CN112650825A

CN112650825A - Method and device for determining abnormal drive receiving behavior, storage medium and electronic equipment

Info

Publication number: CN112650825A
Application number: CN202011615630.1A
Authority: CN
Inventors: 陈君
Original assignee: Beijing Didi Infinity Technology and Development Co Ltd
Current assignee: Beijing Didi Infinity Technology and Development Co Ltd
Priority date: 2020-12-30
Filing date: 2020-12-30
Publication date: 2021-04-13
Anticipated expiration: 2040-12-30
Also published as: CN112650825B

Abstract

The application provides a method and a device for determining abnormal pickup behavior, a storage medium and electronic equipment, wherein the method comprises the following steps: acquiring track information corresponding to the cancelled target travel order; determining a first abnormal stopping probability of the target driver when the target driver stops abnormally based on a difference judgment rule between the current driving position and a preset driving position; if the first abnormal stay probability is larger than a preset probability threshold, determining whether the target driver has abnormal driving receiving behavior or not based on the first abnormal stay probability; and if the first abnormal stay probability is not greater than the preset probability threshold, determining whether the target driver has abnormal driving receiving behavior according to a second abnormal stay probability output by a pre-trained abnormal stay recognition model. Therefore, the abnormal driving receiving behavior of the target driver can be comprehensively judged under different conditions according to the difference judgment rule between the current driving position and the preset driving position and the abnormal stay recognition model, and the accuracy of recognizing the abnormal driving receiving behavior of the target driver is improved.

Description

Method and device for determining abnormal drive receiving behavior, storage medium and electronic equipment

Technical Field

The application relates to the technical field of network appointment vehicles, in particular to a method and a device for determining abnormal driving receiving behaviors, a storage medium and electronic equipment.

Background

With the rapid development of the internet technology, the travel service based on the internet technology brings more and more convenience for people to travel, for example, a user can travel by bus through a network car booking service system. When the network appointment car service system is used at present, after a user submits a travel order, the travel order is cancelled when the travel order is not normally completed, so that an incomplete abnormal order is generated, and the analysis of the cancellation reason of the abnormal order has a good guiding effect on the cancellation of the avoiding order when the order is dispatched next time.

At the present stage, most of the reasons for canceling the order are judged by inputting dimension information of the target driver into a trained recognition model for recognition, but the judgment of the model is generally related to the attributes of the training sample, the actual travel scene is complex and changeable, and the sample may not accurately and completely cover all scenes, so that the recognition accuracy of the abnormal driving receiving behavior of the target driver through the recognition model is low, and how to improve the recognition accuracy of the abnormal driving receiving behavior of the driver in the driving receiving process is a problem to be solved urgently.

Disclosure of Invention

In view of this, an object of the present application is to provide a method, an apparatus, a storage medium, and an electronic device for determining an abnormal pickup behavior, which can perform comprehensive judgment on the abnormal pickup behavior of a target driver under different conditions according to driving positions of a pickup target driver at different times in trajectory information and driving time corresponding to each driving position, according to a difference judgment rule between a current driving position and a preset driving position and an abnormal stay recognition model, and are helpful to improve accuracy of recognition of the abnormal pickup behavior of the target driver.

The embodiment of the application provides a method for determining abnormal driving receiving behaviors, which comprises the following steps:

obtaining order information of the cancelled target trip order; the order information comprises the driving track information of a target driver carrying a target trip order; the driving track information comprises driving positions of a target driver at different time after receiving a target trip order and driving time corresponding to each driving position;

determining a first abnormal stopping probability that the target driver stops abnormally based on a difference judgment rule between a preset current driving position and a preset driving position and the driving track information of the target driver;

detecting whether the first abnormal stay probability is greater than a preset probability threshold;

if the first abnormal stay probability is larger than a preset probability threshold, determining whether the target driver has abnormal driving receiving behaviors or not based on the first abnormal stay probability;

and if the first abnormal stopping probability is not greater than a preset probability threshold, inputting the driving track information into a pre-trained abnormal stopping recognition model to determine a second abnormal stopping probability of the target driver when the target driver abnormally stops in the order receiving process, and determining whether the target driver abnormally receives driving behaviors or not based on the second abnormal stopping probability.

Further, inputting the driving track information into a pre-trained abnormal stay recognition model to determine a second abnormal stay probability that the target driver abnormally stays under the non-extreme condition in the order taking process, including:

inputting the running track information into a pre-trained abnormal stay recognition model to determine a first sub-abnormal stay probability of the target driver for abnormal stay under the non-extreme condition in the order taking process; the abnormal stay recognition model is obtained by training different historical driving positions in the sample target driver historical driving track information in the non-extreme case history cancelled travel order and the historical driving time corresponding to each historical driving position;

determining a second sub-abnormal stay probability of abnormal driving of the target driver based on a preset driving track motion rule of the target driver in the driving process under the extreme condition and the driving track information of the target driver;

and determining a second abnormal stopping probability of the abnormal stopping of the target driver in the order taking process based on the first sub abnormal stopping probability and the second sub abnormal stopping probability.

Further, the extreme case is a case where at least one of the following conditions is present:

the time difference value between the time when the travel order is cancelled and the order receiving time is smaller than a preset time threshold value;

the distance between the order receiving position of the travel order and the boarding position of the passenger is smaller than a preset distance threshold value.

Further, the determining a second sub-abnormal stay probability that the target driver has abnormal driving based on the preset driving track motion rule of the target driver in the driving process under the extreme condition and the driving track information of the target driver includes:

determining at least one abnormal driving position with abnormal driving at a plurality of driving positions included in the driving track; the abnormal running position is a position with a running speed smaller than a preset running speed or a position with a deviation planning navigation route;

and determining a second sub-abnormal stopping probability that the target driver has abnormal driving based on the determined proportion of the at least one abnormal driving position in the plurality of driving positions.

Further, the inputting the information of the driving track into a pre-trained abnormal stay recognition model to determine a first sub-abnormal stay probability of the abnormal stay of the target driver in the order taking process includes:

determining a plurality of driving positions included in the driving track information and driving time corresponding to each driving position;

inputting each driving position and corresponding driving time into a pre-trained abnormal stay recognition model, and determining at least one abnormal driving position;

and determining a first sub-abnormal stopping probability that the target driver has abnormal driving based on the determined proportion of the at least one abnormal driving position in the plurality of driving positions.

Further, the second sub-abnormal stay probability that the target driver has abnormal driving is determined based on the preset driving track motion rule of the target driver in the driving process under the extreme condition and the driving track information of the target driver:

determining an updated mapping relation between a driving position and driving time in a driving track motion rule of a target driver based on a travel scene corresponding to a target travel order; wherein the travel scene comprises at least one condition affecting a target driver driving trajectory;

for each driving position, if the mapping relation between the driving position and the corresponding driving time is inconsistent with the updated mapping relation, determining the driving position as an abnormal driving position;

Further, the cancellation party of the target travel order is a passenger;

the target driver's travel track information is the target driver's travel track information in the time period from the time when the target trip order is placed to the time when the order is cancelled.

Further, the determining method further includes:

acquiring a plurality of historical cancelled travel orders and order attribute information corresponding to each historical cancelled travel order; the order attribute information comprises at least one of order cancellation time between the time when the travel order is cancelled and the order taking time and driving receiving distance between the order taking position of the travel order and the boarding position of the passenger;

determining at least one abnormal order with order cancellation time smaller than preset time and at least one abnormal order with order cancellation distance smaller than a preset distance threshold value for the target driver driving distance from the obtained plurality of historical cancelled travel orders;

filtering out at least one order cancellation time and at least one cancellation time abnormal order from the plurality of historical cancelled travel orders, and determining a plurality of sample cancelled travel orders;

and training and constructing a deep learning network based on different historical driving positions in the historical driving track information corresponding to the cancelled travel order of each sample and the historical driving time corresponding to each historical driving position to obtain an abnormal stay recognition model.

Further, the determining a first abnormal stopping probability that the target driver stops abnormally based on a difference judgment rule between a preset current driving position and a preset driving position and the driving track information of the target driver includes:

determining the running position as an abnormal running position if the distance difference between the current running position and the preset running position is greater than a preset difference threshold value in the running time corresponding to each running position;

and determining a first abnormal stop probability that the target driver has abnormal driving based on the determined proportion of the at least one abnormal driving position in the plurality of driving positions.

The embodiment of the application provides a device for determining abnormal driving receiving behaviors, which comprises:

the order information acquisition module is used for acquiring order information of cancelled target trip orders; the order information comprises the driving track information of a target driver carrying a target trip order; the driving track information comprises driving positions of a target driver at different time after receiving a target trip order and driving time corresponding to each driving position;

the first probability determination module is used for determining a first abnormal stopping probability that the target driver stops abnormally based on a difference judgment rule between a preset current driving position and a preset driving position and the driving track information of the target driver;

the probability detection module is used for detecting whether the first abnormal stay probability is greater than a preset probability threshold value;

the first abnormal behavior determination module is used for determining whether the target driver has abnormal driving receiving behaviors or not based on the first abnormal stopping probability if the first abnormal stopping probability is larger than a preset probability threshold;

and the second abnormal behavior determining module is used for inputting the running track information into a pre-trained abnormal stay recognition model if the first abnormal stay probability is not greater than a preset probability threshold so as to determine a second abnormal stay probability of the target driver when the target driver abnormally stays in the order receiving process, and determining whether the target driver abnormally receives the driving behavior based on the second abnormal stay probability.

An embodiment of the present application further provides an electronic device, including: the electronic equipment comprises a processor, a memory and a bus, wherein the memory stores machine readable instructions executable by the processor, the processor and the memory are communicated through the bus when the electronic equipment runs, and the machine readable instructions are executed by the processor to execute the steps of the method for determining the abnormal drive receiving behavior.

The embodiment of the present application further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the steps of the method for determining abnormal drive receiving behavior as described above are performed.

According to the method, the device, the storage medium and the electronic equipment for determining the abnormal pickup behavior, the order information of the cancelled target trip order including the driving track information of the target driver carrying the target trip order is obtained; determining a first abnormal stopping probability of the target driver when the target driver abnormally stops according to a preset difference judgment rule between the current driving position and a preset driving position, detecting whether the first abnormal stopping probability is greater than a preset probability threshold, and determining whether the target driver abnormally connects the driving according to the first abnormal stopping probability if the first abnormal stopping probability is greater than the preset probability threshold; if the first abnormal stopping probability is not greater than the preset probability threshold, the driving track information is input into a pre-trained abnormal stopping recognition model to determine a second abnormal stopping probability, whether the target driver has abnormal driving receiving behavior is determined according to the second abnormal stopping probability, the abnormal driving receiving behavior of the target driver can be comprehensively determined under different conditions according to a difference judgment rule between the current driving position and the preset driving position and the abnormal stopping recognition model, and the accuracy of recognizing the abnormal driving receiving behavior of the target driver is improved.

Furthermore, in the processing process of the abnormal stay recognition model, a first sub abnormal stay probability of abnormal stay of a target driver in the process of receiving driving under non-extreme conditions is determined in advance through the abnormal stay recognition model, a second sub abnormal stay probability of abnormal stay of the target driver in the process of receiving driving is determined according to the driving track motion rule of the target driver in the driving process under extreme conditions, the abnormal stay probability output by the abnormal recognition model is determined for the target driver by combining the first sub abnormal stay probability and the second sub abnormal stay probability, whether abnormal receiving driving behaviors occur or not is judged according to the abnormal stay probability output by the abnormal recognition model, and the condition that all conditions cannot be completely covered by the abnormal stay recognition model and misjudgment is further generated can be compensated by combining the judgment under non-extreme conditions and the judgment of the rule under extreme conditions, the method is beneficial to improving the accuracy of identifying the abnormal driving receiving behavior of the target driver.

In order to make the aforementioned objects, features and advantages of the present application more comprehensible, preferred embodiments accompanied with figures are described in detail below.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are required to be used in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and therefore should not be considered as limiting the scope, and for those skilled in the art, other related drawings can be obtained from the drawings without inventive effort.

Fig. 1 is a flowchart of a method for determining an abnormal pickup behavior according to an embodiment of the present disclosure;

fig. 2 is a flowchart of another method for determining abnormal pickup behavior according to an embodiment of the present disclosure;

fig. 3 is a schematic structural diagram of an apparatus for determining an abnormal pickup behavior according to an embodiment of the present disclosure;

fig. 4 is a second schematic structural diagram of an apparatus for determining an abnormal pickup behavior according to an embodiment of the present application;

fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present application.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all the embodiments. The components of the embodiments of the present application, generally described and illustrated in the figures herein, can be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present application, presented in the accompanying drawings, is not intended to limit the scope of the claimed application, but is merely representative of selected embodiments of the application. Every other embodiment that can be obtained by a person skilled in the art without making creative efforts based on the embodiments of the present application falls within the protection scope of the present application.

It should be noted that in the embodiments of the present application, the term "comprising" is used to indicate the presence of the features stated hereinafter, but does not exclude the addition of further features.

The terms "passenger," "requestor," "service requestor," and "customer" are used interchangeably in this application to refer to an individual, entity, or tool that can request or order a service. The terms "target driver," "provider," "service provider," and "provider" are used interchangeably in this application to refer to an individual, entity, or tool that can provide a service. The term "user" in this application may refer to an individual, entity or tool that requests a service, subscribes to a service, provides a service, or facilitates the provision of a service. For example, the user may be a passenger, a driver, an operator, etc., or any combination thereof. In the present application, "passenger" and "passenger terminal" may be used interchangeably, and "driver" and "driver terminal" may be used interchangeably.

The terms "service request" and "order" are used interchangeably herein to refer to a request initiated by a passenger, a service requester, a driver, a service provider, or a supplier, the like, or any combination thereof. Accepting the "service request" or "order" may be a passenger, a service requester, a driver, a service provider, a supplier, or the like, or any combination thereof. The service request may be charged or free.

It is worth noting that before the application is provided in the present application, the abnormal pickup behavior of the target driver side is generally determined by setting a plurality of determination rules according to different trip scenes, because a plurality of trip scenes exist in the trip process, and the trip problems encountered in each trip scene are different, different determination rules need to be set for different trip scenes and different trip problems, for example, different rules need to be set for different road congestion situations corresponding to different trip times for determination, rule 1: under the condition that a road is not congested, the target driver stays in the same position for too long time, which is an abnormal driving behavior; rule 2: under the condition of road congestion, the target driver stays at the congestion position for a longer time, the congestion condition of the road is referred, the target driver is determined to normally stay at the congestion position for a long time, and therefore the number of judgment rules which need to be set aiming at different travel scenes is large, when the abnormal driving receiving behavior of the target driver is judged subsequently, a large number of judgment rules need to be matched one by one, the problem of inaccurate matching is easy to occur in the matching process, and the efficiency and the accuracy of judging the abnormal driving receiving behavior of the target driver through the large number of judgment rules are low;

with the development of the neural network deep learning technology, the identification of the abnormal pickup behavior of the target driver side may also be performed by using a pre-trained identification model, but the pre-trained identification model learns data characteristics in a large amount of sample data to provide a corresponding judgment result on a statistical level, the identification result of the identification model depends on the characteristics of the sample data, for the accuracy of the model in the model training process, the sample data under some extreme scenes (for example, a scene in which the existing time of a travel order is short and the driving behavior of the target driver cannot be estimated) which may affect the accuracy of the model is not included in the model training process, and due to the complexity of the travel scene, the identification model may have misjudgment on the pickup data under some extreme scenes in the specific application process, the accuracy rate of judging the abnormal driving receiving behavior of the target driver is low.

Based on this, the embodiment of the application provides a method for determining abnormal pickup behavior, which can comprehensively determine the abnormal pickup behavior of a target driver under different conditions according to a difference determination rule between a current driving position and a preset driving position and an abnormal stay recognition model, and is beneficial to improving the accuracy of recognizing the abnormal pickup behavior of the target driver.

Referring to fig. 1, fig. 1 is a flowchart illustrating a method for determining an abnormal pickup behavior according to an embodiment of the present disclosure. As shown in fig. 1, a method for determining an abnormal pickup behavior provided in an embodiment of the present application includes:

s101, obtaining order information of a cancelled target trip order; the order information comprises the driving track information of a target driver carrying a target trip order; the driving track information comprises driving positions of the target driver at different time after receiving the target trip order and driving time corresponding to each driving position.

S102, determining a first abnormal stopping probability that the target driver stops abnormally based on a difference judgment rule between a preset current driving position and a preset driving position and the driving track information of the target driver.

S103, detecting whether the first abnormal stay probability is larger than a preset probability threshold value.

And S104, if the first abnormal stopping probability is larger than a preset probability threshold, determining whether the target driver has abnormal driving receiving behavior or not based on the first abnormal stopping probability.

And S105, if the first abnormal stay probability is not greater than a preset probability threshold, inputting the running track information into a pre-trained abnormal stay recognition model to determine a second abnormal stay probability of the target driver when the target driver abnormally stays in the order receiving process, and determining whether the target driver abnormally receives the driving behavior or not based on the second abnormal stay probability.

In step S101, the cancelled target order may be an order in which the passenger issues a cancellation request or an order in which the target driver issues a cancellation request. When the order is determined to be cancelled, the driving positions and the driving time corresponding to each driving position at different times after the target trip order is received during the order cancellation period need to be obtained.

Here, the cancellation party of the target trip order may be a target driver or a passenger who places an order, and when the cancellation party is a target driver end, there may be a case that the target driver finds that the passenger cannot reach the passenger point after receiving the order, and the cancellation from the target driver end is generally not order cancellation performed by the target driver within a long time period, so that the judgment of the abnormal stopping behavior of the target driver is generally not analyzed from the cancellation from the target driver end; in consideration of the perception of the passenger to the service and the objective analysis of the abnormal stopping of the target driver, the method mainly analyzes the abnormal driving receiving behavior of the target driver under the condition that the passenger cancels the order subjectively.

Here, the travel track information of the target driver is the travel track information of the target driver in a time period from a time when the order is placed to a time when the order is cancelled.

The travel order comprises the following five time nodes which sequentially appear from order generation to order cancellation: an order placing node (a time node at which a passenger submits an order request including a travel destination and a travel origin), an order taking node (a time scene at which a target driver receives a target travel order and determines the order taking), an order canceling node (an order cancelled by the passenger through a handheld intelligent terminal, an event scene at which the passenger completes the order), an order completing node (a time node at which the target driver sends the passenger to the travel destination in the order request made by the passenger after the order taking and the passenger completes normal completion of a payment order) and an order worksheet responsible node (a time node at which a responsible party (the target driver or the passenger) has an exception after giving responsibility confirmation to an abnormally completed order platform), nodes at which the order is cancelled and an abnormal dispute exists may generally appear at the order canceling node and the order worksheet responsible node according to the difference of time intervals between each node and the order taking node, the information of the driving track of the target driver acquired at each node is also different.

The driving track information comprises driving positions of the target driver at different time after receiving the target trip order and driving time corresponding to each driving position, and the driving progress of the target driver in the order execution time can be reflected, so that whether abnormal behaviors of abnormal stopping exist in the order receiving process of the target driver or not is determined.

The driving positions of different time in the driving process of the target driver and the driving time corresponding to each driving position are a space-time characteristic, the driving positions of the target driver in a certain driving time are indicated, and the driving condition of the target driver can be determined through the space-time characteristic.

Wherein, the driving condition comprises normal driving, abnormal driving and the like; the normal driving condition comprises that the driving speed of the target driver in the driving receiving section is consistent with the historical driving speed of the target driver in the similar historical driving receiving section, and the target driver does not wander at a certain position or positions within a certain distance from the target driver to the boarding point of the passenger after the order is received; the abnormal driving condition comprises the condition that the target driver wanders at one or a plurality of positions within a certain distance from the target driver to the boarding point of the passenger after the order is picked up; or the system plans the target driver with a pick-up route from the pick-up position of the target driver to the boarding position of the passenger, and the target driver does not drive according to the pick-up route (yaw driving). The situation that the target driver is in normal driving or abnormal driving can be determined directly according to the driving behavior of the target driver in the follow-up process through the definition of normal driving and abnormal driving, and therefore the abnormal driving receiving behavior of the target driver is analyzed more accurately.

Here, the driving position may be an absolute position during driving, for example, a position identified by coordinates on a map or a position area divided by different street areas; a plurality of driving positions determined by the distance between the current position and the order taking position of the target driver can be further adopted, for example, the position 100 meters away from the order taking position, the position 500 meters away from the order taking position and the like; there may also be a plurality of travel positions determined by the distance between the current position and the boarding position of the passenger, for example, a position 100 meters from the boarding position, a position 500 meters from the boarding position, and the like.

Here, step S101 obtains the order information of the order cancelled floor when the order is cancelled, and further obtains the driving track information of the order pickup target driver from the order information, and determines the probability that the target driver has abnormal driving conditions (such as abnormal stay) during the process from order pickup to order cancellation by analyzing the driving positions of the target driver indicated in the driving track information at different times and the driving time corresponding to each driving position, so as to determine the abnormal stay condition of the target driver, so that the platform determines the responsibility of the target driver according to the abnormal stay condition of the target driver when the order is cancelled.

In step S102, according to the difference determination rule between the current driving position and the preset driving position, a mapping relationship between a difference between the current driving position and the preset driving position and the abnormal stopping behavior may be determined, so as to determine a first abnormal stopping probability that the target driver has abnormally stopped.

Here, for different travel scenes, it is possible to appropriately adjust the gap determination rule between the current driving position and the preset driving position to improve the accuracy of determining abnormal driving receiving of the target driver.

Here, the travel scene corresponding to the target travel order may be determined by road congestion conditions of the current target driver, travel time of the target travel order, and a travel area where the target travel order is located.

The road congestion condition of the current target driver may cause that the target driver has a slower driving speed at one or more driving positions during the pickup process, and further causes that the target driver does not reach the corresponding driving position at the corresponding driving time node, and at this time, when a difference judgment rule between the current driving position and the preset driving position is set, the driving time interval between the driving position where congestion may occur and the adjacent driving position needs to be lengthened.

In step S103, it is detected whether the first abnormal stay probability output by the determination rule is greater than a preset probability threshold, so as to determine whether the target driver has abnormal pickup behavior under the determination condition of the determination rule, and further determine a subsequent determination basis for finally confirming the abnormal pickup behavior of the target driver when the target driver travels according to the determination manner (rule determination or model determination).

In step S104, if the first abnormal stopping probability is greater than the preset probability threshold, it indicates that the target driver has an abnormal stopping condition under the difference determination rule between the current driving position and the preset driving position, and therefore, it can be determined whether the target driver has an abnormal driving receiving behavior through the first abnormal stopping probability.

Here, the abnormal driving receiving behavior of the target driver can be determined earlier by the judgment of the difference judgment rule between the current driving position and the preset driving position, and the efficiency is high; meanwhile, the judgment based on the difference judgment rule between the current driving position and the preset driving position is the judgment for the corresponding scene, and the judgment accuracy for the abnormal stopping behavior of the target driver is higher unlike the abnormal stopping recognition model which needs to rely on a history label.

In step S105, if the first abnormal staying probability is not greater than the preset probability threshold, it indicates that the target driver has no abnormal staying condition under the difference determination rule between the current driving position and the preset driving position, and therefore, the abnormal staying identification model is required to determine the abnormal staying behavior of the target driver, and determine whether the target driver has the abnormal driving receiving behavior according to a fourth abnormal staying probability output by the abnormal staying identification model.

After the driving track information of the target driver accepting the cancelled order, which is obtained in step S101, is input into the pre-trained abnormal stay recognition model, and a second abnormal stay probability of the target driver in the order taking process is output through the abnormal stay recognition model, and the second abnormal stay probability of the target driver in the order taking process can be determined through the second abnormal stay probability output by the model.

Here, the abnormal stay recognition model may be any one of deep learning networks, and is trained by a correspondence relationship between different historical travel positions included in the acquired historical cancelled order and the historical travel time corresponding to each historical travel position.

Here, in the processing process of the abnormal stay recognition model, the non-extreme condition is processed first, so that a first sub abnormal stay probability is determined, meanwhile, a second sub abnormal stay probability is determined according to the judgment rule in the extreme condition, the first sub abnormal stay probability is judged and corrected according to the second sub abnormal stay probability, and therefore the second abnormal stay probability with higher accuracy is determined and is used as the output of the abnormal stay recognition model.

Here, the extreme case refers to a case where the travel track information that can be used to analyze the driving behavior of the target driver cannot be acquired, and may be defined by a distance between the order taking position of the target driver and the boarding position of the passenger, or may be determined by a time difference between the order taking time of the target driver and the order cancellation time, the non-extreme case refers to a case where the travel track information that can be used to analyze the driving behavior of the target driver is acquired, and the length of a time period between the order starting (placing) time and the order ending time of the travel order is generally long in the non-extreme case.

When the non-extreme condition is defined as the distance between the target driver pick-up point and the passenger boarding point, a preset distance threshold needs to be set, wherein the preset distance threshold is the minimum distance which the target driver can normally drive and move, and when the distance between the target driver pick-up position and the passenger boarding position is greater than the preset distance threshold, the non-extreme condition is determined; when the definition of the non-extreme condition is the time difference between the order taking time of the target driver and the order cancellation time, a preset time threshold needs to be set, the preset time threshold is the minimum time period for the target driver to move in normal driving, and when the time difference between the order taking time of the target driver and the order cancellation time is greater than the preset time threshold, the non-extreme condition is determined.

When the abnormal stay recognition model starts to recognize the abnormal stay behavior of the target driver, the travel track information of the target driver in the period from the order taking of the target driver to the order canceling is collected after the travel order is cancelled, the travel behavior of the target driver is accumulated in the period, and when the abnormal stay recognition model is used for judging the abnormal stay behavior of the target driver, the abnormal stay recognition model is uniform in time dimension, more observation time is provided for the target driver to take the order, and the accuracy of recognizing the abnormal stay behavior of the target driver under the non-extreme condition is ensured to a certain extent.

Further, in the method for determining an abnormal pickup behavior provided in the embodiment of the present application, the inputting the driving trajectory information into a pre-trained abnormal-stay recognition model to determine a second abnormal-stay probability that the target driver abnormally stays in the pickup process, and determining whether the target driver abnormally picks up the driving behavior based on the second abnormal-stay probability includes:

a 1: inputting the running track information into a pre-trained abnormal stay recognition model to determine a first sub-abnormal stay probability of the target driver for abnormal stay under the non-extreme condition in the order taking process; the abnormal stay recognition model is obtained by training based on different historical driving positions in the sample target driver historical driving track information in the non-extreme-case historical cancelled travel order and the historical driving time corresponding to each historical driving position.

b 1: and determining a second sub-abnormal stopping probability that the target driver has abnormal driving based on a preset driving track motion rule of the target driver in the driving process under the extreme condition and the driving track information of the target driver.

c 1: and determining a second abnormal stopping probability of the abnormal stopping of the target driver in the order taking process based on the first sub abnormal stopping probability and the second sub abnormal stopping probability.

In step a1, after the acquired driving track information of the target driver accepting the cancelled order is obtained, the driving track information is input into a pre-trained abnormal stay recognition model, a first sub-abnormal stay probability of the target driver in the order taking process is output through the abnormal stay recognition model, and the first sub-abnormal stay probability of the target driver in the order taking process can be determined through the first sub-abnormal stay probability output by the model.

The method comprises the steps of firstly determining a first sub-abnormal stay probability under the non-extreme condition, wherein the first sub-abnormal stay probability is obtained by training based on historical sample information of an abnormal stay recognition model, in the process, the determination of the abnormal stay behavior of a target driver based on the first sub-abnormal stay probability can ensure the accuracy under the non-extreme condition, but under some extreme scenes, the judgment of the abnormal stay behavior of the target driver by the first sub-abnormal stay probability has a certain misjudgment rate, and after the first sub-abnormal stay probability is output, the first sub-abnormal stay probability is judged according to a preset rule and then corrected, so that the output of the abnormal stay recognition model with higher accuracy is determined jointly.

In step b1, the driving track motion rule for the target driver under the extreme condition is determined according to the historical behavior track of the target driver under the extreme condition, and the abnormal stopping behavior of the target driver is judged according to the driving track motion rule and the driving track information of the target driver under the extreme condition, so that the probability of misjudgment of the abnormal stopping behavior of the target driver by the abnormal stopping recognition model under the extreme condition is avoided, the inaccuracy of model judgment under the extreme condition by the abnormal stopping recognition model is compensated through the rule judgment, and the accuracy of the final determination of the abnormal stopping behavior of the target driver in the drive receiving section is improved.

Here, the extreme case may include judgment features in both time and space, and the extreme case in the time dimension means that a time difference between a time when the travel order is cancelled and a time when the travel order is taken is less than a preset time threshold; in an extreme case of the space dimension, the distance between the order taking position of the target driver and the boarding position of the passenger is smaller than a preset distance threshold; and the extreme case can also include the case of combining the time characteristic and the space characteristic, namely, the time difference value between the time when the travel order is cancelled and the time when the travel order is picked up is smaller than the preset time threshold value, and the distance between the pick-up position of the target driver and the boarding position of the passenger is smaller than the preset distance threshold value.

The preset distance threshold value is the minimum distance which can be used for acquiring the driving track of the target driver; the preset time threshold is the length of the minimum time period in which the target driver can acquire the target driver's travel track. The case where the target driver running track can be acquired is a case where the target driver starts driving and does not encounter any interruption (road congestion, vehicle trouble, etc.) of the running operation.

For example, in the time dimension, after a passenger places an order through the platform, the target driver receives a target order through the platform, and when the target driver receives the target order and does not start driving to a boarding point where the passenger is located, the passenger suddenly cancels the order; under the spatial dimension, the target driver determines that the pick-up point is the same as the getting-on position of the passenger after receiving the driving, and the target driver does not click the button for confirming the arrival after normally receiving the passenger, so that the target driver is considered to be under the condition of not reaching the getting-on point of the passenger for receiving the order; in the case of a combination of the time dimension and the space dimension, the pickup position of the target driver is near the boarding position of the passenger, and the passenger cancels the order shortly after the passenger places the order, resulting in the target driver being cancelled before the passenger is missed.

When the target driver abnormally stops under the non-extreme condition, the driving track movement rule corresponding to the target order needs to be determined by considering the driving condition of the target driver under the current order, the travel time of the target order, the road congestion condition and the travel scene of the target order in combination with the driving track movement rule of the target driver.

Here, when the target driver abnormal stopping behavior is judged according to the rule, since the rule is not a conclusion obtained by counting probability like an identification model, but a judgment logic set according to a specific service scene, the rule has the characteristics of simple calculation, interpretability, light weight and convenience for deployment, the judgment rule of the rule can be adjusted according to different travel scenes, and the definition scale of the target driver abnormal stopping behavior is obtained.

For example, the travel time of the target order belongs to an early peak period, and the travel route of the passenger passes through a stage of relatively congested roads, at this time, a long-time stay behavior may occur in the driving process of the target driver, but the stay cannot be considered as an abnormal stay condition of the target driver, and an erroneous judgment condition caused by a travel scene and the like needs to be performed when the judgment is performed according to a rule.

Here, since the rule judgment is that a judgment process is given from a data level, that is, a result is given according to a judgment rule when the rule judgment is made, the reason why the target driver abnormally stops can be seen, and the judgment is more explanatory than the judgment that the recognition model directly outputs the probability result that the target driver abnormally stops.

For example, if the rule determines that continuous movement is required for a certain period of time, but the target driver does not move continuously for the certain period of time, and the stay time is too long at a certain position, it can be seen that the target driver is determined that the target driver has an abnormal stay when taking an order because the stay time is too long for the certain period of time.

In step c1, a second abnormal staying probability output by the abnormal staying identification model is determined according to the first sub abnormal staying probability determined in step a1 and the second sub abnormal staying probability determined in step Sb 2. The method comprises the steps of determining a first sub-abnormal stay probability under a non-extreme condition through an abnormal stay recognition model and determining a second sub-abnormal stay probability under an extreme condition through a rule, comprehensively determining a second abnormal stay probability output by the abnormal stay recognition model, considering the probability of abnormal driving receiving behavior of a target driver from the non-extreme condition and the extreme condition, more comprehensively analyzing the situation of the abnormal driving receiving behavior of the target driver, and improving the accuracy of the second abnormal stay probability output by the abnormal stay recognition model through judgment of the non-extreme condition and complementary interpretation of the driving track motion rule of the target driver under the extreme condition.

Here, the abnormal pickup behavior refers to an abnormal situation that a target driver has a driving track in the pickup process, and may be that the target driver does not drive according to a planned pickup route (i.e., a yaw behavior occurs); or the target driver wanders in a certain area in the process of taking over the driving, or the stay time is too long, and the like.

After the target driver receives the order, a plurality of driving positions in a driving track in the normal driving process of the target driver and the driving time corresponding to each driving position are planned according to the current road condition and the driving speed of the target driver for driving the vehicle, and when the target driver does not reach the corresponding driving position at a certain time point, the target driver is considered to have abnormal driving receiving behaviors with a maximum probability.

For example, according to the distance between the order taking position and the boarding position of the passenger, the current road condition and the speed of the target driver when driving the vehicle to normally travel, three traveling positions are determined: according to the distance between the position A, the position B and the position C and the order receiving position of the target driver, determining that the running time when the position A is reached is 30 minutes after the order is received, the running time when the position B is reached is 70 minutes after the order is received, and the running time when the position C is reached is 100 minutes after the order is received; according to the fact that the driving time when the driver arrives at the position A is 30 minutes after order taking, the driving time when the driver arrives at the position B is 120 minutes after order taking, and the driving time when the driver arrives at the position C is 500 minutes after order taking, the driver can know that the driver does not arrive at the position B and the position C at the corresponding time, and the driver can be determined to have abnormal driving receiving behaviors after arriving at the position A.

Here, the second abnormal stay probability that the target driver abnormally stays in the order taking process according to the joint judgment of the first sub-abnormal stay probability and the second sub-abnormal stay probability may be a second abnormal stay probability that the target driver abnormally stays in the order taking process according to the combined judgment of the first sub-abnormal stay probability and the second sub-abnormal stay probability, wherein a comprehensive abnormal stay probability is determined by a weighted sum of the first sub-abnormal stay probability and the second sub-abnormal stay probability, so that the target driver abnormally stays in the order taking process according to the comprehensive abnormal stay probability.

The second sub-abnormal stay probability under the extreme condition is determined according to the first sub-abnormal stay probability under the non-extreme condition and the driving track motion rule of the target driver for judgment, the judgment result of the driving track motion rule of the machine can be set to be a yes or no classification result for output, and the second abnormal stay probability of the abnormal stay in the order receiving process is determined according to the determined first sub-abnormal stay probability and the rule hit condition.

For example, when it is determined that the first sub-abnormal stay probability is greater than the preset probability threshold and the rule result is that abnormal stay occurs, it may be determined that the second abnormal stay probability at which the target driver abnormally stays in the pickup process is the first sub-abnormal stay probability; when the first sub abnormal stopping probability is determined to be larger than the preset probability threshold value and the rule result is that the abnormal stopping does not occur, the second abnormal stopping probability that the target driver abnormally stops in the process of receiving driving can be determined to be 0.

Here, the second abnormal staying probability is determined by the first sub abnormal staying probability and the second sub abnormal staying probability, and the second abnormal staying probability may be determined by performing weighted summation on the first sub abnormal staying probability and the second sub abnormal staying probability by using a first weight coefficient corresponding to the first sub abnormal staying probability and a second weight coefficient corresponding to the second sub abnormal staying probability; or the average probability is calculated according to the first sub-abnormal stay probability and the second sub-abnormal stay probability, and then the second abnormal stay probability is determined.

In the method, different weight coefficients can be set under different conditions, so that the influence degree of the first sub-abnormal stay probability and the second sub-abnormal stay probability on the determination of the final second abnormal stay probability is adjusted, and the accuracy of the determination of the target abnormal stay probability is improved.

After the first sub-abnormal stay probability that the target driver abnormally stays in the order taking process is determined under the non-extreme condition, whether the first sub-abnormal stay probability is larger than a preset probability threshold value or not can be determined, if the first sub-abnormal stay probability is not larger than the preset probability threshold value, the target driver is considered not to have the behavior of abnormal stay, and the determination process of the second sub-abnormal stay probability that the target driver abnormally stays in the order taking process according to the driving track motion rule of the target driver and the driving track information of the target driver under the extreme condition is not carried out any more.

Further, in the method for determining an abnormal pickup behavior provided in the embodiment of the present application, the determining a second sub-abnormal stopping probability that the target driver has undergone abnormal driving based on a preset driving trajectory movement rule of the target driver during driving under an extreme condition and the driving trajectory information of the target driver includes:

a 2: determining at least one abnormal driving position with abnormal driving at a plurality of driving positions included in the driving track; and the abnormal running position is a position with a running speed less than a preset running speed or a position with a deviation planning navigation route.

b 2: and determining a second sub-abnormal stopping probability that the target driver has abnormal driving based on the determined proportion of the at least one abnormal driving position in the plurality of driving positions.

In step a2, considering that the time difference between the trip order cancellation time and the order taking time in the extreme case is smaller than the preset time threshold, in the case that the time from the order placing to the order canceling is short, it is difficult to obtain the travel time corresponding to the track point from the track information of the target driver, and therefore in this case, at least one abnormal travel position of travel abnormalities at a plurality of travel positions needs to be determined, so as to determine the second sub-abnormal stay track in the middle extreme case according to the abnormal travel position.

Here, the at least one abnormal driving position where the driving abnormality occurs may be a position where the driving speed of the target driver is slow, or a position where the target driver deviates from a preset pick-up route, or the like.

The slow driving at a certain position means that the driving speed of the target driver is lower than the normal driving speed when the normal driving of the target driver is not influenced by the road condition; the deviation of the target driver from the preset pick-up route refers to the condition that the target driver is obviously deviated from the pick-up route on the premise that the road does not cause the target driver to have to detour.

Here, in the case where the time difference between the time when the travel order is cancelled and the pick-up time is less than the preset time threshold, the target driver may not travel a long distance during the normal driving, and the division of the plurality of driving positions may be a division according to a corresponding ratio, so that the plurality of driving positions are divided in the pick-up routes of different lengths.

For example, when the pickup driving distance is less than 1000m, 10 driving positions for dividing the pickup driving section into 10 driving positions may be set; and when the driving receiving driving distance is more than 1000m and less than 3000m, dividing the driving receiving road section into 15 driving positions and the like.

The division of the multiple driving positions can be average division of the driving receiving route, and the distance length between every two driving positions can be adjusted in real time according to the real-time road condition of the driving receiving route. For example, the distance between two adjacent driving positions may be divided by a longer length, in which case there may be no movement of the vehicle for a longer period of time on a certain congested road section.

In step a2, considering that the extreme case is that the distance between the pick-up position of the travel order and the boarding position of the passenger is smaller than the preset distance threshold, and when the distance between the pick-up position of the target driver and the boarding position of the passenger is smaller than the preset threshold, the target driver may not have a valid movement track on the boarding position for the case that the pick-up position of the target driver and the boarding position of the passenger are extremely close, and the driving position to be acquired may be an absolute position, which may include the pick-up position of the target driver and the boarding position of the passenger, the distance between the pick-up position and the boarding position needs to be determined, if the distance between the two positions is smaller than the preset distance threshold and the arrival instruction sent by the target driver is not received, which will determine that the target driver is an abnormal driving position at the pick-up position, the target driver is not displaced in this position, nor is there any travel speed.

Here, when the distance between the order taking position of the target driver and the boarding position of the passenger is smaller than the preset threshold, for the case that the order taking position of the target driver and the boarding position of the passenger are not extremely close, the target driver still has a certain driving track between the order taking position of the target driver and the boarding position of the passenger, at this time, the abnormal stopping behavior of the target driver needs to be determined according to the corresponding relationship between the driving position and the driving time, and the determination mode is consistent with the abnormal driving position determination step in which the time difference between the time when the travel order is cancelled and the order taking time is smaller than the preset time threshold, which is not described herein again.

In step b2, a second sub-abnormal stopping probability that the target driver has abnormal driving is determined according to the proportion of the at least one abnormal driving position determined in step a2 in the plurality of driving positions. The abnormal driving position represents the abnormal driving tendency of the target driver in the driving process, the existence of the individual abnormal driving position only represents that the target driver possibly has an abnormal condition at the position, but the abnormal driving condition may not exist in the whole driving process, so the proportion of the abnormal driving position in a plurality of driving positions needs to be determined, and further the second sub abnormal stopping probability is determined.

Here, the proportion of the abnormal driving position in the plurality of driving positions may be directly taken as the second sub abnormal stop probability in the extreme case, that is, the proportion of the abnormal driving position in the plurality of driving positions is equal to the second sub abnormal stop probability; the mapping relationship between the proportion of the abnormal driving position in the plurality of driving positions and the second sub-abnormal stopping probability may be preset, so that the second sub-abnormal stopping probability is determined according to the proportion of the abnormal driving position in the plurality of driving positions through the mapping relationship.

Preferably, the second sub-abnormal stopping probability is determined according to a mapping relation between the proportion of the plurality of driving positions and the second abnormal stopping probability, because the second sub-abnormal stopping probability, which is relatively similar and represented based on the proportion of the determined at least one abnormal driving position in the plurality of driving positions and is used for the target driver to have abnormal driving, is consistent, and the second sub-abnormal stopping probability is determined more accurately according to the mapping relation.

Here, since the abnormal stopping behavior is a continuous process, the number of the continuously abnormal driving positions may be determined, and the second sub abnormal stopping probability that the target driver has undergone abnormal driving may be determined according to the number of the abnormal driving positions and the mapping relationship between the preset number of the abnormal driving positions and the second abnormal stopping probability.

In the process of determining responsibility through the driving track motion rule of the target driver and the driving track information of the target driver in the driving process, the abnormal driving position of the target driver is determined according to the time-space characteristics (corresponding relation between the driving position and the driving time) of the target driver in the historical process, then the second sub abnormal stopping probability of abnormal driving of the target driver is determined, the judgment is carried out according to the time-space characteristics in the historical driving track, the dependence of abnormal stopping behaviors on real-time track services is reduced, and the usability of abnormal stopping behavior identification can be improved to a certain extent.

Further, in the method for determining an abnormal pickup behavior provided by the embodiment of the present application, the inputting the driving trajectory information into a pre-trained abnormal parking recognition model to determine a first sub-abnormal parking probability that the target driver abnormally stops under a non-extreme condition in a pickup process includes:

a 3: a plurality of driving positions included in the driving track information and a driving time corresponding to each driving position are determined.

b 3: and inputting each driving position and the corresponding driving time into a pre-trained abnormal stop recognition model, and determining at least one abnormal driving position.

c 2: and determining a first sub-abnormal stopping probability that the target driver has abnormal driving based on the determined proportion of the at least one abnormal driving position in the plurality of driving positions.

In step a3, a plurality of travel positions and a travel time for each travel position included in the travel locus information are determined, and a first sub-abnormal stop probability is determined by determining a plurality of abnormal travel positions among the plurality of travel positions.

Here, the division into the plurality of driving positions may be division according to respective proportions such that the plurality of driving positions are divided in the different lengths of the pick-up route; the division of the multiple driving positions can be average division of the driving receiving route, and the distance length between every two driving positions can be adjusted in real time according to the real-time road condition of the driving receiving route. For example, the distance between two adjacent driving positions may be divided by a longer length, in which case there may be no movement of the vehicle for a longer period of time on a certain congested road section.

In step b3, each driving position and the corresponding driving time are input into a pre-trained abnormal stop recognition model, so as to obtain at least one abnormal driving position, and further determine a first sub-abnormal stop probability according to the abnormal driving position.

Here, since the determination of whether or not the travel position is the abnormal travel position by the abnormal stop recognition model needs to be made in conjunction with the travel positions before and after the travel position and the travel time reference determination, when each travel position is input to the abnormal stop recognition model, it is necessary to input the travel positions to the abnormal stop recognition model in order.

The sequence of inputting each driving position can be input according to the sequence of the distance between the driving position and the order receiving position from near to far, and can also be input according to the sequence of the driving time from first to last.

In step c2, a first sub-abnormal stopping probability that the target driver has abnormally driven is determined according to the proportion of the at least one abnormal driving position determined in step b3 in the plurality of driving positions. The abnormal driving position represents the abnormal driving tendency of the target driver in the driving process, the existence of the individual abnormal driving position only represents that the target driver possibly has an abnormal condition at the position, but the abnormal driving condition may not exist in the whole driving process, so the proportion of the abnormal driving position in a plurality of driving positions needs to be determined, and further the first sub-abnormal stopping probability is determined.

Here, the proportion of the abnormal driving position in the plurality of driving positions may be directly taken as the first sub abnormal stop probability in the extreme case, that is, the proportion of the abnormal driving position in the plurality of driving positions is equal to the first sub abnormal stop probability; the mapping relationship between the proportion of the abnormal driving position in the plurality of driving positions and the first sub-abnormal stopping probability may be preset, so that the first sub-abnormal stopping probability is determined according to the proportion of the abnormal driving position in the plurality of driving positions through the mapping relationship.

Preferably, the first sub-abnormal stopping probability is determined according to a mapping relation between the proportion of the plurality of driving positions and the first sub-abnormal stopping probability, because the first sub-abnormal stopping probability, which is relatively similar and represented based on the proportion of the determined at least one abnormal driving position in the plurality of driving positions, of the target driver having abnormal driving is consistent, and the first sub-abnormal stopping probability is determined more accurately according to the mapping relation.

a 4: determining an updated mapping relation between a driving position and driving time in a driving track motion rule of a target driver based on a travel scene corresponding to a target travel order; wherein the travel scene includes at least one condition affecting a target driver driving trajectory.

b 4: and determining the running position as an abnormal running position when the mapping relation between the running position and the corresponding running time is inconsistent with the updated mapping relation for each running position.

c 3: and determining a second abnormal stop probability that the target driver has abnormal driving based on the determined proportion of the at least one abnormal driving position in the plurality of driving positions.

In step a4, according to the trip scene corresponding to the determined target trip order, an updated mapping relationship between the driving position and the driving time in the driving track motion rule of the target driver is determined, and considering that interference of different driving conditions may occur in the process of taking over the driving by the target driver, which may result in that the target driver stays at a certain driving position for too long time, but the stay cannot be regarded as the stay of the target driver with subjective awareness, and therefore the abnormal stay behavior occurring to the target driver cannot be determined, and therefore the mapping relationship between the driving position and the driving time in the driving track motion rule of the target driver needs to be updated in combination with the real-time driving road condition, so as to improve the accuracy of the determination of the abnormal stay behavior of the target driver.

The road congestion condition of the current target driver driving may cause that the driving speed of the target driver at one or more driving positions is slow in the process of driving and receiving, and further cause that the corresponding driving position is not reached at the corresponding driving time node, and at this time, when the mapping relationship is set to be updated, the driving time interval between the driving position where congestion may occur and the adjacent driving position needs to be lengthened; under the condition of travel time of the target travel order, due to the fact that certain differences exist among congestion conditions of roads at different travel times (the travel roads are congested at the peak in the morning and at the evening in one day, and the congestion conditions of the travel roads at the peak balancing period are general), the congestion conditions of the roads can be determined through different travel times, the updated mapping relation between the travel position and the travel time in the driving track motion rule of the target driver is further determined, the determination process is consistent with the confirmation step under the congestion condition of the roads, and the detailed description is omitted; the travel area represents the particularity of travel time of different areas, which may affect the established mapping relationship between the travel position and the travel time of the target driver, for example, in european countries, the time length of each traffic light is longer than that in the country, so when the mapping relationship is subjected to migration calculation of different areas, a change needs to be made, and for the above example, the change is made in the following manner: the driving positions corresponding to the possible traffic posts and the driving time intervals between the adjacent driving positions need to be lengthened.

In step b4, for each driving position during the pickup of the target driver, if the driving time corresponding to the driving position is not consistent with the updated mapping relationship, the driving position is determined to be an abnormal driving position.

Here, the map between the travel position and the corresponding travel time does not match the updated map, and this is reflected in a case where the travel time at the travel position does not match the travel time corresponding to the travel position in the updated map and is later than the travel time in the updated map.

For example, the travel time for the target driver to reach the a travel position during pickup is indicated in the updated map as 8: 00, but the travel time for the target driver to reach the a travel position during pickup is 9: 30, it is possible to determine that the running position is an abnormal running position.

In step c3, the proportion of the abnormal driving position in the plurality of driving positions can be directly used as the second sub-abnormal stopping probability in the extreme case, that is, the proportion of the abnormal driving position in the plurality of driving positions is equal to the second sub-abnormal stopping probability; the mapping relationship between the proportion of the abnormal driving position in the plurality of driving positions and the second abnormal stopping probability may be preset, so that the second sub abnormal stopping probability is determined according to the proportion of the abnormal driving position in the plurality of driving positions through the mapping relationship.

Preferably, the second sub-abnormal stopping probability is determined according to a mapping relation between the proportion of the plurality of driving positions and the second abnormal stopping probability, because the second sub-abnormal stopping probability, which is relatively similar and represented based on the proportion of the determined at least one abnormal driving position in the plurality of driving positions and is used for the target driver to drive abnormally, is consistent, and the second sub-abnormal stopping probability is determined more accurately according to the mapping relation.

Further, referring to fig. 2, fig. 2 is a flowchart of another method for determining abnormal pickup behavior according to an embodiment of the present application, where the method for determining abnormal pickup behavior further includes:

s201, obtaining a plurality of historical cancelled travel orders and order attributes corresponding to each historical cancelled travel order; the order attribute comprises at least one of order cancellation time between the time when the travel order is cancelled and the order taking time and driving receiving distance between the order taking position of the travel order and the boarding position of the passenger.

S202, determining at least one abnormal order with order cancellation time smaller than preset time and at least one abnormal order with order cancellation distance smaller than a preset distance threshold value for a target driver driving distance from the obtained plurality of historical cancelled travel orders.

S203, filtering out at least one order cancellation time and at least one cancellation time abnormal order from the plurality of historical cancelled travel orders, and determining a plurality of sample cancelled travel orders.

S204, training and constructing a deep learning network based on different historical driving positions in the historical driving track information corresponding to the cancelled travel order of each sample and the historical driving time corresponding to each historical driving position to obtain an abnormal stay recognition model.

In step S201, while obtaining a plurality of history cancelled travel orders, it is further required to determine order attribute information corresponding to each history cancelled travel order, and screen out orders under extreme conditions existing in the history cancelled travel orders according to the order attribute information, so as to avoid a problem that data of the history cancelled travel orders under extreme conditions affects a training effect in a model training process, and further affects accuracy of model training.

Here, the order attribute information includes at least one of an order cancellation time between a time when the travel order is cancelled and an order taking time, and a pickup distance between an order taking position of the travel order and a boarding position of the passenger.

In step S202, according to the order attribute information of each historical cancelled order determined in step S201, a cancellation time abnormal order and a cancellation route abnormal order are determined from the obtained multiple historical cancelled orders, and then the influence of the sample cancelled travel order in an extreme case on model training is eliminated by cancelling the cancellation time abnormal order and the cancellation route abnormal order.

In step S203, the at least one cancellation time exception order and the at least one cancellation distance exception order determined in step S202 are filtered from the plurality of acquired historical cancellation trip orders, so as to determine that a plurality of samples of models for non-extreme cases have cancelled trip orders.

Here, for an order cancelled from a history, which may be an order with abnormal cancellation time or an order with abnormal cancellation distance, in order to reduce the workload of filtering the abnormal order, a deduplication step may be performed before filtering, and for an order identifier (order number, etc.) of the same history order cancelled, orders with different order attributes for the same order identifier (that is, an order with abnormal cancellation time or an order with abnormal cancellation distance) are deduplicated.

For example, if a history cancelled travel order is both a cancellation time exception order and a cancellation distance exception order, only the order attribute of the history cancelled travel order which is a cancellation distance exception order may be retained after the deduplication.

In step S204, the constructed deep learning network is trained according to the travel order cancelled from the plurality of samples determined in step S203, so as to obtain an abnormal stay recognition model.

Here, the training process may be as follows:

obtaining sample target driver historical driving track information in a plurality of sample cancelled orders and actual probability of abnormal stay of a target driver corresponding to each cancelled order in the driving process;

inputting historical driving track information of the target driver of each sample to the constructed deep learning network aiming at each cancelled order to obtain the prediction probability of abnormal stay of the target driver corresponding to the cancelled order of the sample in the driving process;

for each cancelled order of the sample, determining a deviation value between the predicted probability and the actual probability of the cancelled order of the sample;

if the deviation value corresponding to the cancelled order of the sample is larger than a preset deviation threshold value, adjusting parameters in the deep learning network until the deviation value corresponding to the cancelled order of each sample is smaller than or equal to the preset deviation threshold value, determining that the deep learning network is completely trained, and determining the deep learning network after the deep learning network is completely trained as the trained recognition model.

Further, in the method for determining an abnormal pickup behavior provided in the embodiment of the present application, a first abnormal stopping probability that the target driver stops abnormally is determined based on a difference determination rule between a preset current driving position and a preset driving position and the driving track information of the target driver, and the method includes:

a 5: and on the running time corresponding to each running position, if the distance difference between the current running position and the preset running position is greater than a preset difference threshold value, determining that the running position is an abnormal running position.

b 5: and determining a first abnormal stop probability that the target driver has abnormal driving based on the determined proportion of the at least one abnormal driving position in the plurality of driving positions.

In step a5, according to the current travel scene, a plurality of driving positions and driving time corresponding to each driving position in the process of receiving driving by the target driver are preset, according to the difference between the current driving position and the preset driving position, whether the current driving position is an abnormal driving position or not can be determined, and then the first abnormal stopping probability that the target driver has abnormally driven is determined according to the determined abnormal driving position.

After the target driver receives the order, a plurality of driving positions in a driving track in the normal driving process of the target driver and the driving time corresponding to each driving position are planned according to the current road condition and the driving speed of the target driver for driving the vehicle, and when the target driver does not reach the corresponding driving position at a certain time point, whether the current driving position is an abnormal driving position or not is considered, and then the determined abnormal driving position is passed.

In step b5, a first abnormal stop probability that the driver is abnormally driven is determined according to the proportion of the at least one abnormal driving position determined in step a5 in the plurality of driving positions. The abnormal driving position represents the abnormal driving tendency of the target driver in the driving process, the existence of the individual abnormal driving position only represents that the target driver possibly has an abnormal condition at the position, but the abnormal driving condition may not exist in the whole driving process, so the proportion of the abnormal driving position in a plurality of driving positions needs to be determined, and further the first abnormal stop probability is determined.

Here, the proportion of the abnormal driving position in the plurality of driving positions may be directly taken as the first abnormal stop probability in the extreme case, that is, the proportion of the abnormal driving position in the plurality of driving positions is equal to the first abnormal stop probability; the mapping relationship between the proportion of the abnormal driving position in the plurality of driving positions and the first abnormal stopping probability may be preset, so that the first abnormal stopping probability is determined according to the proportion of the abnormal driving position in the plurality of driving positions through the mapping relationship.

Preferably, the first abnormal stopping probability is determined according to a mapping relation between the proportions of the plurality of driving positions and the first abnormal stopping probability, and because the first abnormal stopping probabilities, which are relatively similar and represented based on the proportions of the determined at least one abnormal driving position in the plurality of driving positions, of the target driver having abnormal driving are consistent, the first abnormal stopping probability is determined to be more accurate according to the mapping relation.

Taking a scene that an order is cancelled in a travel process as an example, a process for determining abnormal pickup behavior in the technical scheme of the application is explained, and the determination of the abnormal pickup behavior comprises the following steps:

step 1: generating a target travel order according to the determined travel starting place of the passenger and the travel destination input by the passenger, and matching a corresponding pickup target driver for the target travel order;

step 2: after the target driver takes an order, monitoring the driving track of the target driver in the process of going to the travel origin where the passenger is located from the order taking position in real time, and acquiring the driving track information of the target driver in the process of going to the travel origin where the passenger is located from the order taking position when receiving an order cancel instruction of the passenger;

and step 3: according to a preset bottom-holding strategy: determining whether a first abnormal stopping probability of the target driver, which is calculated under a bottom-in-pocket strategy and has abnormal stopping, is greater than a preset probability threshold value or not according to a difference judgment rule between the current driving position and a preset driving position and driving track information of the target driver;

and 4, step 4: if the first abnormal stay probability of the target driver which is calculated under the bottom-catching strategy and is subjected to the abnormal stay is larger than the preset probability threshold, determining that the target driver hits the bottom-catching rule in the driving-catching process, judging whether the target driver has the abnormal driving-catching behavior according to the first abnormal stay probability, and determining that the target driver has the abnormal driving-catching behavior in the driving-catching process when the first abnormal stay probability is larger than the preset abnormal judgment probability.

And 5: if the first abnormal stay probability of the target driver, which is calculated under the bottom-catching strategy and has the abnormal stay, is not larger than the preset probability threshold, the driving positions of different time in the acquired driving track information and the driving time of each driving position are input into a pre-trained abnormal stay recognition model, and according to the output result of the abnormal stay recognition model, a second abnormal stay probability of the target driver, which has the abnormal stay in the driving receiving process, is determined after the judgment of the abnormal stay recognition model;

step 5-1: inputting the running track information into a pre-trained abnormal stay recognition model to determine a first sub-abnormal stay probability of the target driver for abnormal stay under non-extreme payment in the order receiving process;

step 5-2: determining whether the target driver meets an extreme abnormal condition in the order receiving process, and if the time difference between the order cancelling time and the order placing time of the passenger is found to be short, determining a second sub-abnormal stay probability that the target driver abnormally drives under the extreme condition according to the driving track information of the target driver and the driving track motion rule of the target driver;

step 5-3: and determining a second abnormal stopping probability of the abnormal stopping of the target driver in the order receiving process according to the first sub abnormal stopping probability and the second sub abnormal stopping probability, and determining that the abnormal driving receiving behavior of the target driver occurs in the driving receiving process when the second abnormal stopping probability is greater than a preset probability threshold.

Step 6: after the target driver is determined to have abnormal pickup behavior in the pickup process, the target driver is determined to be the responsible party in the order cancellation process, the target driver is asked for responsibility, meanwhile, the abnormal pickup behavior of the target driver is recorded, and resources are considered not to incline towards the target driver when orders are subsequently distributed.

According to the method for determining the abnormal drive receiving behavior, the cancelled target trip orders are obtained, and the order information comprises the driving track information of the target driver carrying the target trip orders; determining a first abnormal stopping probability of the target driver when the target driver abnormally stops according to a preset difference judgment rule between the current driving position and a preset driving position, detecting whether the first abnormal stopping probability is greater than a preset probability threshold, and determining whether the target driver abnormally connects the driving according to the first abnormal stopping probability if the first abnormal stopping probability is greater than the preset probability threshold; if the first abnormal stopping probability is not greater than the preset probability threshold, the driving track information is input into a pre-trained abnormal stopping recognition model to determine a second abnormal stopping probability, whether the target driver has abnormal driving receiving behavior is determined according to the second abnormal stopping probability, the abnormal driving receiving behavior of the target driver can be comprehensively determined under different conditions according to a difference judgment rule between the current driving position and the preset driving position and the abnormal stopping recognition model, and the accuracy of recognizing the abnormal driving receiving behavior of the target driver is improved.

Referring to fig. 3 to 4, fig. 3 is a first schematic structural diagram of a device for determining an abnormal pickup behavior provided in the embodiment of the present application, and fig. 4 is a second schematic structural diagram of the device for determining an abnormal pickup behavior provided in the embodiment of the present application. As shown in fig. 3, the determining means 300 includes:

an order information obtaining module 310, configured to obtain order information of the cancelled target trip order; the order information comprises the driving track information of a target driver carrying a target trip order; the driving track information comprises driving positions of a target driver at different time after receiving a target trip order and driving time corresponding to each driving position;

a first probability determination module 320, configured to determine a first abnormal stopping probability that a target driver stops abnormally based on a difference determination rule between a preset current driving position and a preset driving position and the driving track information of the target driver;

a probability detection module 330, configured to detect whether the first abnormal staying probability is greater than a preset probability threshold;

a first abnormal behavior determination module 340, configured to determine whether an abnormal pickup behavior occurs to a target driver based on the first abnormal stopping probability if the first abnormal stopping probability is greater than a preset probability threshold;

and a second abnormal behavior determining module 350, configured to, if the first abnormal stopping probability is not greater than a preset probability threshold, input the driving trajectory information into a pre-trained abnormal stopping recognition model to determine a second abnormal stopping probability that the target driver abnormally stops in the order receiving process, and determine whether the target driver abnormally receives a driving behavior based on the second abnormal stopping probability.

Further, as shown in fig. 4, the determining apparatus 300 further includes a model training module 360, where the model training module 360 is configured to:

Further, a second abnormal behavior determination module 350 is configured to input the driving track information into a pre-trained abnormal stopping recognition model to determine a second abnormal stopping probability that the target driver abnormally stops in the order taking process, where the second abnormal behavior determination module 350 is configured to:

Further, when the second abnormal behavior determination module 350 is configured to determine a second sub-abnormal stopping probability that the target driver has undergone abnormal driving based on the driving track motion rule of the target driver and the driving track information of the target driver during driving under the preset extreme conditions, the second abnormal behavior determination module 350 is configured to:

Further, when the second abnormal behavior determination module 350 is configured to input the driving track information into a pre-trained abnormal stopping recognition model to determine a first sub-abnormal stopping probability that the target driver abnormally stops in a non-extreme situation during the order taking process, the second abnormal behavior determination module 350 is configured to:

Further, when the first abnormal behavior determination module 340 is configured to determine a first abnormal stopping probability that the target driver stops abnormally based on a difference determination rule between a preset current driving position and a preset driving position and the driving track information of the target driver, the first abnormal behavior determination module 340 is configured to:

The abnormal pickup behavior determining device provided by the embodiment of the application acquires the cancelled target trip orders and order information of the travel track information of a target driver carrying the target trip orders; determining a first abnormal stopping probability of the target driver when the target driver abnormally stops according to a preset difference judgment rule between the current driving position and a preset driving position, detecting whether the first abnormal stopping probability is greater than a preset probability threshold, and determining whether the target driver abnormally connects the driving according to the first abnormal stopping probability if the first abnormal stopping probability is greater than the preset probability threshold; if the first abnormal stopping probability is not greater than the preset probability threshold, the driving track information is input into a pre-trained abnormal stopping recognition model to determine a second abnormal stopping probability, whether the target driver has abnormal driving receiving behavior is determined according to the second abnormal stopping probability, the abnormal driving receiving behavior of the target driver can be comprehensively determined under different conditions according to a difference judgment rule between the current driving position and the preset driving position and the abnormal stopping recognition model, and the accuracy of recognizing the abnormal driving receiving behavior of the target driver is improved.

Referring to fig. 5, fig. 5 is a schematic structural diagram of an electronic device according to an embodiment of the present disclosure. As shown in fig. 5, the electronic device 500 includes a processor 510, a memory 520, and a bus 530.

The memory 520 stores machine-readable instructions executable by the processor 510, when the electronic device 500 runs, the processor 510 and the memory 520 communicate through the bus 530, and when the machine-readable instructions are executed by the processor 510, the steps of the method for determining abnormal pickup behavior in the embodiment of the method shown in fig. 1 and fig. 2 may be performed.

An embodiment of the present application further provides a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the step of the method for determining an abnormal driving receiving behavior in the method embodiments shown in fig. 1 and fig. 2 may be executed.

It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.

In the several embodiments provided in the present application, it should be understood that the disclosed system, apparatus and method may be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one logical division, and there may be other divisions when actually implemented, and for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection of devices or units through some communication interfaces, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer-readable storage medium executable by a processor. Based on such understanding, the technical solution of the present application or portions thereof that substantially contribute to the prior art may be embodied in the form of a software product stored in a storage medium and including instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present application. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.

Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present application, and are used for illustrating the technical solutions of the present application, but not limiting the same, and the scope of the present application is not limited thereto, and although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope disclosed in the present application; such modifications, changes or substitutions do not depart from the spirit and scope of the exemplary embodiments of the present application, and are intended to be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A method for determining abnormal pickup behavior is characterized by comprising the following steps:

2. The method for determining according to claim 1, wherein the inputting the information of the driving track into a pre-trained abnormal stay recognition model to determine a second abnormal stay probability of the target driver having abnormal stay in the order taking process comprises:

3. The determination method according to claim 2, characterized in that the extreme case is a case where at least one of the following conditions is present:

4. The method for determining according to claim 2, wherein the determining of the second sub-abnormal stay probability that the target driver has undergone abnormal driving based on the preset driving track motion rule of the target driver during driving in the extreme case and the driving track information of the target driver comprises:

5. The method for determining according to claim 2, wherein the inputting the information of the driving track into a pre-trained abnormal stay recognition model to determine a first sub-abnormal stay probability of the target driver having abnormal stay in the non-extreme condition during the order taking process comprises:

6. The determination method according to claim 2, wherein the second sub-abnormal stay probability that the target driver has undergone abnormal driving is determined based on the preset driving track motion rule of the target driver during driving in the extreme case and the driving track information of the target driver:

7. The method of determining according to claim 1, wherein the cancellation party of the target travel order is a passenger;

8. The determination method according to claim 1, characterized in that the determination method further comprises:

9. The method according to claim 1, wherein the determining a first abnormal stay probability that the target driver has abnormally stopped based on a difference determination rule between a preset current driving position and a preset driving position and the driving track information of the target driver comprises:

10. An abnormal pickup behavior determination device, characterized in that the determination device comprises:

11. An electronic device, comprising: a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, when an electronic device runs, the processor and the storage medium communicate through the bus, and the processor executes the machine-readable instructions to execute the steps of the method for determining abnormal pickup behavior according to any one of claims 1 to 9.

12. A computer-readable storage medium, having a computer program stored thereon, where the computer program is executed by a processor to perform the steps of the method for determining abnormal pickup behavior according to any one of claims 1 to 9.