A training method of a model, an acquisition method of representation information, and a route planning method

Abstract

Embodiments of the present disclosure relate to a training method and device of a representation information obtaining model, a representation information obtaining method and device, a route planning method and device, and an electronic device, a computer storage medium and a computer program product. The training method comprises: obtaining a navigation behavior feature sequence of a to-be-characterized object, the navigation behavior feature sequence comprising feature information of a planned route of at least two navigation behaviors and feature information of an actual driving route; obtaining feature information of a planned route with the maximum actual driving coverage rate of each navigation behavior of an object, the object comprising the to-be-characterized object and other objects except the to-be-characterized object; generating a training sample based on the navigation behavior feature sequence of the to-be-characterized object and the feature information of the planned route with the maximum actual driving coverage rate of each navigation behavior of the object; and training the representation information obtaining model with the training sample to obtain a trained representation information obtaining model, which can obtain a relatively accurate representation information obtaining model.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of computer, and particularly relates to a training method and device of a representation information acquisition model, a representation information acquisition method and device, a route planning method and device, and an electronic device, computer storage medium and computer program product. BACKGROUND

[0002] With the development of computer technology, the application of navigation technology and navigation software in daily life gradually increases. An object (for example, an end user) can autonomously and conveniently select a route that can reach a destination by using the navigation software, thereby improving the convenience of the user's travel.

[0003] In the prior art, the navigation software recommends all selectable routes between a starting point and a destination to a user according to preset rules, and the user selects a route that meets the needs from the routes. In this way, the route cannot be personalized for different users, and the user experience is poor. SUMMARY

[0004] In order to solve the above technical problems or at least partially solve the above technical problems, the present disclosure provides a training method and device of a representation information acquisition model, a representation information acquisition method and device, a route planning method and device, and an electronic device, computer storage medium and computer program product.

[0005] The present disclosure provides a training method of a representation information acquisition model, comprising:

[0006] obtaining a navigation behavior feature sequence of an object to be represented, the navigation behavior feature sequence comprising feature information of a planning route of at least two navigation behaviors and feature information of an actual driving route;

[0007] obtaining feature information of a planning route with the maximum actual driving coverage rate of each navigation behavior of the object, the object comprising the object to be represented and other objects except the object to be represented;

[0008] generating a training sample based on the navigation behavior feature sequence of the object to be represented and the feature information of the planning route with the maximum actual driving coverage rate of each navigation behavior of the object;

[0009] training the representation information acquisition model with the training sample to obtain a trained representation information acquisition model.

[0010] The present disclosure also provides a representation information acquisition method of an object, comprising:

[0011] obtaining a navigation behavior feature sequence of an object, the navigation behavior feature sequence comprising feature information of a planning route of at least two navigation behaviors and feature information of an actual driving route;

[0012] The representation information of the object is obtained by the representation information obtaining model trained by any of the above methods based on the navigation behavior feature sequence of the object.

[0013] The present disclosure further provides a route planning method, comprising:

[0014] The representation information of the object is obtained based on any of the above object representation information obtaining methods.

[0015] The planning route recommended to the object is determined based on at least the representation information of the object, the start and end points selected by the object, and the road condition information.

[0016] The present disclosure further provides a representation information obtaining model training device, comprising:

[0017] The sequence obtaining module is configured to obtain a navigation behavior feature sequence of the object to be represented, the navigation behavior feature sequence comprising feature information of a planning route of at least two navigation behaviors and feature information of an actual driving route.

[0018] The feature information obtaining module is configured to obtain feature information of a planning route with the maximum actual driving coverage rate of each navigation behavior of the object, the object comprising the object to be represented and other objects except the object to be represented.

[0019] The sample generating module is configured to generate a training sample based on the navigation behavior feature sequence of the object to be represented and the feature information of the planning route with the maximum actual driving coverage rate of each navigation behavior.

[0020] The training module is configured to train the representation information obtaining model with the training sample to obtain a trained representation information obtaining model.

[0021] The present disclosure further provides an object representation information obtaining device, comprising:

[0022] The sequence obtaining module is configured to obtain a navigation behavior feature sequence of the object to be represented, the navigation behavior feature sequence comprising feature information of a planning route of at least two navigation behaviors and feature information of an actual driving route.

[0023] The representation information obtaining module is configured to obtain the representation information of the object by a representation information obtaining model trained by any of the above devices based on the navigation behavior feature sequence of the object to be represented.

[0024] The present disclosure further provides a route planning device, comprising:

[0025] The representation information obtaining module is configured to obtain the representation information of the object obtained by any of the above object representation information obtaining devices.

[0026] a route planning module configured to determine a recommended planning route to the object based on at least the representation information of the object and the start and end points selected by the object.

[0027] The embodiments of the present disclosure further provide a map navigation system, comprising any of the route planning devices.

[0028] The embodiments of the present disclosure further provide a network car-hailing platform system, comprising any of the route planning devices.

[0029] The embodiments of the present disclosure further provide an electronic device, comprising:

[0030] a processor;

[0031] a memory configured to store executable instructions of the processor;

[0032] the processor is configured to read the executable instructions from the memory and execute the instructions to implement any of the training methods of the representation information acquisition model, any of the methods of acquiring the representation information of the object, or any of the route planning methods.

[0033] The embodiments of the present disclosure further provide a computer readable storage medium, which stores a computer program, and the computer program is configured to execute any of the training methods of the representation information acquisition model, any of the methods of acquiring the representation information of the object, or any of the route planning methods.

[0034] The embodiments of the present disclosure further provide a computer program product, which is configured to execute any of the training methods of the representation information acquisition model, any of the methods of acquiring the representation information of the object, or any of the route planning methods.

[0035] Compared with the prior art, the technical scheme provided by the embodiments of the present disclosure has at least the following advantages: in the embodiments of the present disclosure, the navigation behavior feature sequence of the to-be-characterized object is obtained, the navigation behavior feature sequence includes feature information of a planned route of at least two navigation behaviors and feature information of an actual driving route; feature information of a planned route with the maximum actual travel coverage of each navigation behavior of an object including the to-be-characterized object and other objects except the to-be-characterized object is obtained; and a training sample is generated based on the navigation behavior feature sequence of the to-be-characterized object and the feature information of the planned route with the maximum actual travel coverage of each navigation behavior of the object, and the training sample is used to train a characterization information acquisition model to obtain a trained characterization information acquisition model. Based on the navigation behavior feature sequence and the feature information of the planned route with the maximum actual travel coverage of each navigation behavior, information for characterizing the preferences of the object, i.e., object characterization information, can be extracted. Thus, for different to-be-characterized objects, the navigation behavior feature sequence of the to-be-characterized object and the feature information of the planned route with the maximum actual travel coverage of the to-be-characterized object and other objects except the to-be-characterized object are obtained, and a training sample for training the model is constructed based thereon. Further, the training sample is used to train the characterization information acquisition model, and a characterization information acquisition model that can accurately determine the characterization information of the to-be-characterized object can be obtained, so that the trained characterization information acquisition model can be used to automatically capture object characterization information and determine object preferences, which is conducive to realizing personalized route planning based on object preferences and improving the object navigation experience. At the same time, the characterization information can be directly obtained based on the original information of the navigation behavior feature sequence of the object, and there is no problem of limitations of model training by the experience of data analysts, so that the characterization information acquisition model is easy to maintain and can realize effective utilization of the navigation behavior feature sequence of the object. BRIEF DESCRIPTION OF DRAWINGS

[0036] The above and other features, advantages, and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the sizes and elements are not necessarily drawn to scale.

[0037] Figure 1 A flowchart of a training method of a characterization information acquisition model provided by the embodiments of the present disclosure is shown in the figure;

[0038] Figure 2 For Figure 1 In the method shown in the figure, a specific flowchart of S103 is shown in the figure;

[0039] Figure 3 A flowchart of another training method of a characterization information acquisition model provided by the embodiments of the present disclosure is shown in the figure;

[0040] Figure 4 A structural schematic diagram of a representation information acquisition model provided by an embodiment of the present disclosure is shown in the figure;

[0041] Figure 5 For Figure 1 The specific flowchart of S104 in the flow shown in the figure is shown in the figure;

[0042] Figure 6 For Figure 5 The specific flowchart of S301 in the flow shown in the figure is shown in the figure;

[0043] Figure 7 A flowchart of an object representation information acquisition method provided by an embodiment of the present disclosure is shown in the figure;

[0044] Figure 8 A flowchart of a route planning method provided by an embodiment of the present disclosure is shown in the figure;

[0045] Figure 9 A structural schematic diagram of a training device of a representation information acquisition model provided by an embodiment of the present disclosure is shown in the figure;

[0046] Figure 10 A structural schematic diagram of another training device of a representation information acquisition model provided by an embodiment of the present disclosure is shown in the figure;

[0047] Figure 11 A structural schematic diagram of an object representation information acquisition device provided by an embodiment of the present disclosure is shown in the figure;

[0048] Figure 12 A structural schematic diagram of a route planning device provided by an embodiment of the present disclosure is shown in the figure;

[0049] Figure 13 A structural schematic diagram of an electronic device provided by an embodiment of the present disclosure is shown in the figure. DETAILED DESCRIPTION

[0050] Embodiments of the present disclosure will be described in more detail by making reference to the accompanying drawings. While certain embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be embodied in various forms and should not be construed as being limited to the embodiments set forth herein, rather these embodiments are provided so as to more thoroughly and completely understand the present disclosure. It is understood that the drawings and embodiments of the present disclosure are for exemplary purposes only and are not intended to limit the scope of protection of the present disclosure.

[0051] It should be understood that each step described in the method embodiments of the present disclosure can be executed in different order and / or in parallel. In addition, the method embodiments can include additional steps and / or omit the execution of the steps shown. The scope of the present disclosure is not limited in this respect.

[0052] As used herein, the term "includes" and its variants are open-ended, meaning that "includes but is not limited to." The term "based on" means "based, at least in part, on." The term "one embodiment" means "at least one embodiment." The term "another embodiment" means "at least one additional embodiment." The term "some embodiments" means "at least some embodiments." Related terms have corresponding meanings.

[0053] It should be noted that the terms "first", "second", and the like in the present disclosure are merely used to distinguish different devices, modules or units, and do not imply the order or interdependence of the functions performed by these devices, modules or units.

[0054] It should be noted that the modification of "one" or "multiple" in the present disclosure is illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise explicitly indicated in the context, it should be understood as "one or more".

[0055] In order to solve the problem that the prior art cannot realize personalized route pushing for different objects, the related technology proposes an identification method based on object historical navigation behavior statistics, which can specifically include: obtaining object historical navigation information; statistics of feature indicators in the object historical navigation information, such as the number of times the object selects a route with shorter time, the number of times the object selects a route with less charge, etc.; the statistical feature indicators are provided as input of a training model to provide object personalized information. However, in this method, the statistics of feature indicators are based on the experience of data analysts to specify statistical rules, which has certain limitations in the accuracy and recall of object preference extraction; at the same time, object preference will change over time, and the method based on feature indicator statistics is difficult to capture such changes, resulting in the model failing to adapt in time when the object preference changes, and leading to poor accuracy of object preference extraction.

[0056] In order to solve at least part of the above problems, the embodiments of the present disclosure provide a method for generating training samples based on a navigation behavior feature sequence of a to-be-characterized object and feature information of a planning route with the maximum coverage rate of the object (including the to-be-characterized object and other objects except the to-be-characterized object); and training a characterization information acquisition model using the training samples to obtain a trained characterization information acquisition model capable of accurately extracting object preference. The method can be regarded as a training method of an object personalized preference automatic mining learning model, and the trained characterization information acquisition model can acquire object characterization information based on the navigation behavior feature sequence of the object, that is, extract the personalized preference (i.e., "object preference") of the object, which can be used to assist personalized route pushing, so that the route planning can meet the personalized needs of different objects and improve the object navigation experience.

[0057] The training method of the representation information acquisition model, the representation information acquisition method, and the route planning method can be applied to various application scenarios of route planning technology and navigation technology, for example, navigation when an object autonomously drives, rides, or walks, and navigation when the object takes a taxi offline, takes a network taxi, or hires a driver. The trained representation information acquisition model can determine the representation information of the object and directly apply it to a downstream model, such as route planning, route recommendation, crowd representation of the object (for example, the representation information of the object can determine whether the object is sensitive to money, and based on this, the platform can be guided to issue coupons), and the like. For example, the representation information of the object indicates that the object prefers to save money, and the recommended route does not pass through a highway; or the representation information of the object indicates that the object prefers to save time, and the recommended route preferentially includes a highway. The above method can be locally executed on a terminal device where an application or a mini-program with a navigation function is located, or can be executed by a server that interacts with the terminal device, and is not limited herein.

[0058] The training method of the representation information acquisition model, the representation information acquisition method, and the route planning method can be applied to various application scenarios of route planning technology and navigation technology, for example, navigation when an object autonomously drives, rides, or walks, and navigation when the object takes a taxi offline, takes a network taxi, or hires a driver. The trained representation information acquisition model can determine the representation information of the object and directly apply it to a downstream model, such as route planning, route recommendation, crowd representation of the object (for example, the representation information of the object can determine whether the object is sensitive to money, and based on this, the platform can be guided to issue coupons), and the like. For example, the representation information of the object indicates that the object prefers to save money, and the recommended route does not pass through a highway; or the representation information of the object indicates that the object prefers to save time, and the recommended route preferentially includes a highway. The above method can be locally executed on a terminal device where an application or a mini-program with a navigation function is located, or can be executed by a server that interacts with the terminal device, and is not limited herein. Figures 1-13 The training method of the representation information acquisition model, the representation information acquisition method, and the route planning method can be applied to various application scenarios of route planning technology and navigation technology, for example, navigation when an object autonomously drives, rides, or walks, and navigation when the object takes a taxi offline, takes a network taxi, or hires a driver. The trained representation information acquisition model can determine the representation information of the object and directly apply it to a downstream model, such as route planning, route recommendation, crowd representation of the object (for example, the representation information of the object can determine whether the object is sensitive to money, and based on this, the platform can be guided to issue coupons), and the like. For example, the representation information of the object indicates that the object prefers to save money, and the recommended route does not pass through a highway; or the representation information of the object indicates that the object prefers to save time, and the recommended route preferentially includes a highway. The above method can be locally executed on a terminal device where an application or a mini-program with a navigation function is located, or can be executed by a server that interacts with the terminal device, and is not limited herein.

[0059] Figure 1 A flowchart of a training method of a representation information acquisition model is provided for the embodiments of the present disclosure. Refer to Figure 1 The training method of the representation information acquisition model comprises the following steps.

[0060] S101, obtaining a navigation behavior feature sequence of a to-be-represented object.

[0061] The navigation behavior feature sequence can also be referred to as a historical navigation behavior feature sequence, and comprises feature information of a planned route of at least two navigation behaviors and feature information of an actual driving route.

[0062] A navigation behavior includes the entire process of an object from a planned route at the beginning to following a navigation in the middle to reaching a destination to end the navigation. In one navigation behavior, the number of planned routes can be one, two, or more, which is determined based on the number of optional routes between a starting point (i.e., a starting point) and a destination (i.e., an ending point); the number of actual driving routes is only one, which is the route actually traveled by the object from the starting point to the destination. The navigation behavior feature sequence can be features associated with at least two navigation behaviors concatenated in time sequence, which can ensure a large amount of data for model training in the subsequent steps, thereby making the representation information acquisition model obtained by training have high extraction accuracy of object representation information and accurately determining the preferences of the object.

[0063] The feature information is used to represent the characteristics of the planned route and the actual driving route. Correspondingly, the feature information can be obtained based on extraction, statistics, and other methods known to those skilled in the art, which are not described or limited herein.

[0064] For example, the feature information can include time length, distance, number of traffic lights, number of navigation actions, and cost information. The number of navigation actions can include the number of straight driving, turning, U-turn, entering ramp, entering loop, and exiting loop, which are only exemplary and do not constitute a limitation on the embodiments of the present disclosure.

[0065] For example, the navigation behavior feature sequence can be stored locally on the terminal device and / or on a cloud server. In this step, the stored navigation behavior feature sequence can be obtained based on a data retrieval instruction. When the navigation behavior feature sequence is stored on the terminal device and the method is executed on the terminal device, the data retrieval can be directly performed. Similarly, when the navigation behavior feature sequence is stored on the cloud server and the method is executed on the cloud server, the data retrieval can also be directly performed. When the navigation behavior data is stored on the terminal device and the method is executed on the cloud server, the cloud server can issue a data retrieval instruction to the terminal device, and the terminal device can upload the navigation behavior feature sequence to the cloud server based on the received data retrieval instruction.

[0066] S102, obtaining feature information of a planned route with the maximum actual driving coverage rate of each navigation behavior of the object.

[0067] The object includes the object to be represented and other objects except the object to be represented. The feature information of the planned route with the maximum actual driving coverage rate of the object can reflect the preference of the object when selecting the planned route. Therefore, by obtaining the feature information of the planned route with the maximum actual driving coverage rate of each navigation behavior of the object to be represented and other objects except the object to be represented, the object preference of the object to be represented can be distinguished from the object preference of other objects, and data for constructing positive samples and negative samples in the training sample can be provided, which will be described in detail below.

[0068] The actual driving coverage rate is used to represent the coverage degree of the actual driving route to the planned route, that is, the proportion of the overlapping part of the route in the planned route, which can also be understood as the effective utilization rate of the planned route in the actual driving process of the object.

[0069] Therefore, the planned route with the maximum actual driving coverage rate is the planned route with the highest effective utilization rate in the driving process of the object, that is, the planned route with the highest proportion of the overlapping part of the route.

[0070] Exemplarily, for each navigation behavior, the actual driving route and all the planning routes can be calculated respectively to obtain the actual travel coverage, and the relative size of the actual travel coverage obtained by comparison can be compared by using the difference with 0 or by using the ratio with 1. The planning route corresponding to the maximum actual travel coverage is the planning route with the maximum actual travel coverage for the navigation behavior.

[0071] Exemplarily, the actual travel coverage can be expressed in the form of percentage and take a value between 0-100%. Wherein, the actual travel coverage value is 0, which represents that there is no overlap between the planning route and the actual driving route; the actual travel coverage value is 100%, which represents that the planning route is completely covered by the actual driving route; when the actual travel coverage value is greater than 0 and less than 100%, the greater the actual travel coverage value, the more the part of the planning route that overlaps with the actual driving route.

[0072] It can be understood that the planning route with the maximum actual travel coverage can be the planning route with the actual travel coverage value of 100%, or other planning routes with the actual travel coverage value less than 100%. As long as the actual travel coverage of the planning route is the maximum among all the planning routes in a single navigation behavior, the specific value is not limited.

[0073] In other embodiments, the actual travel coverage can also be expressed in other forms known to those skilled in the art, which are not limited herein.

[0074] S103, generating a training sample based on the navigation behavior feature sequence of the to-be-characterized object and the feature information of the planning route with the maximum actual travel coverage of each navigation behavior of the object.

[0075] Wherein, the representation information acquisition model is used to extract the object representation information, and the object representation information is used for the personalized preference of the to-be-characterized object. Generally, the personalized preferences of different objects are different, and the corresponding object representation information is not the same. Specifically, the demand of the object for the route is multi-objective, such as short time, short distance, few navigation actions or low cost, and different objects tend to different objectives, which is called the personalized preference of the object.

[0076] In this step, a training sample is generated based on the navigation sequence of the navigation behavior of the to-be-characterized object obtained in the foregoing step and the feature information of the planning route with the maximum actual travel coverage of each navigation behavior of the object; so as to train the representation information acquisition model using the training sample in the subsequent step, which will be described below in combination with S104.

[0077] S104, training the representation information acquisition model with the training sample to obtain the trained representation information acquisition model.

[0078] In combination with the foregoing, the training sample generated in S103 is used to train the representation information obtaining model, that is, the representation information obtaining model is trained based on data capable of representing the object preference, so that the representation information obtaining model obtained after training can accurately determine the object representation information, extract the object individual preference, and facilitate the realization of individual route planning for different objects to improve the object navigation experience.

[0079] The training method of the representation information obtaining model provided in the embodiments of the present disclosure can generate training samples based on the navigation behavior feature sequence representing the object and the feature information of the route plan with the maximum actual travel coverage of each navigation behavior of the object (including the object to be represented and other objects except the object to be represented) for different objects to be represented. The training samples are generated based on data capable of representing the object preference. Thus, the representation information obtaining model can be trained using the training samples, and the representation information obtaining model capable of accurately determining the object representation information of the object can be obtained, so that the object representation information can be automatically captured using the trained representation information obtaining model, and the object individual preference can be determined. This facilitates individual route planning based on the object individual preference, and improves the object navigation experience. Moreover, the object representation information can be directly obtained based on the original information of the navigation behavior feature sequence of the object, so that the model training method does not have the problem of limitations of the experience of data analysts on model training, making the representation information obtaining model easy to maintain and enabling effective use of the navigation behavior feature sequence of the object.

[0080] In some embodiments, based on Figure 1 In S101, the feature information of the route plan of at least two navigation behaviors can include at least two of the following:

[0081] the feature information of the route plan with the maximum actual travel coverage;

[0082] the feature information of the route plan ranked first;

[0083] the feature information of the route plan with the shortest static time consumption.

[0084] The route plan with the maximum actual travel coverage is the route plan that can best meet the individual needs of the object, that is, the route plan that best meets the preference of the object to be represented. In combination with the foregoing, the details are not repeated here.

[0085] The route plan ranked first is the route plan ranked first according to a preset rule.

[0086] Exemplarily, the preset rule can include time priority, distance priority, cost priority, navigation action number priority, or a comprehensive rule combining at least two factors. Among them, the time priority can be in the order of time from short to long, the distance priority can be in the order of distance from short to long, the cost priority can be in the order of cost from less to more, and the navigation action number priority can be in the order of navigation action number from less to more; the comprehensive rule can be a comprehensive sorting of the planned route by combining the weights corresponding to different factors, which is not described or limited here.

[0087] Among them, the planned route with the shortest static time consumption is the planned route with the shortest time consumption without considering any dynamic information; it can also be understood as the planned route with the shortest time consumption in the case of no congestion and free flow, that is, the "regular fastest planned route" or "regular fastest recall first route", which is distinguished from the "dynamic fastest planned route". It can be understood that the "dynamic fastest planned route" is the planned route with the shortest time consumption under the condition of combining dynamic information such as actual road conditions and traffic flow.

[0088] In this step, the planned route with the largest actual walking coverage rate, the planned route ranked first in the recommendation, and the planned route with the shortest static time consumption can all have an associated relationship with the object representation information. By obtaining the feature information of at least two of the above planned routes, data can be provided for subsequent training models, and the training effect is better when the number of planned routes is larger.

[0089] In other embodiments, the planned route in the object navigation behavior can also include other planned routes for associating object representation information, which are not described or limited here.

[0090] In some embodiments, the training sample can include positive samples and negative samples. Among them, the feature information of the planned route with the largest actual walking coverage rate of each navigation behavior of the object to be represented and the navigation behavior feature sequence of the object to be represented are generated to generate a training sample as a positive sample; the feature information of the planned route with the largest actual walking coverage rate of each navigation behavior of other objects and the navigation behavior feature sequence of the object to be represented are generated to generate a training sample as a negative sample.

[0091] Exemplarily, Figure 2 For Figure 1 In the method shown, the specific flowchart of S103. On the basis of Figure 1 Referring to Figure 2 , S103 can include:

[0092] S201, based on the navigation behavior feature sequence of the object to be represented and the feature information of the planned route with the largest actual walking coverage rate of each navigation behavior of the object to be represented, generate a positive sample in the training sample.

[0093] S202, based on the navigation behavior feature sequence of the to-be-characterized object and the feature information of the route with the maximum actual travel coverage of each navigation behavior of the other object, generate a negative sample in the training sample.

[0094] The positive sample can reflect the preference of the to-be-characterized object, and the negative sample can reflect the preference of the other object except the to-be-characterized object, so as to distinguish the other object from the to-be-characterized object and accurately obtain the personalized preference of the to-be-characterized object.

[0095] In some embodiments, based on Figure 2 S201 can include: generating a first feature pair of each navigation behavior as a positive sample in the training sample; the first feature pair includes a navigation behavior feature sequence up to the last navigation behavior and feature information of the route with the maximum actual travel coverage of the last navigation behavior.

[0096] The first feature pair of each navigation behavior includes a navigation behavior feature sequence up to the last navigation behavior and feature information of the route with the maximum actual travel coverage of the last navigation behavior.

[0097] The navigation behavior feature sequence up to the last navigation behavior can also be referred to as historical statistical information, and the feature information of the route with the maximum actual travel coverage of the last navigation behavior can also be referred to as single navigation information; based on this, the first feature pair of each navigation behavior can be a pair of historical statistical information and single navigation information of the object navigation behavior.

[0098] For example, the navigation behavior feature sequence can be represented as [navi1, navi2, …, navin]; the feature information of the route with the maximum actual travel coverage can be represented as unit navim'. Wherein, navin represents the nth single navigation behavior, n≥1 and is an integer, 1≤m'≤n and is an integer. Based on this, the first feature pair can include ([navi1, navi2, …, navin], navi1'), …, ([navi1, navi2, …, navin], navim'), …, ([navi1, navi2, …, navin], navin').

[0099] For example, when the value of n is 2, the first feature pair can include ([navi1, navi2], navi1') and ([navi1, navi2], navi2'); that is, for single navigation, it forms a pair with the navigation behavior feature sequence of this navigation and the navigation behavior of the object before this navigation.

[0100] The first feature pair of the navigation behavior, i.e., the pari pair corresponding to the constructed pari pair, is used as a positive sample of a training sample of a representation information acquisition model.

[0101] In some embodiments, the navigation behavior feature sequence in the first feature pair of the navigation behavior and the feature information of the route with the maximum real travel coverage of each navigation behavior of the object to be characterized are as similar as possible, so as to accurately characterize the personalized preference of the object and extract the personalized preference of the object.

[0102] For example, "as similar as possible" herein is distinguished from "as dissimilar as possible" between the navigation behavior feature sequence in the second feature pair of the navigation behavior and the feature information of the route with the maximum real travel coverage of each navigation behavior of the other object in the latter part, which is exemplarily described below.

[0103] In some embodiments, based on the above, Figure 2 S202 can include: generating a second feature pair of each navigation behavior as a negative sample in the training sample; the second feature pair includes the navigation behavior feature sequence up to the last navigation behavior, and the feature information of the route with the maximum real travel coverage of the navigation behavior of the randomly selected other object.

[0104] The navigation behavior of the randomly selected other object can be set to be different from the last navigation behavior, and specifically can be the navigation behavior corresponding to different routes of the same object; or can be set to be different from the current object, and specifically can be the navigation behavior made by the other object in the above except the object to be characterized, so as to distinguish the navigation behavior of the object to be characterized.

[0105] Meanwhile, the feature information of the route with the maximum real travel coverage of the navigation behavior of the other object can indicate the preference of the other object, so as to better distinguish the object to be characterized.

[0106] The second feature pair of each navigation behavior includes the navigation behavior feature sequence up to the last navigation behavior of the object to be characterized, and the feature information of the route with the maximum real travel coverage of the other object.

[0107] For example, taking n=2 as an example, for each pair of positive samples, two routes with the maximum real travel coverage of the other object are randomly selected from the routes with the maximum real travel coverage of the remaining other objects except the object to be characterized, navim' is replaced by the feature information of the route with the maximum real travel coverage of the other object, to form the second feature pair of the navigation behavior. For example, the feature information of the route with the maximum real travel coverage of each navigation behavior of the other object and the historical statistical information of the object navigation behavior are combined to form the pari pair of the negative sample.

[0108] The second feature pair of the navigation behavior, i.e., the corresponding constructed pari pair, is a negative sample of a training sample of a representation information acquisition model.

[0109] In some embodiments, in the second feature pair of the navigation behavior, the feature information of the planning route with the maximum real walking coverage of each navigation behavior of the other object is as different as possible from the navigation behavior feature sequence of the object to be represented, so as to accurately distinguish the object to be represented from the other object and accurately extract the personalized preference of the object.

[0110] By way of example, "as different as possible" herein is distinguished from "as similar as possible" in the first feature pair of the navigation behavior in the foregoing description. "As different as possible" mainly refers to that the routes in the positive sample and the negative sample are different. If multiple negative samples are selected, the routes in the negative samples can be the same or different, which is not limited herein. Specifically, the routes are different in that at least one of the starting point and the ending point of the routes is different. By way of example, in the case that the objects are the same or different, the starting point of one planning route is address A and the ending point is address B, and the starting point of another planning route is address C and the ending point is address D, wherein address A is different from address C, and / or address B is different from address D, so that the feature information of the planning route with the maximum real walking coverage of each navigation behavior of the object can be determined for different planning routes in the positive sample and the negative sample; for another example, other routes of other objects except the object to be represented can be used, for example, the object to be represented is user 1 and the other object is user 2, so that the feature information of the planning route with the maximum real walking coverage of each navigation behavior of the object can be determined for different objects in the positive sample and the negative sample.

[0111] In some embodiments, Figure 3 Another flowchart of a training method of a representation information acquisition model provided by the embodiments of the present disclosure is provided. Based on the foregoing Figure 1 , refer to Figure 3 , the method can include S101, S102, S103 and S104, and specifically, S102 can include S1021 and S1022, S103 can include S1031 and S1032, and S104 can be replaced by S1041.

[0112] S101, obtaining a navigation behavior feature sequence of an object to be represented.

[0113] S1021, obtaining feature information of a planning route with the maximum real walking coverage of each navigation behavior of the object to be represented.

[0114] S1022, obtaining feature information of a planning route with the maximum real walking coverage of each navigation behavior of an object other than the object to be represented.

[0115] The characteristic information of the planning route with the maximum actual travel coverage of each navigation behavior of the other object, in addition to the object to be characterized, can represent the personalized preference of the other object, in addition to the current object, and is used to distinguish from the object to be characterized.

[0116] In combination with the above, since the personalized preferences of different objects are different, the object representation information corresponding thereto is also different. Based on this, the characteristic information of the planning route with the maximum actual travel coverage of the navigation behavior of the other object in this step provides a data basis for the negative samples of the object representation information acquisition model constructed in the subsequent step.

[0117] S1031, generating a first feature pair of each navigation behavior based on the navigation behavior feature sequence of the object to be characterized and the characteristic information of the planning route with the maximum actual travel coverage of each navigation behavior of the object to be characterized, as a positive sample of the training sample.

[0118] S1032, generating a second feature pair of each navigation behavior based on the navigation behavior feature sequence of the object to be characterized and the characteristic information of the planning route with the maximum actual travel coverage of each navigation behavior of the other object, in addition to the object to be characterized, as a negative sample of the training sample.

[0119] S1041, training the representation information acquisition model with the positive sample and the negative sample in the training sample to obtain the trained representation information acquisition model.

[0120] In the step S1031, the first feature pair of the navigation behavior constructed is input into the representation information acquisition model as a positive sample, and the second feature pair of the navigation behavior constructed in the step S1032 is input into the representation information acquisition model as a negative sample, and the output is the corresponding object representation information, so as to train the object representation information acquisition model, facilitate subsequent determination of the object representation information by using the object representation information acquisition model, and accurately extract the personalized preference of the object.

[0121] In other embodiments, the navigation behavior feature sequence obtained in the foregoing steps, the characteristic information of the planning route with the maximum actual travel coverage of each navigation behavior of the object to be characterized, and the characteristic information of the planning route with the maximum actual travel coverage of each navigation behavior of the other object, in addition to the object to be characterized, can be directly used as the training data of the representation information acquisition model, the first feature pair of the navigation behavior and the second feature pair of the navigation behavior are constructed by using the representation information acquisition model, and the training step in the step S1041 is further implemented, which is not limited herein.

[0122] In some embodiments, for each navigation behavior of the object to be characterized, the planning route with the maximum actual travel coverage of the other object in the second feature pair has a large difference from the planning route with the maximum actual travel coverage of the object to be characterized in the first feature pair.

[0123] The planning route with the maximum actual travel coverage of the to-be-characterized object in the first feature pair is the planning route representing the personalized preference of the to-be-characterized object, and the planning route with the maximum actual travel coverage of the other object in the second feature pair is the planning route representing the personalized preference of the other object. Therefore, the difference between the two planning routes in the same navigation behavior is used to represent the difference between the personalized preference of the other object and the personalized preference of the to-be-characterized object. By setting a large difference, the personalized preference of the other object and the personalized preference of the to-be-characterized object can be clearly distinguished, thereby effectively training the object characterization information acquisition model to accurately extract the object characterization information of the to-be-characterized object.

[0124] For example, the difference can be represented in the form of a percentage and has a value between 0 and 100%. When the difference is 0, it means that there is no difference between the two planning routes in the same navigation behavior, that is, they are completely the same. When the difference is 100%, it means that the difference between the two planning routes in the same navigation behavior is the largest, that is, they are completely different. When the difference is greater than 0 and less than 100%, the larger the difference, the greater the difference between the two planning routes in the same navigation behavior. Correspondingly, the difference threshold can be 80%, 90%, or other percentage values, which can be set based on the requirements of the training method of the object characterization information acquisition model, which is not limited here.

[0125] In other embodiments, the difference can also be represented in other forms known to those skilled in the art, which is not limited here.

[0126] In some embodiments, for each navigation behavior of the to-be-characterized object, the ratio of the number of the second feature pair to the first feature pair in the same navigation behavior is N:1, and N is a positive integer greater than or equal to 1.

[0127] In combination with the above, the ratio of the number of negative samples to positive samples for training the object characterization information acquisition model can be 1:1 or greater than 1, that is, the number of negative samples can be more.

[0128] In the negative sample, the planning route with the maximum actual travel coverage of the other object can be the planning route with the maximum actual travel coverage of the same other object, or can be multiple planning routes with the maximum actual travel coverage of different objects corresponding to the same navigation behavior, which is not limited here.

[0129] In this way, more positive samples and negative samples can be used to train the characterization information acquisition model, which is beneficial to improve the training accuracy so as to accurately extract the personalized preference of the object by using the trained characterization information acquisition model in subsequent applications.

[0130] In some embodiments, Figure 4A structural schematic diagram of a representation information obtaining model is provided for an embodiment of the present disclosure. Refer to Figure 4 The representation information obtaining model 10 can include an object preference extraction module 11, a route feature extraction module 12, and a similarity mapping module 13.

[0131] The object preference extraction module 11 is configured to encode the navigation behavior feature sequence of each navigation behavior to obtain object preference encoding. The route feature extraction module 12 is configured to encode the feature information of the route with the maximum actual coverage of each navigation behavior to obtain route feature encoding. The similarity mapping module 13 is trained by the object preference encoding and the route feature encoding to obtain similarity mapping parameters in the similarity mapping module 13.

[0132] Based on this, Figure 5 For Figure 1 The specific flowchart of S104 in the flowchart is shown. Based on Figure 1 and Figure 4 , refer to Figure 5 S104 can include:

[0133] S301, encode the navigation behavior feature sequence of each navigation behavior by the object preference extraction module to obtain object preference encoding.

[0134] The object preference encoding is object representation information, which can be used as the representation of the object and used to represent the tendency of the object, also known as the personalized preference of the object.

[0135] For example, the object preference encoding can be an object vector, which can be represented based on 64 dimensions or represented in other ways, which is not limited herein.

[0136] For example, the object preference extraction module can be implemented by a navigation user personalized deep embedding network (NUPDEN), that is, the navigation behavior feature sequence of each navigation behavior of the object to be represented is encoded by the NUPDEN to obtain the object preference encoding.

[0137] In other embodiments, other neural networks can also be used to implement the object preference extraction module and the route feature extraction module and the similarity mapping module in the representation information obtaining model, which is not limited herein.

[0138] S302, encode the feature information of the route with the maximum actual coverage of each navigation behavior by the route feature extraction module to obtain route feature encoding.

[0139] The route feature code is associated with the feature information of the route with the maximum actual travel coverage of each navigation behavior, and different route feature codes can represent the route with the maximum actual travel coverage of different feature information.

[0140] S303, training the similarity mapping module based on the object preference code and the route feature code to obtain a similarity mapping parameter in the similarity mapping module.

[0141] The similarity mapping parameter is used to represent the association between the object and the planning route, and the similarity between the object preference code and the route feature code can be calculated, and the similarity mapping parameter in the similarity mapping module can be calculated and optimized, that is, the similarity mapping module is trained using the object preference code and the route feature code to obtain the similarity mapping parameter therein.

[0142] After that, the trained similarity mapping module can be used to obtain the corresponding object representation information, that is, the object preference code, based on the input navigation behavior feature sequence of the object to be represented, that is, the personalized preference of the object is extracted.

[0143] In some embodiments, Figure 6 For Figure 5 The specific flowchart of S301 is shown in the flowchart. Based on Figure 5 Referring to Figure 6 S301 can include:

[0144] S401, determining the weight value of each navigation behavior.

[0145] The weight value of each navigation behavior is used to represent the influence degree of the navigation behavior on the object preference code; the greater the weight value, the greater the influence of the navigation behavior on the object preference code.

[0146] In some embodiments, this step can include:

[0147] The weight value of the navigation behavior is determined based on the occurrence time of the navigation behavior; wherein the weight value of the navigation behavior is greater the closer to the time of the last navigation behavior.

[0148] Therefore, the influence of the navigation behavior on the object preference code is greater the closer to the time of the last navigation behavior, and the influence of the navigation behavior on the object preference code is smaller the farther from the time of the last navigation behavior, that is, the object preference code can change over time, so that the representation information acquisition model can capture the change of the personalized preference of the object in the process of continuous updating and iteration, and thus facilitate the planning route to meet the personalized needs of the object changing over time.

[0149] S402, based on the weight value of each navigation behavior, the navigation behavior feature sequence of each navigation behavior is encoded by the object preference extraction module to obtain object preference encoding.

[0150] The object preference extraction module encodes the navigation behavior feature sequence based on the weight value of each navigation behavior to obtain the object preference encoding, so that the object preference encoding more accurately represents the object's personalized preference.

[0151] Therefore, after the navigation behavior feature sequence is input into the representation information acquisition model, the object preference extraction module in the representation information acquisition model will automatically set and select the importance of the navigation behavior feature sequence based on the weight value of each navigation behavior. For example, 3 out of 10 navigation behaviors can express the object's personalized preference, so the weight value of these 3 navigation behaviors will be relatively large. At the same time, the navigation behavior feature sequence that is closer to the last navigation behavior has a greater importance. Based on this, the object preference extraction module encodes the navigation behavior feature sequence of each navigation behavior to obtain the object preference encoding.

[0152] At the same time, the input of the route feature extraction module is the feature information of the planning route with the maximum coverage rate of each navigation behavior, and the encoded result is route feature encoding.

[0153] Finally, the similarity between the object and the route is obtained by calculating the similarity between the object preference encoding and the route feature encoding, and the similarity mapping parameters in the similarity mapping module are calculated and optimized to train the similarity mapping module, that is, to train the representation information acquisition model.

[0154] The training method of the representation information acquisition model provided by the embodiments of the present disclosure can automatically capture the correlation between different navigation behavior feature sequences and the time-decaying correlation coefficient (i.e., the setting of the weight value being larger the closer to the time), and automatically extract the object's personalized preference that changes over time in the process of continuous updating and iteration, thereby achieving accurate acquisition of the object's personalized preference.

[0155] The embodiments of the present disclosure also provide an object representation information acquisition method for acquiring the representation information of an object based on the navigation behavior feature sequence of the object by using the representation information acquisition model trained by any of the above methods to determine the object's preference.

[0156] In some embodiments, Figure 7 A flowchart of an object representation information acquisition method provided by the embodiments of the present disclosure is shown. Referring to Figure 7 The object representation information acquisition method comprises the following steps.

[0157] S111, acquire a navigation behavior feature sequence of the object.

[0158] The navigation behavior feature sequence includes feature information of a planned route of at least two navigation behaviors and feature information of an actual driving route.

[0159] In one navigation behavior, the planned route can be one, two or more, which is determined based on the number of optional routes between the starting point and the destination; the number of the actual driving route is only one, which is the route actually traveled by the object from the starting point to the destination, and can be associated with the personalized preference of the object. The navigation behavior feature sequence includes the feature information of the planned route of at least two navigation behaviors and the feature information of the actual driving route, which provides basic data for determining the representation information of the object by the representation information acquisition model subsequently.

[0160] S112, acquire the representation information of the object by the representation information acquisition model trained by the training method of the representation information acquisition model based on the navigation behavior feature sequence of the object.

[0161] The training method of the representation information acquisition model can be any of the methods in the above embodiments, which is used to train the representation information acquisition model. The trained representation information acquisition model can determine the representation information of the object based on the navigation behavior feature sequence of the object acquired in S111, so as to determine the personalized preference of the object.

[0162] The method for acquiring the representation information of the object provided by the embodiments of the present disclosure can acquire the representation information of the object by the representation information acquisition model trained by any of the above training methods of the representation information acquisition model based on the acquired navigation behavior feature sequence of the object, and has high accuracy.

[0163] In some embodiments, the object is a vehicle driver or a platform object of a network car.

[0164] When the object is a vehicle driver, in the navigation scenario from the starting point to the destination, the navigation software or program can determine the representation information of the vehicle driver, i.e., the personalized preference, such as the tendency to select a planned route with short time, short distance or less navigation actions, etc., based on the navigation behavior feature sequence of the vehicle driver by the representation information acquisition model, and further perform route planning and recommendation based on the acquired representation information of the vehicle driver, so as to meet the needs of the vehicle driver and improve the driving experience.

[0165] When the object is a ride-hailing platform object, after the ride-hailing platform object inputs a departure location and a destination, a route planning software or program can determine the characterization information of the ride-hailing platform object, i.e., the individualized preference of the ride-hailing platform object, such as a preference for a route planning with a short time, a short distance, or a low cost, based on the navigation behavior characteristic sequence of the ride-hailing platform object through the characterization information acquisition model, and further perform route planning and recommendation based on the obtained characterization information of the ride-hailing platform object, so as to meet the demand of the ride-hailing platform object and improve the driving experience.

[0166] In other embodiments, the object can also be a walking object, a cycling object, or an object traveling in other ways, which is not limited herein.

[0167] The object characterization information acquisition method provided by the embodiments of the present disclosure can automatically extract the individualized preference of the object from the navigation behavior characteristic sequence of the object through the characterization information acquisition model, and can also capture the change of the individualized preference of the object over time, so as to accurately extract the individualized preference of the object.

[0168] The embodiments of the present disclosure also provide a route planning method, which determines a planned route based on the characterization information of the object obtained by any of the above object characterization information acquisition methods, so that the planned route has a high degree of agreement with the preference of the object and improves the navigation experience of the object.

[0169] In some embodiments, Figure 8 A flowchart of a route planning method provided by the embodiments of the present disclosure is shown. Referring to Figure 8 The route planning method includes:

[0170] S121, obtaining the characterization information of the object obtained based on the object characterization information acquisition method.

[0171] The characterization information of the object can be automatically determined based on any of the above object characterization information acquisition methods, and the individualized preference of the object is determined in this step.

[0172] For example, the individualized preference of the object can include a preference of the object for not taking a highway, a short time, a small number of navigation actions, a low cost, or a small number of traffic lights.

[0173] S122, determining a planned route recommended to the object based on at least the characterization information of the object, a start and end point selected by the object, and road condition information.

[0174] The road condition information is used to represent real-time road conditions, and can include information related to route planning, such as vehicle congestion, the number of intersections, traffic light conditions, and weather conditions.

[0175] The planning route recommended to the object is a planning route that meets the personalized preferences of the object. In combination with the foregoing S121, in this step, the planning route that meets the personalized preferences of the object is determined based on the start and end points selected by the object, the road condition information, and the representation information of the object, so as to realize personalized route planning for different objects.

[0176] The route planning method provided by the embodiments of the present disclosure can obtain the representation information of the object obtained by any one of the above-mentioned representation information obtaining methods, and determine the planning route recommended to the object in combination with the start and end points selected by the object and the road condition information. The planning route has a high degree of fit with the personalized preferences of the object, is beneficial to meet the personalized needs of the object, and improves the navigation experience of the object.

[0177] The embodiments of the present disclosure also provide a training device of a representation information obtaining model, which can be used to execute the flow steps of any one of the above-mentioned training methods of the representation information obtaining model, and realize the corresponding effects.

[0178] In some embodiments, Figure 9 A structural schematic diagram of a training device of a representation information obtaining model provided by the embodiments of the present disclosure is provided. Referring to Figure 9 The training device 600 of the representation information obtaining model can include:

[0179] A sequence obtaining module 610 is configured to obtain a navigation behavior feature sequence of a to-be-represented object. The navigation behavior feature sequence includes feature information of a planning route of at least two navigation behaviors and feature information of an actual driving route.

[0180] A feature information obtaining module 620 is configured to obtain feature information of a planning route with the maximum actual driving coverage rate for each two navigation behaviors of an object. The object includes the to-be-represented object and other objects except the to-be-represented object.

[0181] A sample generating module 630 is configured to generate a training sample based on the navigation behavior feature sequence of the to-be-represented object and the feature information of the planning route with the maximum actual driving coverage rate for each navigation behavior of the object.

[0182] A training module 640 is configured to train the representation information obtaining model with the training sample, and obtain a trained representation information obtaining model.

[0183] The training apparatus of the representation information obtaining model provided in the embodiments of the present disclosure can generate training samples based on the navigation behavior feature sequence of the object and the feature information of the planning route with the maximum actual walking coverage of each navigation behavior of the object (including the object to be represented and other objects except the object to be represented) for different objects, and the training samples are generated based on the data capable of representing the preferences of the object. Thus, the representation information obtaining model can be trained by using the training samples, and the representation information obtaining model capable of accurately determining the object representation information of the object can be obtained, so that the object representation information can be automatically captured by using the trained representation information obtaining model, and the personalized preferences of the object can be determined. The object representation information can be directly obtained based on the original information of the navigation behavior feature sequence of the object, and there is no problem of limitations of the experience of data analysts on the model training, so that the representation information obtaining model is easy to maintain and the effective use of the navigation behavior feature sequence of the object can be realized.

[0184] In some embodiments, the feature information obtaining module 620 can include at least two sub-modules as follows:

[0185] The first obtaining sub-module is configured to obtain the feature information of the planning route with the maximum actual walking coverage.

[0186] The second obtaining sub-module is configured to obtain the feature information of the planning route ranked first.

[0187] The third obtaining sub-module is configured to obtain the feature information of the planning route with the shortest static time consumption.

[0188] In some embodiments, Figure 10 Another structure diagram of the training apparatus of the representation information obtaining model provided in the embodiments of the present disclosure is provided. Based on Figure 9 , referring to Figure 10 , the sample generation module 630 can include:

[0189] The positive sample generation sub-module 631 is configured to generate the positive sample in the training sample based on the navigation behavior feature sequence of the object to be represented and the feature information of the planning route with the maximum actual walking coverage of each navigation behavior of the object to be represented.

[0190] The negative sample generation sub-module 632 is configured to generate the negative sample in the training sample based on the navigation behavior feature sequence of the object to be represented and the feature information of the planning route with the maximum actual walking coverage of each navigation behavior of the other objects.

[0191] In some embodiments, continuing to refer to Figure 10The positive sample generation submodule 631 in the device 600 is specifically configured to:

[0192] The first feature pair of each navigation behavior is generated as a positive sample in the training sample; the first feature pair includes a navigation behavior feature sequence up to the last navigation behavior, and feature information of a planning route with the maximum actual walking coverage of the last navigation behavior.

[0193] In some embodiments, continuing to refer to Figure 10 The negative sample generation submodule 632 in the device 600 is specifically configured to:

[0194] The second feature pair of each navigation behavior is generated as a negative sample in the training sample; the second feature pair includes a navigation behavior feature sequence up to the last navigation behavior, and feature information of a planning route with the maximum actual walking coverage of a randomly selected navigation behavior of another object.

[0195] In some embodiments, for each navigation behavior of the object to be characterized, the ratio of the number of the second feature pair to the first feature pair is N:1, N being a positive integer greater than or equal to 1.

[0196] In some embodiments, in combination with Figure 4 and Figure 10 The training module 640 can be configured to:

[0197] Encode the navigation behavior feature sequence of each navigation behavior by the object preference extraction module to obtain object preference encoding;

[0198] Encode the feature information of the planning route with the maximum actual walking coverage of each navigation behavior of the object by the route feature extraction module to obtain route feature encoding;

[0199] Train the similarity mapping module based on the object preference encoding and the route feature encoding to obtain similarity mapping parameters in the similarity mapping module.

[0200] In some embodiments, the training module 640 is configured to encode the navigation behavior feature sequence of each navigation behavior by the object preference extraction module to obtain object preference encoding, which can specifically include:

[0201] Determine a weight value of each navigation behavior;

[0202] Encode the navigation behavior feature sequence of each navigation behavior by the object preference extraction module based on the weight value of each navigation behavior to obtain object preference encoding.

[0203] In some embodiments, the training module 640 determines the weight value of each navigation behavior, which can specifically include:

[0204] The weight value of the navigation behavior is determined based on the occurrence time of the navigation behavior, and the weight value of the navigation behavior is greater when the navigation behavior is closer to the last navigation behavior time.

[0205] The training device of the route representation information acquisition model disclosed in the above embodiments can implement the training method of the representation information acquisition model disclosed in the above method embodiments, and has the same or corresponding beneficial effects. To avoid repetition, details are not repeated here.

[0206] The embodiments of the present disclosure also provide an object representation information acquisition device, which can be used to execute the process steps of any one of the above object representation information acquisition methods, and achieve the corresponding effects.

[0207] In some embodiments, Figure 11 A structural schematic diagram of an object representation information acquisition device provided by the embodiments of the present disclosure is provided. Referring to Figure 11 The object representation information acquisition device 700 includes:

[0208] The sequence acquisition module 710 is configured to acquire a navigation behavior feature sequence of the object to be represented, the navigation behavior feature sequence including feature information of a planned route of at least two navigation behaviors and feature information of an actual driving route.

[0209] The representation information acquisition module 720 is configured to acquire the representation information of the object based on the navigation behavior feature sequence of the object to be represented and the representation information acquisition model trained by the training device of any one of the above representation information acquisition models.

[0210] The object representation information acquisition device provided by the embodiments of the present disclosure can acquire the representation information of the object based on the acquired navigation behavior feature sequence of the object and the representation information acquisition model trained by the training device of any one of the above representation information acquisition models through the synergistic effect of the above functional modules, and has high accuracy.

[0211] The embodiments of the present disclosure also provide a route planning device, which can be used to execute any one of the above route planning methods and achieve the corresponding effects.

[0212] In some embodiments, Figure 12 A structural schematic diagram of a route planning device provided by the embodiments of the present disclosure is provided. Referring to Figure 12 The route planning device 800 includes:

[0213] The representation information acquisition module 810 is configured to acquire the representation information of the object obtained based on any one of the above object representation information acquisition devices.

[0214] The route planning module 820 is configured to determine a planned route recommended to the object based on at least the representation information of the object, the start and end points selected by the object, and road condition information.

[0215] The route planning device provided by the embodiments of the present disclosure can obtain the characterization information of the object obtained by the above-mentioned any object characterization information obtaining device, and determine a recommended planning route for the object in combination with the start and end points selected by the object and the road condition information. The planning route has a higher degree of fit with the personalized preferences of the object, is beneficial to meet the personalized needs of the object, and improves the navigation experience of the object.

[0216] The embodiments of the present disclosure also provide a map navigation system, which comprises the above-mentioned any route planning device.

[0217] For example, the object can be a vehicle driver. The route planning device can determine a planning route that meets the personalized preferences of the vehicle driver based on the characterization information of the vehicle driver and the current road condition information. The map navigation system can further comprise a display module configured to display the planning route to the vehicle driver for viewing. The map navigation system can further comprise a voice module configured to play a navigation prompt audio based on the planning route and the real-time position of the vehicle driver to prompt the vehicle driver to perform a navigation action.

[0218] In other embodiments, the map navigation system can further comprise other functional modules known to those skilled in the art, which are not described or limited herein.

[0219] The map navigation system provided by the embodiments of the present disclosure can determine a planning route that meets the personalized preferences of the object based on the above-mentioned any route planning device, and further implement navigation.

[0220] The embodiments of the present disclosure also provide a network car-hailing platform system, which comprises the above-mentioned any route planning device.

[0221] For example, the object can be a network car-hailing platform object. The route planning device can determine a planning route that meets the personalized preferences of the network car-hailing platform object based on the characterization information of the network car-hailing platform object and the current road condition information. The network car-hailing platform system can further comprise a display module configured to display the planning route to the network car-hailing platform object for confirmation.

[0222] In other embodiments, the network car-hailing platform system can further comprise other functional modules known to those skilled in the art, which are not described or limited herein.

[0223] Figure 13This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. The training method, object representation information acquisition method or route planning method for acquiring representation information described above in the embodiments of the present disclosure can be implemented on a cloud server or a local host. Therefore, the electronic device can be a cloud server, a local host, or a client device or terminal device as described in the above embodiments.

[0224] The following detailed reference illustrates a structural diagram suitable for implementing the electronic device 500 in the embodiments of this disclosure. The electronic device 500 in the embodiments of this disclosure may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. The illustrated electronic device is merely an example and should not impose any limitation on the functionality or scope of the embodiments of this disclosure.

[0225] like Figure 13 As shown, the electronic device 500 may include a processor (e.g., a central processing unit, a graphics processing unit, etc., also referred to as a processing device) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage device 508 into a random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the electronic device 500. The processor 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0226] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 508 including, for example, magnetic tapes, hard disks, etc.; and communication devices 509. Communication device 509 allows electronic device 500 to communicate wirelessly or wiredly with other devices to exchange data. Although an electronic device 500 with various devices is shown, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0227] In particular, according to embodiments of the present disclosure, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, embodiments of the present disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for executing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network through the communication apparatus 509, or installed from the storage apparatus 508, or installed from the ROM 502. When the computer program is executed by the processing apparatus 501, the above-mentioned processes defined in the training method of a feature information acquisition model, the acquisition method of object feature information, or the route planning method of embodiments of the present disclosure are executed, realizing the corresponding functions.

[0228] It should be noted that the computer-readable medium described above in the present disclosure can be a computer-readable signal medium or a computer-readable storage medium or any combination thereof. The computer-readable storage medium may, for example, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any combination thereof. More specific examples of the computer-readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, the computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus or device. In the present disclosure, the computer-readable signal medium can include a data signal carried in a baseband or as a part of a carrier wave, which carries computer-readable program code. Such a propagated data signal can take many forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium other than the computer-readable storage medium that can send, propagate or transmit the program for use by or in connection with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium can be transmitted by any suitable medium, including but not limited to a wire, an optical fiber, an RF (radio frequency) or the like, or any suitable combination thereof.

[0229] In some embodiments, the client and the server can communicate using any currently known or future developed network protocol, such as HTTP (HyperText Transfer Protocol), and can be interconnected with digital data communication (e.g., a communication network) of any form or medium. Examples of communication networks include local area networks ("LAN"), wide area networks ("WAN"), internetworks (e.g., the Internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future developed networks.

[0230] The computer readable medium described above can be included in the electronic device described above; or can exist separately, without being assembled into the electronic device.

[0231] The computer readable medium described above carries one or more programs, which, when executed by the electronic device, cause the electronic device to:

[0232] Obtain a navigation behavior feature sequence of the object to be characterized, the navigation behavior feature sequence including feature information of a planned route of at least two navigation behaviors, and feature information of an actual driving route;

[0233] Obtain feature information of a planned route with the maximum actual travel coverage for each navigation behavior of the object, the object including the object to be characterized and other objects than the object to be characterized;

[0234] Generate a training sample based on the navigation behavior feature sequence of the object to be characterized, and the feature information of the planned route with the maximum actual travel coverage for each navigation behavior of the object;

[0235] Train the characterization information acquisition model with the training sample to obtain a trained characterization information acquisition model.

[0236] Alternatively, the computer readable medium described above carries one or more programs, which, when executed by the electronic device, cause the electronic device to:

[0237] Obtain a navigation behavior feature sequence of the object, the navigation behavior feature sequence including feature information of a planned route of at least two navigation behaviors, and feature information of an actual driving route;

[0238] Obtain characterization information of the object based on the navigation behavior feature sequence of the object, by the characterization information acquisition model trained by any of the above methods.

[0239] Alternatively, the computer readable medium described above carries one or more programs, which, when executed by the electronic device, cause the electronic device to:

[0240] obtaining the object representation information of the object based on the object representation information acquisition method;

[0241] determining a recommended route to the object based on at least the object representation information, the object selection start and end points, and the road condition information.

[0242] Alternatively, the computer readable medium carries one or more programs which, when executed by the electronic device, cause the electronic device to perform the training method of the representation information acquisition model, the object representation information acquisition method, or the route planning method. To avoid repetition, details are not described here.

[0243] Wherein, the computer program code for performing the operations of the present disclosure can be written in one or more programming languages or combinations thereof, including but not limited to object-oriented programming languages such as Java, Smalltalk, C++, and conventional procedural programming languages such as "C" language or similar programming languages. Program code can be executed entirely on the object computer, partially on the object computer, as an independent software package, partially on the object computer and partially on a remote computer, or entirely on a remote computer or server. In the case of remote computers, the remote computer can be connected to the object computer through any kind of network, including local area network (LAN) or wide area network (WAN), or can be connected to external computers (for example, through the Internet by using an Internet service provider).

[0244] The flowcharts and block diagrams in the drawings illustrate the possible implementation architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram can represent a module, program segment, or a part of code containing one or more executable instructions for implementing the specified logic function. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in different order than that noted in the drawings. For example, two blocks represented in succession can actually be executed substantially in parallel, and sometimes in reverse order, depending on the function involved. It should also be noted that each block in the block diagram and / or flowchart, and the combination of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a combination of dedicated hardware and computer instructions.

[0245] The units described in the embodiments of the present disclosure can be implemented in the form of software, or can be implemented in the form of hardware. In some cases, the names of units, modules and sub-modules do not constitute a limitation on the units, modules and sub-modules themselves.

[0246] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, and without limitation, example types of hardware logic components that can be used include Field-programmable Gate Arrays (FPGAs), Application-specific Integrated Circuits (ASICs), Application-specific Standard Products (ASSPs), System-on-a-chip systems (SOCs), Complex Programmable Logic Devices (CPLDs), etc.

[0247] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0248] According to one or more embodiments of the present disclosure, the present disclosure provides a computer-readable storage medium, which stores a computer program. The computer program is configured to execute any one of the training methods of the feature information acquisition model provided by the embodiments of the present disclosure, or is configured to execute any one of the methods of acquiring object feature information provided by the embodiments of the present disclosure, or is configured to execute any one of the route planning methods provided by the embodiments of the present disclosure.

[0249] The above description is merely preferred embodiments of the present disclosure and a description of principles of applied technologies. Those skilled in the art should understand that the disclosure range of the present disclosure is not limited to the technical solutions formed by the specific combinations of the above technical features, and should also cover other technical solutions formed by any combinations of the above technical features or equivalent features without departing from the above disclosed concept. For example, the above technical features can be replaced with technical features disclosed in the present disclosure (but not limited to) having similar functions to form technical solutions.

[0250] Moreover, while operations are depicted in a particular order, this should not be understood as requiring such an order nor infringing on the scope of the disclosure. Certain of the operations described in the discussion are combinable into a single operation, and certain operations can be separated into several operations. In some embodiments, the operations described in the discussion can be performed in an order different than presented in the discussion. In some embodiments, the operations described in the discussion can be performed concurrently. Also, while several specific implementation details are discussed in the discussion, these should not be interpreted as limiting the scope of the disclosure. Rather, certain features described in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable sub-combination.

[0251] Although the subject matter has been described in language specific to structural features and / or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as example forms of implementing the claims.

Claims

1. A training method of a representation information acquisition model, comprising: obtaining a navigation behavior feature sequence of a to-be-represented object, the navigation behavior feature sequence comprising feature information of a planned route of at least two navigation behaviors and feature information of an actually traveled route; obtaining feature information of a planned route with a maximum actual travel coverage of each navigation behavior of an object, the object comprising the to-be-represented object and other objects except the to-be-represented object; generating a training sample based on the navigation behavior feature sequence of the to-be-represented object and the feature information of the planned route with the maximum actual travel coverage of each navigation behavior of the object, wherein the training sample comprises positive samples and negative samples, the training sample generated by using the feature information of the planned route with the maximum actual travel coverage of each navigation behavior of the to-be-represented object and the navigation behavior feature sequence of the to-be-represented object is a positive sample, and the training sample generated by using the feature information of the planned route with the maximum actual travel coverage of each navigation behavior of other objects and the navigation behavior feature sequence of the to-be-represented object is a negative sample; training the representation information acquisition model by using the training sample to obtain a trained representation information acquisition model.

2. The method of claim 1, wherein, The feature information of the planned route of the at least two navigation behaviors comprises at least two of: feature information of a planned route with a maximum actual travel coverage; feature information of a planned route ranked first; feature information of a planned route with a shortest static time consumption.

3. The method of claim 1, wherein, The generating of the training sample based on the navigation behavior feature sequence of the to-be-represented object and the feature information of the planned route with the maximum actual travel coverage of each navigation behavior of the object comprises: generating a positive sample in the training sample based on the navigation behavior feature sequence of the to-be-represented object and the feature information of the planned route with the maximum actual travel coverage of each navigation behavior of the to-be-represented object; and generating a negative sample in the training sample based on the navigation behavior feature sequence of the to-be-represented object and the feature information of the planned route with the maximum actual travel coverage of each navigation behavior of other objects.

4. The method of claim 3, wherein, The generating of the positive sample in the training sample based on the navigation behavior feature sequence of the to-be-represented object and the feature information of the planned route with the maximum actual travel coverage of each navigation behavior of the to-be-represented object comprises: generating a first feature pair of each navigation behavior as the positive sample in the training sample, the first feature pair comprising a navigation behavior feature sequence up to a last navigation behavior and the feature information of the planned route with the maximum actual travel coverage of the last navigation behavior.

5. The method of claim 4, wherein, The generating of the negative sample in the training sample based on the navigation behavior feature sequence of the to-be-represented object and the feature information of the planned route with the maximum actual travel coverage of each navigation behavior of other objects comprises: generating a second feature pair of each navigation behavior as the negative sample in the training sample, the second feature pair comprising the navigation behavior feature sequence up to the last navigation behavior and the feature information of the planned route with the maximum actual travel coverage of a navigation behavior of a randomly selected other object.

6. The method of claim 4, wherein, For each navigation behavior of the object to be characterized, a ratio of a number of second feature pairs to a number of first feature pairs is N:1, N being a positive integer greater than or equal to 1.

7. The method of claim 1, wherein, The characterization information acquisition model comprises an object preference extraction module, a route feature extraction module, and a similarity mapping module. The training of the characterization information acquisition model with the training samples comprises: The object preference extraction module encodes the navigation behavior feature sequence of each navigation behavior to obtain object preference encoding. The route feature extraction module encodes the feature information of the planning route with the maximum actual coverage rate of each navigation behavior to obtain route feature encoding. The similarity mapping module is trained based on the object preference encoding and the route feature encoding to obtain similarity mapping parameters in the similarity mapping module.

8. The method of claim 7, wherein, The object preference extraction module encodes the navigation behavior feature sequence of each navigation behavior to obtain object preference encoding, comprising: Based on the occurrence time of the navigation behavior, a weight value of each navigation behavior is determined; wherein the weight value of the navigation behavior is greater the closer to the time of the last navigation behavior; Based on the weight value of each navigation behavior, the object preference extraction module encodes the navigation behavior feature sequence of each navigation behavior to obtain object preference encoding.

9. An object characterization information acquisition method, comprising: Obtaining a navigation behavior feature sequence of an object, the navigation behavior feature sequence comprising feature information of a planning route of at least two navigation behaviors, and feature information of an actual driving route; Based on the navigation behavior feature sequence of the object, a characterization information acquisition model trained by the method of any one of claims 1-8 is used to obtain characterization information of the object.

10. A route planning method, comprising: Obtaining characterization information of an object based on the method of claim 9; Based on at least the characterization information of the object, a start and end point selected by the object, and road condition information, a planning route recommended to the object is determined.

11. A computer program product for executing the training method of the characterization information acquisition model of any one of claims 1-8, the object characterization information acquisition method of claim 9, or the route planning method of claim 10.

Images