicle is 21,000÷100÷30 = 7 times / day. Or, the third average value can also be understood as the average value after calculating the individual average values, that is, first calculate the daily average travel times of each vehicle in the preset area within a certain time, and then take the average value of the daily average travel times of all vehicles. For example, there are 100 vehicles in the preset area, and the total sum of the daily average times of each vehicle is 150 times, then the daily travel times of each vehicle is 4.5 times / day.

[0143] It should also be noted that one day in the third average value or the third standard deviation can also refer to a specific day. In this case, the third average value can be understood as follows: on a certain day within a preset time range (such as 30 days), the number of trips made by all vehicles in the preset area on that day is counted, and the average value of the number of trips made by all vehicles on that day is calculated. For example, if 100 vehicles made a total of 800 trips on a certain day, the third average value is 800÷100 = 8 trips / day. Correspondingly, the third standard deviation can be understood as follows: for a specific day within a preset time range (such as 30 days), first calculate the third average value of the number of trips made by all vehicles on that day according to the method described above, then count the number of trips made by each vehicle in the preset area on that day, calculate the difference between the number of trips made by each vehicle and this third average value, square these differences, then find the average value of these squared values (i.e., the variance), and finally take the square root of the variance. The result obtained is the third standard deviation. The number of trips made by a vehicle within a recent preset duration exceeds the trip threshold, and the number of trips per day is less than the first trip threshold. The first trip threshold is equal to the third average value minus the product of the third standard deviation and the second coefficient.

[0144] The average value of the cumulative travel time of a vehicle is less than the second reference average value. The second reference average value is equal to the fourth average value minus the product of the fourth standard deviation and the first coefficient. The fourth average value is the average value of the cumulative travel time of all vehicles in the preset area in one day, and the fourth standard deviation is the standard deviation of the cumulative travel time of all vehicles in the preset area in one day.

[0145] The number of trips made by a vehicle within a recent preset duration exceeds the trip threshold, and the cumulative travel time per day is less than the first duration threshold. The first duration threshold is equal to the fourth average value minus the product of the fourth standard deviation and the second coefficient.

[0146] A possible implementation method is that a professional operation vehicle is a vehicle that meets at least one of the following:

[0147] The average value of the number of trips made by a vehicle per day is greater than the third reference average value; the third reference average value is equal to the product of the third standard deviation and the first coefficient plus the third average value.

[0148] The number of trips made by a vehicle within a recent preset duration exceeds the trip threshold, and the number of trips per day is greater than the second trip threshold; the second trip threshold is equal to the third average value plus the product of the third standard deviation and the second coefficient.

[0149] The average value of the cumulative travel time of a vehicle is greater than the fourth reference average value; the fourth reference average value is equal to the fourth average value plus the product of the fourth standard deviation and the first coefficient.

[0150] The number of trips made by a vehicle within a recent preset duration exceeds the trip threshold, and the cumulative travel time per day is greater than the second duration threshold; the second duration threshold is equal to the fourth average value plus the product of the fourth standard deviation and the second coefficient.

[0151] It should be noted that the first coefficient can be 1, 2, or 3, and the second coefficient can also be 3, 4, or 5. The specific values of the first coefficient and the second coefficient can be set according to the actual situation in the preset area, and the embodiments of the present application do not limit this.

[0152] Exemplarily, the summary of the multiple averages and standard deviations in the above-mentioned multiple conditions can be referred to Table 3:

[0153] Table 3

[0154]

[0155] Exemplarily, the summary of the above-mentioned multiple conditions can be referred to Table 4:

[0156] Table 4

[0157]

[0158]

[0159] It can be understood that the core principle of this screening condition is to measure whether the travel behavior of a vehicle is stable, frequent, or inefficient through the mean and standard deviation, compare the data of a single vehicle with the data distribution of the overall vehicles in the area, and use "mean ± coefficient × standard deviation" as the dynamic preset threshold. According to the density of vehicles in the actual area, different multiples can be selected to flexibly distinguish different types of vehicles.

[0160] In some embodiments, for step S404, the process of performing clustering processing on the trip data of travel-regular vehicles to obtain clustering clusters and screening out frequently visited locations from the clustering clusters, and obtaining training samples based on the trip data of the frequently visited locations, as Figure 5 shown, S404 specifically includes the following steps:

[0161] S501. Cluster the destinations in the trip data of travel-regular vehicles to obtain multiple destination clustering clusters.

[0162] In some embodiments, the clustering algorithm for clustering the destinations in the trip data of travel-regular vehicles can be the K-means clustering algorithm, the density-based spatial clustering of applications with noise (DBSCAN) algorithm, and the hierarchical clustering algorithm.

[0163] A possible implementation manner is that the Canopy algorithm can be first used to preprocess the trip data, which can quickly filter out destinations that are obviously not in the same cluster, and then the hierarchical clustering algorithm is used to perform clustering analysis on the preprocessed results to obtain multiple destination clustering clusters. Specifically, this process can be referred to below Figure 7 , and details are not described herein again.

[0164] S502. Determine the first cluster, the second cluster, and the third cluster from multiple destination clusters.

[0165] Among them, the first cluster is the destination cluster with the most destinations falling into the target departure place cluster. The target departure place cluster is the departure place cluster with the most departure places among multiple departure place clusters. The multiple departure place clusters are obtained by clustering the departure places of the first trips of each day in the trip data of regularly traveling vehicles.

[0166] It can be seen that because the trips of the vehicle are relatively regular, regular trips mean that the vehicle owner has a fixed living track. The departure place with the largest number of first trips of each day can be regarded as the residence (such as home) of the vehicle owner. Therefore, the destination label of the training samples extracted based on the first cluster can be set to home.

[0167] The second cluster is the target place cluster including the most target destinations. The target destination is the destination of the previous trip among two trips with an interval duration greater than the duration.

[0168] It can be seen that when the interval duration between two trips is relatively long, it usually means that the vehicle owner has stayed at a certain place for a long time. The destination of the previous trip is usually the place where the vehicle owner has stayed for a long time. For vehicle owners with regular trips, the place where they stay for a long time is usually the workplace (such as a company). Therefore, the destination label of the training samples extracted based on the second cluster can be set to company.

[0169] The third cluster is the destination cluster including the most driving energy replenishment places. The vehicle drive needs to replenish energy regularly (such as refueling or charging). Therefore, the vehicle owner will frequently go to these places. Therefore, the destination label of the training samples extracted based on the third cluster can be set to frequently visited driving energy replenishment places.

[0170] It can be understood that from the above description of the label settings in the first cluster, the second cluster, and the third cluster, it can be seen that by dividing regularly traveling vehicles, the division of destinations can be made more accurate, providing a high-quality data basis for model training.

[0171] S503. Extract training samples based on the trip data corresponding to the destinations in the first cluster, the second cluster, and the third cluster.

[0172] In some embodiments, the trip data corresponding to the destinations in the first cluster, the second cluster, and the third cluster includes the daily necessary trips of regularly traveling vehicle owners, namely home, company, and energy replenishment places. These trip data reflect the fixed and high-frequency travel patterns of vehicle owners and can be used as effective samples for model training.

[0173] In some other embodiments, in addition to the first clustering cluster, the second clustering cluster, and the third clustering cluster, other clustering clusters can also be divided to obtain more destination labels. For example, frequently visited locations such as shopping malls, schools, and gyms can be identified to further enrich the diversity and coverage of the training samples, thereby enhancing the generalization ability and prediction accuracy of the model. Specifically, this process can refer to the following text Figure 6 , which will not be elaborated here.

[0174] As can be seen from the above steps S501 - S503, setting the labels of the training samples corresponding to the trip data in the first clustering cluster, the second clustering cluster, and the third clustering cluster as home, company, and energy replenishment locations can extract the daily essential destinations of the vehicle from multiple clustering clusters. The daily essential destinations can represent the travel patterns and the destinations with relatively high probabilities when the vehicle travels. By enabling the destination prediction model to learn the training samples corresponding to these several labels, the training efficiency of the destination prediction model can be improved. In some embodiments, the process of obtaining the destination clustering clusters other than the first clustering cluster, the second clustering cluster, and the third clustering cluster, such as Figure 6 shown, method S503 may further include the following steps:

[0175] S601. Select the top M target clustering clusters with the largest number of destinations from the destination clustering clusters other than the first clustering cluster, the second clustering cluster, and the third clustering cluster.

[0176] Where M is a positive integer.

[0177] Specifically, the destinations in the destination clustering clusters can be sorted from largest to smallest in terms of the number of destinations, and then the top M target clustering clusters are selected from the sorting results.

[0178] It should be noted that M represents the number of selected target clustering clusters, which can be 5, 8, or 10. The specific value of M in the embodiments of the present application is not limited.

[0179] S602. Extract training samples based on the trip data corresponding to the destinations in the top M target clustering clusters.

[0180] In some embodiments, training samples can be extracted based on the trip data corresponding to the destinations in the first clustering cluster, the second clustering cluster, and the third clustering cluster, as well as the M target clustering clusters other than the first clustering cluster, the second clustering cluster, and the third clustering cluster.

[0181] As can be seen from the above steps S601 - S602, in addition to the daily necessary trips, there may be other high - frequency destinations for the travel - pattern vehicles, such as schools, libraries, gyms, etc. In this case, other destination clusters outside the first, second, and third clusters can be considered to provide richer training data for the destination prediction model.

[0182] The following introduces the specific process of clustering the destinations in the trip data of the travel - pattern vehicles in step S501 with reference to the accompanying drawings, as Figure 7 shown, S501 specifically includes the following steps:

[0183] S701. Use the Canopy algorithm to perform preliminary clustering on the trip data to obtain multiple Canopy clusters.

[0184] Among them, each Canopy cluster contains multiple destination coordinates.

[0185] In some embodiments, after obtaining multiple Canopy clusters, the number of each Canopy cluster can also be screened to exclude the Canopy clusters with a small number.

[0186] A possible implementation method is to compare the number of destinations included in each Canopy cluster with a preset destination - number threshold, and exclude the Canopy clusters with a number less than the preset destination - number threshold.

[0187] Exemplarily, the preset destination - number threshold can be 3, 5, or 8. The specific value of the preset destination - number threshold in the embodiments of the present application is not limited.

[0188] S702. Use the single - link algorithm in the hierarchical clustering algorithm to cluster multiple Canopy clusters to obtain clustering clusters.

[0189] Specifically, the single - link algorithm determines the distance between Canopy clusters through the two closest points among the Canopy clusters.

[0190] A possible implementation method is to first calculate the distances between multiple Canopy clusters based on the single - link algorithm, then find the two closest Canopy clusters and merge them into a new Canopy cluster. Repeat the above merging operation until the number of Canopy clusters reaches the preset clustering - cluster number threshold, and regard the merged Canopy clusters as clustering clusters.

[0191] Exemplarily, the number of clustering clusters can be preset to 15, 20, or 30. The specific number of clustering clusters in the embodiments of the present application is not limited.

[0192] It should be understood that by presetting the threshold of the number of clustering clusters, the number of clustering clusters can be limited, and the computing resources and clustering accuracy can be better balanced.

[0193] In some embodiments, since the parking location of a vehicle for a destination may not be fixed. For example, when the destination is a shopping mall, the vehicle may stay in the underground parking lot, the above-ground parking lot, or other places. However, because the actual destination of the vehicle owner is the shopping mall, taking these different staying places as a clustering cluster and further as the destination label of the training sample may cause redundant calculations. In this case, the clustering clusters can be merged based on the positional relationship between the clustering clusters. The specific process is as Figure 8 shown, step S702 may further include the following steps:

[0194] S801. Obtain the central coordinates of each clustering cluster.

[0195] A possible implementation manner is to determine the central coordinates of each clustering cluster by performing weighted average processing on the coordinates of all trip destinations included in the clustering cluster.

[0196] S802. Based on the central coordinates of each clustering cluster, calculate the central distance between any two clustering clusters.

[0197] A possible implementation manner is to calculate the central distance between two clustering clusters through the Euclidean distance formula.

[0198] Exemplarily, the process of calculating the central distance between any two clustering clusters can be expressed as:

[0199]

[0200] where d represents the central distance between two clustering clusters, x1, y1 represent the central coordinates of one clustering cluster, and x2, y2 represent the central coordinates of another clustering cluster.

[0201] S803. Based on the coordinates of the easternmost, westernmost, southernmost, and northernmost destinations of each clustering cluster, calculate the maximum distance between any two clustering clusters.

[0202] Specifically, calculate the distances between the easternmost, westernmost, southernmost, and northernmost destinations of any two clustering clusters through the Euclidean distance formula, and select the farthest pair as the maximum distance between the two clustering clusters.

[0203] Exemplarily, assume that the easternmost, westernmost, southernmost, and northernmost destinations in cluster A are AE, AW, AN, and AS respectively, and the coordinates of the easternmost, westernmost, southernmost, and northernmost destinations in cluster B are BE, BW, BN, and BS respectively. Then calculate the distances between AE and BE, BW, BN, and BS respectively, the distances between AW and BE, BW, BN, and BS respectively, the distances between AN and BE, BW, BN, and BS respectively, and the distances between AS and BE, BW, BN, and BS respectively. That is, a total of 4 * 4 = 16 distances. Select the farthest distance as the maximum distance between the two clusters.

[0204] S804. Based on the central distance and the maximum distance between any two clusters, merge any two clusters that meet the merging conditions.

[0205] Specifically, the merging conditions include at least any one of the following: the central distance between the two clusters does not exceed the central distance threshold, and the maximum distance between the two clusters does not exceed the maximum distance threshold.

[0206] Exemplarily, the central distance threshold can be set to 500 meters, 800 meters, or 1000 meters. The maximum distance threshold can be set to 600 meters, 800 meters, or 1000 meters.

[0207] The above mainly introduces the solution provided by the embodiments of the present application from the perspective of the method. To implement the above functions, the destination prediction device includes the corresponding hardware structure and / or software module for executing each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of the examples described in the embodiments disclosed herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the way of hardware or computer software driving hardware depends on the specific application and design constraints of the technical solution. Skilled professionals can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0208] In an exemplary embodiment, the embodiments of the present application further provide a destination prediction model training device in the form of a virtual device, as Figure 9 shown. The device includes: a first acquisition module 901 and a first processing module 902.

[0209] The first acquisition module 901 is used to acquire a training sample set. The training sample set includes multiple training samples and the destination label corresponding to each sample. Each training sample includes the location information of the departure place of the travel data of the travel-regular vehicle, the remaining state of the driving energy of the vehicle at the time of departure, and the time information at the time of departure. The travel-regular vehicle is determined based on the daily travel times and the cumulative travel time of the vehicle per day.

[0210] The first processing module 902 is configured to train an initial model of the destination prediction model based on a training sample set to obtain the destination prediction model.

[0211] Furthermore, the first processing module 902 is specifically configured to obtain a training sample set, including: obtaining multiple pieces of original trip data within a preset area. Determining valid trip data that meets valid screening conditions from the multiple pieces of original trip data. Based on the daily trip frequency and the cumulative daily trip duration of the vehicles corresponding to the valid trip data, determining the trip data of regularly traveling vehicles from the valid trip data. Extracting training samples based on the trip data of regularly traveling vehicles to obtain the training sample set.

[0212] Furthermore, the valid screening conditions include at least one of the following: the duration of the trip data exceeds a preset duration threshold. The mileage of the trip data exceeds a preset mileage threshold. The distance between the start point and the end point of the trip data exceeds a preset distance threshold.

[0213] Furthermore, the first processing module 902 is specifically configured to determine regularly traveling vehicles from the vehicles corresponding to the valid trip data based on the daily trip frequency and the cumulative daily trip duration of the vehicles corresponding to the valid trip data, including: determining irregularly traveling vehicles, low-regularity traveling vehicles, and professional operation vehicles from the vehicles corresponding to the valid trip data based on the daily trip frequency and the cumulative daily trip duration of the vehicles corresponding to the valid trip data. Determining the vehicles other than the irregularly traveling vehicles, low-regularity traveling vehicles, and professional operation vehicles among the vehicles corresponding to the valid trip data as regularly traveling vehicles.

[0214] Furthermore, the irregularly traveling vehicles are vehicles that meet at least one of the following:

[0215] The standard deviation of the daily trip frequency of the vehicle is greater than the first reference standard deviation; the first reference standard deviation is equal to the product of the first standard deviation and the first coefficient plus the first average value; the first standard deviation is obtained by calculating the standard deviation of the daily trip frequency for each vehicle within the preset area and then calculating the standard deviation of the daily trip frequencies of all vehicles within the preset area again; the first average value is the average value of the daily trip frequencies of all vehicles within the preset area. The standard deviation of the cumulative daily trip duration of the vehicle is greater than the second reference standard deviation; the second reference standard deviation is based on the product of the second standard deviation and the first coefficient plus the second average value; the second standard deviation is obtained by calculating the standard deviation of the cumulative daily trip duration for each vehicle within the preset area and then calculating the standard deviation of the cumulative daily trip durations of all vehicles within the preset area again; the second average value is the average value of the cumulative daily trip durations of all vehicles within the preset area.

[0216] The low-regularity traveling vehicles are vehicles that meet at least one of the following:

[0217] The average number of daily trips of the vehicle is less than the first reference average value; the first reference average value is equal to the third average value minus the product of the third standard deviation and the first coefficient; the third average value is the average number of daily trip times of all vehicles in the preset area; the third standard deviation is the standard deviation of the daily trip times of all vehicles in the preset area. The number of trips of the vehicle within the recent preset duration exceeds the trip number threshold, and the number of trips per single day is less than the first number threshold; the first number threshold is equal to the third average value minus the product of the third standard deviation and the second coefficient. The average value of the cumulative travel time of the vehicle is less than the second reference average value; the second reference average value is equal to the fourth average value minus the product of the fourth standard deviation and the first coefficient; the fourth average value is the average value of the cumulative travel time of all vehicles in the preset area in one day; the fourth standard deviation is the standard deviation of the cumulative travel time of all vehicles in the preset area in one day. The number of trips of the vehicle within the recent preset duration exceeds the trip number threshold, and the cumulative travel time per single day is less than the first duration threshold; the first duration threshold is equal to the fourth average value minus the product of the fourth standard deviation and the second coefficient.

[0218] The professional operation vehicle is a vehicle that meets at least one of the following: The average number of daily trips of the vehicle is greater than the third reference average value; the third reference average value is equal to the product of the third standard deviation and the first coefficient plus the third average value. The number of trips of the vehicle within the recent preset duration exceeds the trip number threshold, and the number of trips per single day is greater than the second number threshold; the second number threshold is equal to the third average value plus the product of the third standard deviation and the second coefficient. The average value of the cumulative travel time of the vehicle is greater than the fourth reference average value; the fourth reference average value is equal to the fourth average value plus the product of the fourth standard deviation and the first coefficient. The number of trips of the vehicle within the recent preset duration exceeds the trip number threshold, and the cumulative travel time per single day is greater than the second duration threshold; the second duration threshold is equal to the fourth average value plus the product of the fourth standard deviation and the second coefficient.

[0219] Further, the first processing module 902 is specifically configured to extract training samples based on the trip data of the travel pattern vehicles to obtain the training sample set, including: clustering the destinations in the trip data of the travel pattern vehicles to obtain multiple destination clustering clusters. Determining a first clustering cluster, a second clustering cluster, and a third clustering cluster from the multiple destination clustering clusters. Among them, the first clustering cluster is the destination clustering cluster with the most destinations falling into the target departure place clustering cluster, and the target departure place clustering cluster is the departure place clustering cluster with the most departure places among the multiple departure place clustering clusters, and the multiple departure place clustering clusters are obtained by clustering the departure places of the first trips of each day in the trip data of the travel pattern vehicles; the second clustering cluster is the target place clustering cluster including the most target destinations, and the target destination is the destination of the previous trip in two trips with an interval duration greater than the duration; the third clustering cluster is the destination clustering cluster including the most driving energy replenishment places. Extracting training samples based on the trip data corresponding to the destinations in the first clustering cluster, the second clustering cluster, and the third clustering cluster.

[0220] Further, the first processing module 902 is further configured to: set the destination label of the training samples extracted based on the first clustering cluster to home. Set the destination label of the training samples extracted based on the second clustering cluster to company. Set the destination label of the training samples extracted based on the third clustering cluster to frequently visited driving energy replenishment place.

[0221] Further, the first processing module 902 is further configured to select the top M target clustering clusters including the most destinations from the destination clustering clusters other than the first clustering cluster, the second clustering cluster, and the third clustering cluster in the multiple destination clustering clusters, where M is a positive integer. Extracting training samples based on the trip data corresponding to the destinations in the top M target clustering clusters.

[0222] In an exemplary embodiment, the embodiment of the present application further provides a destination prediction device in the form of a virtual device, as Figure 10 shown. The device includes: a second acquisition module 1001 and a second processing module 1002.

[0223] The second acquisition module 1001 is configured to acquire the current position information of the vehicle, the remaining state of the driving energy of the vehicle, and the current time information.

[0224] The second processing module 1002 is configured to predict the trip destination based on the position information, the remaining state of the driving energy, the time information, and the destination prediction model; where the destination prediction model is trained based on a training sample set, and the training sample set includes multiple training samples and the destination label corresponding to each sample, and each training sample includes the position information of the departure place of the trip data of the travel pattern vehicle, the remaining state of the driving energy of the vehicle at the time of departure, and the time information at the time of departure.

[0225] Further, the destination prediction model includes a Bayesian probability model. The second processing module 1002 is specifically configured to predict the trip destination based on the location information, the remaining state of the driving energy, the time information, and the destination prediction model, including: predicting the travel probability of each of multiple candidate destinations based on the location information, the remaining state of the driving energy, the time information, and the Bayesian probability model. Determining the top N candidate destinations with the highest travel probabilities as the trip destination, where N is a positive integer.

[0226] Further, the second acquisition module 1001 is further configured to, when detecting that the vehicle is powered on, acquire the current location information of the vehicle, the remaining state of the current driving energy of the vehicle, and the current time information.

[0227] Regarding the device in the above embodiments, the specific manners in which each module performs operations have been described in detail in the embodiments related to the method, and will not be elaborated herein.

[0228] Figure 11 is a block diagram of a vehicle shown according to an exemplary embodiment. As Figure 11 shown, the vehicle 1100 includes but is not limited to: a processor 1101 and a memory 1102.

[0229] Among them, the above-mentioned memory 1102 is used to store the executable instructions of the above-mentioned processor 1101. It can be understood that the above-mentioned processor 1101 is configured to execute instructions to implement the bypass flow method in the above embodiments.

[0230] It should be noted that those skilled in the art can understand that Figure 11 the vehicle structure shown in Figure 11 does not constitute a limitation on the vehicle. The vehicle may include more or fewer components than

[0231] shown, or combine certain components, or have different component arrangements.

[0232] The memory 1102 can be used to store software programs and various data. The memory 1102 mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required by at least one functional module (such as a determination unit, a processing unit, etc.). In addition, the memory 1102 can include high-speed random access memory and can also include non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0233] In an exemplary embodiment, a computer-readable storage medium including instructions is also provided. For example, the memory 1102 including instructions, and the above instructions can be executed by the processor 1101 of the vehicle 1100 to implement the destination prediction method in the above embodiment.

[0234] In actual implementation, Figure 9 the functions of the first acquisition module 901 and the first processing module 902 in Figure 10 the functions of the second acquisition module 1001 and the first processing module 1002 in Figure 11 can be implemented by the processor 1101 in

[0235] calling a computer program stored in the memory 1102. The specific execution process can refer to the description of the method part in the above embodiment and will not be elaborated here.

[0235] Optionally, the computer-readable storage medium can be a non-temporary computer-readable storage medium. For example, the non-temporary computer-readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0236] In an exemplary embodiment, the embodiment of the present application also provides a computer program product including one or more instructions, and the one or more instructions can be executed by the processor 1201 of the vehicle to complete the destination prediction method in the above embodiment.

[0237] It should be noted that when the instructions in the above computer-readable storage medium or the one or more instructions in the computer program product are executed by the processor of the vehicle, each process of the above method embodiment is implemented, and the same technical effects as the above method can be achieved. To avoid repetition, it will not be elaborated here.

[0238] Through the description of the above embodiments, those skilled in the art can clearly understand that for the convenience and conciseness of description, only the above division of each functional module is used as an example. In actual applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above.

[0239] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection between each other can be through some interfaces. The indirect coupling or communication connection of devices or units can be in electrical, mechanical or other forms.

[0240] The units described as separate components may or may not be physically separated. The components displayed as units can be one physical unit or multiple physical units, that is, they can be located in one place, or they can be distributed to multiple different places. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0241] In addition, each functional unit in various embodiments of the present application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0242] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solution of the embodiments of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to enable a device (which can be a single-chip microcomputer, a chip, etc.) or a processor to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: USB flash drives, mobile hard disks, ROM, RAM, magnetic disks, or optical discs and other various media that can store program codes.

[0243] As described above, it is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions within the technical scope disclosed in the present application should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for training a destination prediction model, characterized in that The method includes: Obtaining a training sample set; The training sample set includes a plurality of training samples and a destination label corresponding to each sample. Each training sample includes the location information of the departure place of the travel data of the regularly traveling vehicle, the remaining state of the driving energy of the vehicle at the time of departure, and the time information at the time of departure; the regularly traveling vehicle is determined based on the daily travel times and the cumulative travel time of the vehicle per day; Training an initial model of the destination prediction model based on the training sample set to obtain the destination prediction model.

2. The method for training a destination prediction model according to claim 1, wherein The obtaining of the training sample set includes: Obtaining a plurality of original travel data within a preset area; Determining valid travel data that meets the valid screening conditions from the plurality of original travel data; Based on the daily travel times and the cumulative travel time of the vehicle per day corresponding to the valid travel data, determining regularly traveling vehicles from the vehicles corresponding to the valid travel data; Extracting training samples based on the travel data of the regularly traveling vehicles to obtain the training sample set; Extracting training samples based on the valid travel data to obtain the training sample set.

3. The method for training a destination prediction model according to claim 2, wherein The valid screening conditions include at least one of the following: The travel time of the travel data exceeds a preset time threshold; The mileage of the travel data exceeds a preset mileage threshold; The distance between the starting point and the ending point of the travel data exceeds a preset distance threshold.

4. The method for training a destination prediction model according to claim 2, wherein The determining of regularly traveling vehicles from the vehicles corresponding to the valid travel data based on the daily travel times and the cumulative travel time of the vehicle per day corresponding to the valid travel data includes: Based on the daily travel times and the cumulative travel time of the vehicle per day corresponding to the valid travel data, determining irregularly traveling vehicles, low-regularity traveling vehicles, and professional operation vehicles from the vehicles corresponding to the valid travel data; Determining that the vehicles other than the irregularly traveling vehicles, the low-regularity traveling vehicles, and the professional operation vehicles among the vehicles corresponding to the valid travel data are the regularly traveling vehicles.

5. The method for training a destination prediction model according to claim 4, wherein The irregularly traveling vehicles are vehicles that meet at least one of the following: The standard deviation of the daily travel times of the vehicle is greater than a first reference standard deviation; the first reference standard deviation is equal to the product of a first standard deviation and a first coefficient plus a first average value; the first standard deviation is obtained by calculating the standard deviation of the daily travel times of each vehicle within a preset area and then calculating the standard deviation of the standard deviations of the daily travel times of all vehicles within the preset area again; the first average value is the average value of the standard deviations of the daily travel times of all vehicles within the preset area; The standard deviation of the cumulative travel time of the vehicle per day is greater than a second reference standard deviation; the second reference standard deviation is based on the product of a second standard deviation and a first coefficient plus a second average value; the second standard deviation is obtained by calculating the standard deviation of the cumulative travel time of each vehicle per day within a preset area and then calculating the standard deviation of the standard deviations of the cumulative travel times of all vehicles within the preset area again; the second average value is the average value of the standard deviations of the cumulative travel times of all vehicles within the preset area; The low-trip-pattern vehicles are vehicles that meet at least one of the following conditions: The average number of daily trips of the vehicle is less than the first reference average; the first reference average is equal to the third average minus the product of the third standard deviation and the first coefficient; the third average is the average number of trips of all vehicles in a preset area in one day; the third standard deviation is the standard deviation of the number of trips of all vehicles in a preset area in one day; The number of trips of the vehicle within a preset recent period exceeds the trip number threshold, and the number of trips in a single day is less than the first trip number threshold; the first trip number threshold is equal to the third average minus the product of the third standard deviation and the second coefficient; The average cumulative travel time of the vehicle is less than the second reference average; the second reference average is equal to the fourth average minus the product of the fourth standard deviation and the first coefficient; The fourth average is the average cumulative travel time of all vehicles in a preset area in one day; the fourth standard deviation is the standard deviation of the average cumulative travel time of all vehicles in a preset area in one day; The number of trips of the vehicle within a preset recent period exceeds the trip number threshold, and the cumulative travel time in a single day is less than the first duration threshold; The first duration threshold is equal to the fourth average minus the product of the fourth standard deviation and the second coefficient; The professional operation vehicles are vehicles that meet at least one of the following conditions: The average number of daily trips of the vehicle is greater than the third reference average; the third reference average is equal to the product of the third standard deviation and the first coefficient plus the third average; The number of trips of the vehicle within a preset recent period exceeds the trip number threshold, and the number of trips in a single day is greater than the second trip number threshold; the second trip number threshold is equal to the third average plus the product of the third standard deviation and the second coefficient; The average cumulative travel time of the vehicle is greater than the fourth reference average; the fourth reference average is equal to the fourth average plus the product of the fourth standard deviation and the first coefficient; The number of trips of the vehicle within a preset recent period exceeds the trip number threshold, and the cumulative travel time in a single day is greater than the second duration threshold; The second duration threshold is equal to the fourth average plus the product of the fourth standard deviation and the second coefficient.

6. The method for training a destination prediction model according to claim 2, wherein Extracting training samples from the travel data of the travel-pattern vehicles to obtain the training sample set includes: Clustering the destinations in the travel data of the travel-pattern vehicles to obtain multiple destination clustering clusters; Determining a first clustering cluster, a second clustering cluster, and a third clustering cluster from the multiple destination clustering clusters; Among them, the first clustering cluster is the destination clustering cluster with the most destinations falling into the target departure clustering cluster, and the target departure clustering cluster is the departure clustering cluster with the most departures among multiple departure clustering clusters. The multiple departure clustering clusters are obtained by clustering the departure places of the first trips of each day in the travel data of the travel-pattern vehicles; the second clustering cluster is the target destination clustering cluster with the most target destinations, and the target destination is the destination of the previous trip in two trips with an interval duration greater than the duration; the third clustering cluster is the destination clustering cluster with the most driving energy replenishment places; Extract training samples based on the trip data corresponding to the destinations in the first clustering cluster, the second clustering cluster, and the third clustering cluster.

7. The method for training a destination prediction model according to claim 6, wherein The method further includes: Set the destination label of the training samples extracted based on the first clustering cluster to home; Set the destination label of the training samples extracted based on the second clustering cluster to company; Set the destination label of the training samples extracted based on the third clustering cluster to frequently visited driving energy replenishment places.

8. The method for training a destination prediction model according to claim 6, wherein The method further includes: Select the top M target clustering clusters with the most destinations from the destination clustering clusters other than the first clustering cluster, the second clustering cluster, and the third clustering cluster among the multiple destination clustering clusters; Extract training samples based on the trip data corresponding to the destinations in the top M target clustering clusters; M is a positive integer.

9. A destination prediction method, characterized in that, The method includes: Obtain the current location information of the vehicle, the remaining state of the driving energy of the vehicle, and the current time information; Predict the trip destination based on the location information, the remaining state of the driving energy, the time information, and the destination prediction model; wherein, the destination prediction model is trained based on a training sample set, the training sample set includes multiple training samples and the destination label corresponding to each sample, and each training sample includes the location information of the departure place of the trip data of the vehicle with travel patterns, the remaining state of the driving energy of the vehicle at the time of departure, and the time information at the time of departure.

10. The destination prediction method according to claim 9, wherein The destination prediction model includes a Bayesian probability model. The predicting the trip destination based on the location information, the remaining state of the driving energy, the time information, and the destination prediction model includes: Predict the travel probability of each of multiple candidate destinations based on the location information, the remaining state of the driving energy, the time information, and the Bayesian probability model; Determine the top N candidate destinations with the highest travel probability as the trip destination; N is a positive integer.

11. The destination prediction method according to claim 9, wherein The obtaining the current location information of the vehicle, the remaining state of the driving energy of the vehicle, and the current time information includes: When it is detected that the vehicle is fully powered on, obtain the current location information of the vehicle, the remaining state of the driving energy of the vehicle, and the current time information.

12. A destination prediction model training device, characterized in that, The device includes: A first obtaining module, configured to obtain a training sample set; the training sample set includes multiple training samples and the destination label corresponding to each sample, and each training sample includes the location information of the departure place of the trip data of the vehicle with travel patterns, the remaining state of the driving energy of the vehicle at the time of departure, and the time information at the time of departure; the vehicle with travel patterns is determined based on the daily travel times and the cumulative daily travel duration of the vehicle; A first processing module, configured to train an initial model of the destination prediction model based on the training sample set to obtain the destination prediction model.

13. A destination prediction device, characterized in that, The device includes: A second obtaining module, configured to obtain the current location information of the vehicle, the remaining state of the driving energy of the vehicle, and the current time information; A second processing module, configured to predict a trip destination based on the location information, the remaining state of the driving energy, the time information, and a destination prediction model; wherein the destination prediction model is obtained by training based on a training sample set, the training sample set includes a plurality of training samples and destination labels corresponding to each sample, and each training sample includes the location information of the departure place of the trip data of the travel pattern vehicle, the remaining state of the driving energy of the vehicle at the time of departure, and the time information at the time of departure.

14. A vehicle, characterized in that, Comprising: a processor and a memory; the memory stores instructions executable by the processor; when the processor is configured to execute the instructions, the vehicle implements the destination prediction method according to any one of claims 9-11.

Images