CN117789460A

CN117789460A - Road condition prediction method, device, equipment and storage medium

Info

Publication number: CN117789460A
Application number: CN202311733421.0A
Authority: CN
Inventors: 窦祖俊; 孙明芳; 白红霞; 张岩; 刘子昊; 刘艳荣; 邱亚星
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2023-12-15
Filing date: 2023-12-15
Publication date: 2024-03-29

Abstract

The disclosure provides a road condition prediction method, a device, equipment and a storage medium, relates to the technical field of artificial intelligence, and particularly relates to the technical field of automatic driving. The specific implementation scheme is as follows: responding to the situation that the vehicle enters a traffic light influence road section, and acquiring historical vehicle track information and traffic light state information of the traffic light influence road section in a preset period; according to the historical vehicle track information and the traffic light state information, determining the queue length information of a traffic light influence queue corresponding to the traffic light influence road section; and responding to the parking operation of the vehicle, and predicting and obtaining the road condition information of the road section affected by the traffic light according to the queue length information of the traffic light affected queue.

Description

Road condition prediction method, device, equipment and storage medium

Technical Field

The disclosure relates to artificial intelligence technology, in particular to automatic driving technology, and especially relates to a road condition prediction method, device, equipment and storage medium.

Background

In modern traffic systems, accurate determination of traffic conditions (e.g., clear, creep, or congestion) at traffic light intersections is critical to traffic management and travel decisions.

The main methods for identifying the road conditions currently in common are as follows: 1. a sensor-based method: by deploying sensor devices such as vehicle detectors, cameras, radars and the like on roads, traffic data are collected in real time, and analysis and recognition of road conditions are performed by using the data. 2. Deep learning method based on feature processing: according to different road grades, the congestion conditions (smooth, slow and congestion) are defined based on the passing speed, so that the road conditions are identified.

The sensor-based method has the problems of high cost, small application range and incapability of effectively acquiring accurate images, and the feature processing-based deep learning method has the problem that the road condition prediction at a certain traffic light lamp cap is difficult to accurately learn, so that the two modes can not effectively provide road condition information for vehicles in time, and the vehicles are helped to select an optimal driving strategy.

Disclosure of Invention

The disclosure provides a road condition prediction method, a device, equipment and a storage medium.

According to a first aspect of the present disclosure, there is provided a road condition prediction method, the method including:

responding to the situation that a vehicle enters a traffic light influence road section, and acquiring historical vehicle track information and traffic light state information of the traffic light influence road section in a preset period;

determining queue length information of a traffic light influence queue corresponding to the traffic light influence road section according to the historical vehicle track information and the traffic light state information;

and responding to the parking operation of the vehicle, and predicting and obtaining the road condition information of the traffic light influence road section according to the queue length information of the traffic light influence queue.

Further, determining, according to the historical vehicle track information and the traffic light state information, queue length information of a traffic light influence queue corresponding to the traffic light influence road section, including:

Acquiring historical parking information in the historical vehicle track information;

and determining a traffic light influence queue corresponding to the traffic light influence road section and queue length information of the traffic light influence queue according to the historical parking information and the traffic light state information.

Further, acquiring historical parking information in the historical vehicle track information includes:

grouping the historical vehicle track information to obtain a plurality of groups of historical vehicle track information, wherein each group of historical vehicle track information corresponds to a period;

performing parking detection on each set of historical vehicle track information to determine parking operations of the traffic light affected road sections in each time period;

and determining historical vehicle parking information corresponding to each parking operation according to the parking operation of the traffic light influence road section in each time period.

Further, determining a traffic light influence queue corresponding to the traffic light influence road section according to the historical parking information and the traffic light state information, including:

performing nuclear density estimation on each set of historical vehicle track information in a plurality of sets of historical vehicle track information respectively to obtain probability density distribution values of the historical vehicle passing through each coordinate point in a period corresponding to each set of historical vehicle track information;

Comparing the probability density distribution values of each of a plurality of continuous time periods to determine the traffic flow change rule of the traffic light influence road section in different time periods;

and determining the traffic light influence queues corresponding to the traffic light influence road sections in different time periods according to the traffic flow change rules, the historical parking information and the traffic light state information in different time periods.

Further, according to the traffic flow change rule, the historical parking information and the traffic light state information of different time periods, determining the traffic light influence queue corresponding to the traffic light influence road section in different time periods includes:

screening out target historical parking information influenced by traffic lights in each time period according to the traffic light state information in different time periods;

according to the target historical parking information in each time period, determining a first traffic light influence queue of the traffic light influence road section in each time period;

determining a second traffic light influence queue of the traffic light influence road section in each time period according to the traffic flow change rules of different time periods;

and correspondingly fusing the first traffic light influence queue and the second traffic light influence queue in each time period to obtain the traffic light influence queues corresponding to the traffic light influence road sections in different time periods.

Further, according to the target historical parking information in each period, determining a first traffic light influence queue of the traffic light influence road section in each period includes:

acquiring a plurality of historical parking positions in the target historical parking information in each period;

and generating the first traffic light influence queue of the traffic light influence road section in each period according to the plurality of historical parking positions.

Further, predicting the road condition information of the traffic light affected road section according to the queue length information of the traffic light affected queue, including:

inputting the queue length information of the traffic light influence queue into a road condition prediction model obtained through pre-training to obtain the road condition information of the traffic light influence road section output by the road condition prediction model;

the road condition prediction model is obtained by training a pre-training initial neural network model, and each pair of enhanced training data used for training comprises: the sample traffic light influences the queue length information of the queue and the corresponding traffic light influences the sample road condition information of the road section.

Further, the road condition prediction model learns time sequence information of each pair of labeled training data in the training process, and adopts a plurality of classifiers to estimate road condition information in different driving directions based on each pair of labeled training data.

Further, in the training process, the road condition prediction model predicts and scores each pair of unlabeled training data, and fuses each pair of unlabeled training data meeting a scoring threshold with road condition information in different driving directions to obtain each pair of enhanced training data.

Further, the predicting, according to the queue length information of the traffic light influencing queue, the road condition information of the traffic light influencing road section includes:

and predicting the road condition information of the traffic light influence road section according to the historical vehicle track information and the queue length information of the traffic light influence queue.

Further, predicting the road condition information of the traffic light affected road section according to the historical vehicle track information and the queue length information of the traffic light affected queue, including:

inputting the historical vehicle track information and the queue length information of the traffic light influence queue into a pre-built pre-training model, so as to predict and obtain the road condition information of the traffic light influence road section according to the historical vehicle track information and the queue length information of the traffic light influence queue by adopting the pre-training model;

The pre-training model is pre-built based on a mask prediction task and a time sequence prediction task, the mask prediction task is used for randomly masking part of contents in each set of historical vehicle track information in the historical vehicle track information, so that the pre-training model learns local dependency relations in each set of historical vehicle track information by predicting the part of contents to be masked, and the part of contents comprises: each node in the traffic light influence queue and the time corresponding to each node in the traffic light influence queue; the time sequence prediction task is used for predicting the time corresponding to each node in the traffic light influence queue when the historical vehicle track of the individual group passes through the traffic light influence queue by adopting the historical vehicle track information of the individual group and the corresponding traffic light influence queue.

Further, the pre-training model marks the historical road condition information corresponding to the historical vehicle track information according to the historical parking information in the historical vehicle track information, the traffic light state information and the time corresponding to each node in the traffic light influence queue.

Further, the training data of the pre-training model includes: the historical vehicle track information and the corresponding queue length information of the traffic light influence queue;

In the training process of the pre-training model, the parameters of the pre-training model are adopted to encode training data, so that encoded training data are obtained; enhancing the encoded training data by adopting a time sequence model to obtain time sequence enhanced training data; and estimating road condition information in different driving directions by adopting a plurality of classifiers based on the training data after the time sequence enhancement.

According to a second aspect of the present disclosure, there is provided a road condition prediction apparatus, the apparatus comprising:

the acquisition unit is used for responding to the fact that the vehicle enters the traffic light influence road section and acquiring historical vehicle track information and traffic light state information of the traffic light influence road section in a preset period;

the determining unit is used for determining the queue length information of the traffic light influence queue corresponding to the traffic light influence road section according to the historical vehicle track information and the traffic light state information;

and the prediction unit is used for responding to the parking operation of the vehicle and predicting and obtaining the road condition information of the traffic light influence road section according to the queue length information of the traffic light influence queue.

Further, the determining unit includes:

an acquisition subunit, configured to acquire historical parking information in the historical vehicle track information;

And the determining subunit is used for determining a traffic light influence queue corresponding to the traffic light influence road section and queue length information of the traffic light influence queue according to the historical parking information and the traffic light state information.

Further, the acquiring subunit includes:

the dividing module is used for grouping the historical vehicle track information to obtain a plurality of groups of historical vehicle track information, wherein each group of historical vehicle track information corresponds to one period;

the detection module is used for carrying out parking detection on each group of historical vehicle track information so as to determine parking operation of the traffic light affected road section in each time period;

and the first determining module is used for determining historical vehicle parking information corresponding to each parking operation according to the parking operation of the traffic light influence road section in each time period.

Further, the determining subunit includes:

the estimating module is used for respectively carrying out nuclear density estimation on each set of historical vehicle track information in the plurality of sets of historical vehicle track information to obtain probability density distribution values of the historical vehicle passing through each coordinate point in a period corresponding to each set of historical vehicle track information;

The second determining module is used for comparing the probability density distribution values of each of a plurality of continuous time periods and determining the traffic flow change rule of the traffic light influence road section in different time periods;

and the third determining module is used for determining the traffic light influence queues corresponding to the traffic light influence road sections in different time periods according to the traffic flow change rules, the historical parking information and the traffic light state information in different time periods.

Further, the third determining module further includes:

the screening sub-module is used for screening out target historical parking information influenced by traffic lights in each time period according to the traffic light state information in different time periods;

the first determining submodule is used for determining a first traffic light influence queue of the traffic light influence road section in each time period according to the target historical parking information in each time period;

the second determining submodule is used for determining a second traffic light influence queue of the traffic light influence road section in each time period according to the traffic flow change rules of different time periods;

and the fusion submodule is used for correspondingly fusing the first traffic light influence queue and the second traffic light influence queue in each time period to obtain the traffic light influence queues corresponding to the traffic light influence road sections in different time periods.

Further, the first determination submodule is specifically further configured to:

Further, the prediction unit includes:

the first prediction subunit is used for inputting the queue length information of the traffic light influence queue into a road condition prediction model obtained through training in advance so as to obtain the road condition information of the traffic light influence road section output by the road condition prediction model; the road condition prediction model is obtained by training a pre-training initial neural network model, and each pair of enhanced training data used for training comprises: the sample traffic light influences the queue length information of the queue and the corresponding traffic light influences the sample road condition information of the road section.

Further, the road condition prediction model learns time sequence information of each pair of labeled training data in the training process, and adopts a plurality of classifiers to estimate road condition information in different driving directions based on each pair of labeled training data; and in the training process of the road condition prediction model, predicting and scoring each pair of unlabeled training data, and fusing each pair of unlabeled training data meeting a scoring threshold with road condition information in different driving directions to obtain each pair of enhanced training data.

Further, the prediction unit includes:

and the second prediction subunit is used for predicting and obtaining the road condition information of the traffic light influence road section according to the historical vehicle track information and the queue length information of the traffic light influence queue.

Further, the second prediction subunit is specifically configured to:

According to a third aspect of the present disclosure, there is provided an electronic device comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform any one of the methods.

According to a fourth aspect of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the method according to any one of the claims.

According to a fifth aspect of the present disclosure, there is provided a computer program product comprising: a computer program stored in a readable storage medium, from which it can be read by at least one processor of an electronic device, the at least one processor executing the computer program causing the electronic device to perform the method of the first aspect.

According to the technology disclosed by the invention, the historical vehicle track data can be effectively utilized by acquiring the historical vehicle track information and the traffic light state information of the traffic light influence road section in the preset time period, and the accuracy and the reliability of road condition prediction are improved. According to historical vehicle track information and traffic light state information, the queue length information of the traffic light influence queues is determined, traffic flow conditions of road sections affected by the traffic lights can be accurately reflected, and important basis is provided for road condition prediction. And then, according to the queue length information of the traffic light influence queue, the road condition information of the traffic light influence road section is predicted and obtained by responding to the parking operation of the vehicle, so that the road condition information can be provided for the vehicle in time, the vehicle is helped to select the optimal driving strategy, and the driving efficiency and the driving safety are improved.

It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the disclosure, nor is it intended to be used to limit the scope of the disclosure. Other features of the present disclosure will become apparent from the following specification.

Drawings

The drawings are for a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:

FIG. 1 is a flow chart of a road condition prediction method provided according to an embodiment of the present disclosure;

FIG. 2a is a flow chart of an alternative road condition prediction method in which embodiments of the present disclosure may be implemented;

FIG. 2b is a flow chart of another alternative road condition prediction method in which embodiments of the present disclosure may be implemented;

FIG. 3 is a flowchart of an alternative road condition prediction method provided in accordance with an embodiment of the present disclosure;

FIG. 4 is a flowchart of an alternative road condition prediction method provided in accordance with an embodiment of the present disclosure;

FIG. 5 is a flowchart of an alternative road condition prediction method provided in accordance with an embodiment of the present disclosure;

FIG. 6 is a flowchart of an alternative road condition prediction method provided in accordance with an embodiment of the present disclosure;

fig. 7 is a schematic frame diagram of a road condition prediction apparatus according to an embodiment of the disclosure;

Fig. 8 is a block diagram of an electronic device for implementing a road condition prediction method according to an embodiment of the present disclosure.

Detailed Description

Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and should be considered as merely exemplary. Accordingly, one of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

The main methods for identifying the road conditions currently in common are as follows:

1. a sensor-based method: by deploying sensor devices such as vehicle detectors, cameras, radars and the like on roads, traffic data are collected in real time, and analysis and recognition of road conditions are performed by using the data.

2. Deep learning method based on feature processing: according to different road grades, the congestion conditions (smooth, slow and congestion) are defined based on the passing speed, so that the road conditions are identified.

The two ways have different problems in application:

1. sensor-based methods, where the cost of deploying and maintaining the sensor device is high, and a significant amount of manpower and time is required for data processing and analysis, and coverage is low, cannot be applied in a large scale, and the acquisition device may be obscured by the lead vehicle, and no effective image is acquired.

2. The deep learning method based on feature processing generally relies on defining congestion conditions through features such as speed, extracting statistical features in user track information, and finally training through a time sequence model. However, conventional schemes ignore incomplete data processing of the direction tags when building training data. The missing data can lead to unbalance and incompleteness of the training data set, and it is difficult to accurately learn the road condition prediction at a certain traffic light head.

In order to solve the problems, the disclosure provides a road condition prediction method, a device, equipment and a storage medium, which are applied to an artificial intelligence technology, in particular to the technical field of automatic driving, so as to provide road condition information for vehicles in time, help the vehicles to select an optimal driving strategy and improve driving efficiency and safety.

The traffic management department and related institutions can take targeted measures, such as dispatching traffic police or adjusting traffic control measures, so as to reduce accident risks and improve traffic safety; the road condition prediction result can also be applied to an intelligent traffic management system, so that intelligent control and coordination of traffic signals are realized, and traffic fluency and efficiency are improved.

Fig. 1 is a flowchart of a road condition prediction method according to an embodiment of the present disclosure, as shown in fig. 1, the road condition prediction method provided by the present disclosure includes the following method steps:

s101, responding to a vehicle entering a traffic light influence road section, and acquiring historical vehicle track information and traffic light state information of the traffic light influence road section in a preset period;

s102, determining queue length information of a traffic light influence queue corresponding to the traffic light influence road section according to the historical vehicle track information and the traffic light state information;

s103, responding to the parking operation of the vehicle, and predicting and obtaining the road condition information of the traffic light influence road section according to the queue length information of the traffic light influence queue.

Alternatively, in examples of the present disclosure, the vehicle may be any type or model of vehicle, such as an autonomous vehicle, or the like.

Optionally, in the example of the present disclosure, the historical vehicle track information and the traffic light status information may reflect the traffic flow change and the traffic light change of the traffic light affected road section in different time periods, so as to provide valuable reference information for road condition prediction. For example, if the historical data shows that a traffic light affects a road section with a larger traffic flow in an early peak period and a longer red light time, it can be predicted that the road section has a worse road condition in the early peak period, and congestion may occur.

Optionally, in the example of the present disclosure, the queue length information of the traffic light influence queue may indicate the queuing conditions of vehicles of the traffic light influence road segments in different periods, so as to provide an intuitive index for road condition prediction. For example, if the queue length information of the traffic light influence queue indicates that the queue length of a certain traffic light influence road section is longer in a certain period, it can be predicted that the road condition of the road section is worse in the period, and a parking waiting phenomenon may occur.

In this disclosure, the parking operation of the vehicle may indicate whether the vehicle enters a traffic light affecting road section and/or reaches an intersection where a traffic light is provided, for example, the parking operation may be parking caused by excessive vehicles and traffic jams on the traffic light affecting road section, or the parking operation may also be parking caused by reaching the intersection where the traffic light is provided in response to parking caused by a red light, so as to provide a trigger condition for road condition prediction.

According to the queue length information of the traffic light influence queue, the road condition information of the traffic light influence road section, such as the speed, the traffic flow, the congestion degree and the like, can be predicted, so that useful prompt information is provided for vehicles. For example, if the road condition prediction information shows that a certain traffic light affects a low speed of a road section and the traffic flow is large, the vehicle can be prompted to slow down in the road section or select other routes to bypass, so as to avoid traffic accidents or waste of time.

According to the road condition prediction method, the historical vehicle track data can be effectively utilized by acquiring the historical vehicle track information and the traffic light state information of the traffic light influence road section in the preset time period, and the accuracy and the reliability of road condition prediction are improved. According to historical vehicle track information and traffic light state information, the queue length information of the traffic light influence queues is determined, traffic flow conditions of road sections affected by the traffic lights can be accurately reflected, and important basis is provided for road condition prediction. And then, according to the queue length information of the traffic light influence queue, the road condition information of the traffic light influence road section is predicted and obtained by responding to the parking operation of the vehicle, so that the road condition information can be provided for the vehicle in time, the vehicle is helped to select the optimal driving strategy, and the driving efficiency and the driving safety are improved.

Optionally, in an example, determining, according to the historical vehicle track information and the traffic light state information, queue length information of a traffic light influence queue corresponding to the traffic light influence road section includes:

Optionally, the historical parking information may indicate a historical parking position and parking time of the vehicle on the traffic light affected road section, so as to provide accurate data for determining the traffic light affected queue. For example, if the parking information shows that a certain vehicle has been parked for 5 seconds at a certain coordinate point, it may be determined that the vehicle is affected by a traffic light at the coordinate point, thereby bringing the vehicle into a range of a traffic light impact queue.

Optionally, the traffic light influence queue may represent the distribution and arrangement order of vehicles on the traffic light influence road section, so as to provide an intuitive index for road condition prediction. The queue length information of the traffic light influence queue is the longest length of the traffic light influence queue, and the number of vehicles and occupied space on a traffic light influence road section can be represented, so that a quantized index is provided for road condition prediction. For example, if the traffic light influence queue indicates that 10 vehicles are arranged in a certain order on a certain traffic light influence road section, and the queue length information of the traffic light influence queue indicates that the length of the queue is 50 meters, it can be determined that the traffic flow on the road section is large and the distance between the vehicles is small, so that it is predicted that the road condition on the road section is poor and a congestion phenomenon may occur.

In the above optional examples, the present disclosure may effectively identify parking operations on traffic light impact road segments by acquiring historical parking information in historical vehicle track information, and provide necessary information for determining traffic light impact queues. According to the parking information and the traffic light state information, the traffic light influence queues corresponding to the traffic light influence road sections and the queue length information of the traffic light influence queues are determined, so that the vehicle queuing conditions on the traffic light influence road sections can be accurately determined, and important basis is provided for road condition prediction.

There is also an optional example, predicting, according to the queue length information of the traffic light influencing queue, the road condition information of the traffic light influencing road section, including:

inputting the queue length information of the traffic light influence queue into a road condition prediction model obtained by training in advance to obtain the road condition information of the traffic light influence road section output by the road condition prediction model;

the road condition prediction model is obtained by training an initial neural network model in advance, and each pair of enhanced training data for training comprises: the sample traffic light influences the queue length information of the queue and the corresponding traffic light influences the sample road condition information of the road section.

Optionally, according to the sample track information of the vehicle, the speed, the acceleration, the parking times, the parking time, the queue length information of the queue affected by the traffic lights through the traffic lights, the steering prediction, the sample traffic lights and the like feature data can be calculated, the road attribute information such as the traffic light period and the like is calculated, and the final sample road condition information is obtained through regularization operation.

Optionally, by the above example, the machine learning method can be effectively utilized, so as to improve accuracy and intelligence of road condition prediction. The generalization capability and the robustness of the road condition prediction model can be effectively improved by training the road condition prediction model obtained by training the initial neural network model in advance.

Optionally, in the example of the present disclosure, the road condition prediction model may predict road condition information of a road section affected by the traffic light, such as a vehicle speed, a vehicle flow, a congestion degree, etc., according to queue length information of a queue affected by the traffic light, so as to provide useful prompt information for a vehicle. For example, if the road condition prediction model predicts that the speed of the road section affected by the traffic light in a certain period is 20 km/h, the traffic flow is 100 km/min, and the congestion degree is high according to the queue length information of the traffic light affected queue in the certain period, the vehicle can be prompted to slow down or select other routes to bypass in the road section, so as to avoid traffic accidents or time waste.

Optionally, the enhanced training data may include queue length information of various sample traffic light influence queues and sample road condition information of corresponding traffic light influence road sections, so as to provide abundant data resources for road condition prediction. For example, if the enhanced training data includes a combination of various factors such as different traffic light influencing road segments, different time periods, different traffic flows, different traffic light states, and the like, the initial neural network model can learn more features and rules, so that generalization capability and robustness of the road condition prediction model are improved.

In an example, as shown in fig. 2a, in the training process, the road condition prediction model learns the time sequence information of each pair of labeled training data, and adopts a plurality of classifiers to estimate the road condition information in different driving directions based on each pair of labeled training data.

Optionally, the time sequence information may indicate a time sequence and a time interval of each pair of labeled training data, so as to provide important features for road condition prediction. The multiple classifiers can predict the road condition information in different driving directions according to the road condition information in different driving directions, so that multiple options are provided for road condition prediction.

For example, if the time sequence information shows that the time interval of a certain pair of labeled training data is 5 minutes, it can be inferred that the road condition information of the pair of labeled training data has a certain correlation, thereby providing a useful clue for road condition prediction; if the multiple classifiers respectively predict the east-west and north-south road condition information of a road section affected by a certain traffic light, the most suitable road condition information can be selected according to the running direction of the vehicle, so that an optimal result is provided for road condition prediction. For example, three expert networks are used for respectively learning the characteristics of different turning at the traffic lights, so that the road condition prediction (smooth-creep-congestion) at the traffic light intersection (straight-left-right turning) is finally obtained, and the model training uses cross entropy to calculate the loss, so that the loss is continuously reduced through back propagation.

By adopting the training method, the road condition prediction model learns the time sequence information of each pair of labeled training data in the training process, and the road condition information in different driving directions is obtained by adopting a plurality of classifiers based on the prediction of each pair of labeled training data, so that the accuracy and the flexibility of the road condition prediction model can be effectively improved.

In another example, as shown in fig. 2a, in the training process, the road condition prediction model predicts and scores each pair of unlabeled training data, and fuses each pair of unlabeled training data meeting the scoring threshold with the road condition information in the different driving directions to obtain each pair of enhanced training data.

For example, after the information such as the time when the vehicle passes through the traffic light, the state of the traffic light (such as green light, red light period), the speed of the vehicle, the time when the vehicle passes through, etc., because in the actual training data, the traffic information in three directions does not exist at the lamp cap of each traffic light at the same time. Therefore, training data can be enhanced, namely, data with labels missing in a certain direction is supplemented, and road conditions of the real world can be more accurately depicted. Thus achieving the effect of data enhancement.

By the aid of the training method, the label-free training data can be effectively utilized, and generalization capability and robustness of the road condition prediction model are improved.

Optionally, the unlabeled training data may include more queue length information of the traffic light influencing queues and road condition information of corresponding traffic light influencing road sections, so as to provide abundant data resources for road condition prediction. The prediction and scoring can give out corresponding evaluation according to the quality and the credibility of the unlabeled training data, thereby providing effective screening conditions for road condition prediction.

Optionally, the fusion can generate more perfect and fine training data according to the unlabeled training data and road condition information (smooth-inching-congestion) in different driving directions (straight-going-left-turning-right-turning), so as to provide more powerful algorithm support for road condition prediction. For example, if the unlabeled training data includes a combination of various factors such as different traffic light influencing road segments, different time periods, different traffic flows, different traffic light states, etc., the road condition prediction model can learn more features and rules, so that the generalization capability and robustness of the road condition prediction model are improved; if the prediction and scoring gives a score of 0.8 for a pair of unlabeled training data exceeding a set scoring threshold of 0.7, then the quality and reliability of the pair of unlabeled training data may be considered higher, thereby incorporating it into the range of enhanced training data; if the pair of unlabeled training data is combined with road condition information in different driving directions through fusion, each pair of enhanced training data is obtained, the road condition prediction model can learn more perfect and detailed training data, and therefore accuracy and flexibility of the road condition prediction model are improved.

In another example, for the data of the road condition label missing in a certain direction, the road condition of the direction is predicted, and the output of the model is a score of three road conditions, such as: smooth, slow going, congestion ] - [0.1,0.15,0.75], with congestion as a result. When scoring >0.7, the predicted result is used as a direction label of the missing label for training of the subsequent model. Otherwise, the piece of data is temporarily not used for model training.

According to the embodiment of the disclosure, the traffic light road condition information in a certain direction is subjected to data enhancement by constructing heterogeneous traffic light road condition training data. Firstly, constructing heterogeneous traffic light road condition training data which comprises traffic flow state labels (some direction labels are missing) of roads in different directions; training an initial predictive model using the data with complete labels; predicting the data of the missing tag by using an initial model; then, using the complete tag data and random sampling to obtain the supplemented training data for model enhancement training; the prediction and enhancement training steps are repeated, the prediction performance of the model can be gradually improved, and the enhanced traffic light road condition prediction model is finally obtained. The scheme improves the adaptability of the model to heterogeneous scenes through data iteration enhancement, is suitable for complex and diverse actual road traffic scenes, and is not only a certain fixed traffic flow.

Optionally, in this disclosure example, road condition information of a road section affected by the traffic light, such as a vehicle speed, a vehicle flow, and a congestion degree, may be predicted according to the historical vehicle track information and the queue length information of the traffic light affecting queue, so as to provide useful prompt information for the vehicle. For example, if the predicted road condition information shows that a certain traffic light affects a low speed of a road section and the traffic flow is large, the vehicle can be prompted to slow down in the road section or select other routes to bypass, so as to avoid traffic accidents or waste of time.

As an optional example, by responding to the parking operation of the vehicle and predicting the road condition information of the road section affected by the traffic light according to the historical vehicle track information and the queue length information of the traffic light influence queue, the real-time state of the vehicle can be timely fed back, and dynamic update can be provided for road condition prediction. In addition, according to the traffic flow situation of the road section affected by the traffic light is abstracted into one queue, so that complexity of road condition prediction is simplified, sensitivity and instantaneity of road condition prediction are enhanced, influence of the traffic light on the traffic flow can be accurately depicted, and important characteristics are provided for road condition prediction.

An optional example, predicting, according to the historical vehicle track information and the queue length information of the traffic light influence queue, the road condition information of the traffic light influence road section, includes:

and inputting the historical vehicle track information and the queue length information of the traffic light influence queue into a pre-training model built in advance, so as to predict and obtain the road condition information of the traffic light influence road section according to the historical vehicle track information and the queue length information of the traffic light influence queue by adopting the pre-training model.

The pre-training model is pre-built based on a mask prediction task and a time sequence prediction task, the mask prediction task is used for randomly masking part of contents in each set of historical vehicle track information in the historical vehicle track information, so that the pre-training model learns local dependency relations in each set of historical vehicle track information by predicting the masked part of contents, and the part of contents comprise: the traffic light affects nodes in the queue, and the time corresponding to each node in the queue is affected by the traffic light; the time sequence prediction task is used for predicting the time corresponding to the time when the historical vehicle track of the individual group passes through the node in the traffic light influence queue by adopting the historical vehicle track information of the individual group and the corresponding traffic light influence queue.

As shown in fig. 2b, by inputting the historical vehicle track information and the traffic light influence queue into a pre-built pre-training model (which is constructed based on pre-training data), the road condition information of the traffic light influence road section is predicted and obtained by adopting the pre-training model according to the historical vehicle track information and the queue length information of the traffic light influence queue, so that the strong data representation capability and generalization capability of the pre-training model can be effectively utilized, and a more efficient data processing method is provided for road condition prediction. Therefore, the method and the device can avoid the situation that a new model is built and trained again, and the traffic light is directly predicted by using the fine-tuned pre-training model to obtain the road condition information of the road section, so that time and resources are saved, and meanwhile, the road condition prediction efficiency and quality are improved.

According to the embodiment of the disclosure, different levels and parts of the pre-training model are respectively trained by utilizing different targets and difficulties of the mask prediction task and the time sequence prediction task, so that the complexity and diversity of the pre-training model are improved, and meanwhile, the accuracy and stability of road condition prediction are also enhanced. By pre-building a pre-training model based on a mask prediction task and a time sequence prediction task, local dependency and time sequence dependency in historical vehicle track information can be accurately learned, and deeper data learning characteristics are provided for road condition prediction.

In the embodiment of the disclosure, partial contents (such as nodes in a traffic light queue and time passing through the nodes in the traffic light queue) are randomly covered, so that the pre-training model is forced to learn local dependency relations in track data, and real data loss and noise are simulated, thereby improving the robustness and adaptability of the pre-training model and simultaneously improving the reliability and sensitivity of road condition prediction; the self-learning capacity and the self-error correction capacity of the pre-training model can be effectively enhanced, and a more stable data learning method is provided for road condition prediction.

In the embodiment of the disclosure, by adopting the vehicle track information of the individual group and the queue length information of the traffic light influence queue, the time corresponding to each node in the traffic light influence queue when the vehicle track of the individual group passes through is predicted, so that the time sequence change of the vehicle track can be accurately captured, and more sensitive data change characteristics are provided for road condition prediction. The time sequence information of the vehicle track is predicted, so that the running speed and time consumption of the vehicle on a road section affected by traffic lights are reflected, the time sequence prediction capability of the pre-training model is improved, and meanwhile, the dynamic property and the real-time property of road condition prediction are also improved.

According to the traffic light scene traffic prediction method and device, traffic tasks are pre-trained by utilizing massive road information and user track information, a more general pre-training model is self-supervised learning training by using large-scale non-labeling data, and the model is subjected to off-line fine tuning on a downstream traffic light scene traffic prediction task in a transfer learning mode, so that accurate recognition of traffic conditions is achieved, the limitation of manual feature design is overcome, and traffic features of traffic light scenes are more comprehensively and adaptively captured.

In an optional example, the pre-training model marks the historical road condition information corresponding to the historical vehicle track information according to the historical parking information in the historical vehicle track information, the traffic light state information and the time corresponding to each node in the traffic light influence queue.

Optionally, in the disclosure, the pre-training model marks the historical road condition information corresponding to the historical vehicle track information according to the historical parking information and the traffic light state information in the historical vehicle track information and by affecting the time corresponding to each node in the queue through the traffic light, so that the self-supervision capability of the pre-training model can be effectively utilized, and a higher-quality data marking method is provided for road condition prediction. Therefore, the cost and the error of manual marking can be avoided, and the output of the pre-training model is directly used as a label of the historical road condition information, so that time and resources are saved, and meanwhile, the accuracy and the reliability of road condition prediction are improved.

In an alternative example, the training data of the pre-training model includes: historical vehicle track information and corresponding queue length information of traffic light influence queues; in the training process of the pre-training model, the parameters of the pre-training model are adopted to encode training data, so that encoded training data are obtained; enhancing the encoded training data by adopting a time sequence model to obtain time sequence enhanced training data; and estimating road condition information in different driving directions by adopting a plurality of classifiers based on the training data after the time sequence enhancement.

In the disclosed example, the pretrained model is trimmed in the training process of the pretrained model in the plurality of optional manners: the training data is encoded by adopting the parameters of the pre-training model to obtain the encoded training data, the data representation capability of the pre-training model can be effectively utilized, a more efficient data conversion method is provided for road condition prediction, the training data can be converted into the internal representation of the pre-training model from the original vehicle track information and traffic light influence queue, and therefore the dimension and complexity of the data are reduced, and meanwhile the processibility and the interpretability of the data are improved.

Further, in the disclosed example, the time sequence model is adopted to enhance the encoded training data, so that the time sequence information of the labeled training data is learned, the time sequence dependency relationship in the training data can be accurately learned, and deeper data learning characteristics are provided for road condition prediction. Therefore, the time sequence prediction capability of the time sequence model can be utilized to extract and enhance the time sequence information of the encoded training data, so that the time sequence sensitivity and the time sequence integrity of the training data are improved, and the dynamics and the instantaneity of road condition prediction are enhanced.

Further, in the disclosed example, the road condition information in different driving directions is estimated by using a plurality of classifiers based on the training data after time sequence enhancement, so that a plurality of dimensions and aspects in the training data can be comprehensively considered, and more comprehensive data comprehensive characteristics can be provided for road condition prediction. And the multi-classification capability of a plurality of classifiers can be utilized to classify the training data with multiple dimensions and aspects after the time sequence is enhanced, so that the classification precision and the classification coverage rate of the training data are improved, and meanwhile, the accuracy and the stability of road condition prediction are also improved.

In an alternative example, as shown in fig. 3, acquiring the historical parking information in the historical vehicle track information includes:

s301, grouping the historical vehicle track information to obtain a plurality of groups of historical vehicle track information, wherein each group of historical vehicle track information corresponds to a period;

s302, parking detection is carried out on each group of historical vehicle track information so as to determine parking operation of the traffic light influence road section in each period;

s303, according to the parking operation of the traffic light influence road section in each period, determining the historical vehicle parking information corresponding to each parking operation.

Alternatively, in examples of the present disclosure, multiple sets of historical vehicle track information may represent the distribution and variation of the historical vehicle track information over different time periods, thereby providing valuable reference information for parking detection. For example, if multiple sets of historical vehicle track information indicate that a certain traffic light affects more vehicle track information for a road segment in an early rush hour and less vehicle track information for a late rush hour, it may be inferred that the road segment has a greater traffic flow for the early rush hour and a lesser traffic flow for the late rush hour, thereby providing a useful clue to parking detection.

Alternatively, in examples of the present disclosure, the parking operation may indicate when and where the vehicle is parked on the traffic light affecting road segment, thereby providing accurate data for determining vehicle parking information. For example, if the parking detection shows that a certain vehicle has been parked for 5 seconds at a certain coordinate point of a certain period, it may be determined that the vehicle has undergone a parking operation at the certain coordinate point of the period, thereby bringing the parking operation into the range of the vehicle parking information.

Optionally, in the example of the disclosure, the vehicle parking information may represent a historical parking position and parking time of the vehicle on the traffic light affected road section, so as to provide accurate data for determining the traffic light affected queue. For example, if the vehicle parking information shows that a certain vehicle has been parked for 5 seconds at a certain coordinate point of a certain period, it may be determined that the vehicle is affected by a traffic light at the certain coordinate point of the period, thereby bringing the vehicle into a range of a traffic light impact queue.

In the embodiment of the disclosure, first, by grouping the historical vehicle track information to obtain multiple sets of historical vehicle track information, the historical vehicle track information can be effectively divided according to the time dimension, and a condition for facilitating parking detection is provided. And then, by carrying out parking detection on each group of historical vehicle track information to determine the parking operation of the traffic light influence road section in each period, the parking condition on the traffic light influence road section can be effectively identified, and necessary information is provided for determining the vehicle parking information. Finally, by determining historical vehicle parking information corresponding to each parking operation according to the parking operation of the traffic light influence road section in each period, the vehicle parking condition on the traffic light influence road section can be effectively determined, and important information is provided for determining the traffic light influence queue.

As an optional example, as shown in fig. 4, determining, according to the historical parking information and the traffic light status information, a traffic light impact queue corresponding to the traffic light impact road section includes:

s401, performing nuclear density estimation on each set of historical vehicle track information in a plurality of sets of historical vehicle track information respectively to obtain probability density distribution values of the historical vehicle passing through each coordinate point in a period corresponding to each set of historical vehicle track information;

S402, comparing the probability density distribution values of each of a plurality of continuous time periods to determine the traffic flow change rule of the traffic light influence road section in different time periods;

s403, determining the traffic light influence queues corresponding to the traffic light influence road sections in different time periods according to the traffic flow change rules, the historical parking information and the traffic light state information in different time periods.

Optionally, in the example of the present disclosure, the probability density distribution value may represent a probability of occurrence of the vehicle on the traffic light affected road section, so as to provide quantized data for determining a traffic flow change rule. For example, if the probability density distribution value shows a higher probability of occurrence of a vehicle at a certain coordinate point, it may be judged that the traffic flow at that coordinate point is large, thereby providing a useful clue for determining the traffic flow change law.

Optionally, in the example of the present disclosure, the traffic flow change rule may represent a trend of change of the traffic flow and the vehicle speed on the traffic light affecting road section, so as to provide valuable reference information for determining the traffic light affecting queue. For example, if the traffic flow change law shows that the traffic flow and the speed of a traffic light affecting road segment are low in a certain period and high in a next period, it can be inferred that the traffic light affecting road segment is affected by the traffic light in the period, thereby providing a useful clue for determining the traffic light affecting queue.

Optionally, the traffic light influence queue may represent the distribution and arrangement order of vehicles on the traffic light influence road section, so as to provide an intuitive index for road condition prediction. For example, if the traffic light influence queue shows that 10 vehicles on a certain traffic light influence road section are arranged in a certain order over a certain period of time, it can be judged that the traffic flow on the road section is large and the distance between the vehicles is small, so that it is predicted that the road condition on the road section is poor and a congestion phenomenon may occur.

In the disclosed example, first, kernel density estimation is performed on each set of historical vehicle track information in multiple sets of historical vehicle track information, so as to obtain probability density distribution values of historical vehicles passing through each coordinate point in a period corresponding to each set of historical vehicle track information, so that vehicle distribution conditions on road sections affected by traffic lights can be effectively described, and necessary information is provided for determining a traffic flow change rule. And then, comparing probability density distribution values of each of a plurality of continuous time periods, determining the traffic flow change rule of the traffic light influence road section in different time periods, effectively analyzing the traffic flow dynamic condition of the traffic light influence road section, and providing important information for determining the traffic light influence queue. According to the traffic flow change rules, the parking information and the traffic light state information of different time periods, the traffic light influence queues corresponding to the traffic light influence road sections in different time periods are determined, the vehicle queuing conditions on the traffic light influence road sections can be accurately determined, and important basis is provided for road condition prediction.

Optionally, as shown in fig. 5, in one example, determining the traffic light impact queue corresponding to the traffic light impact road section in different periods according to the traffic flow change rule, the historical parking information and the traffic light status information in different periods includes:

s501, screening out target historical parking information influenced by traffic lights in each period according to the traffic light state information in different periods;

s502, determining a first traffic light influence queue of the traffic light influence road section in each time period according to the target historical parking information in each time period;

s503, determining a second traffic light influence queue of the traffic light influence road section in each time period according to the traffic flow change rules of different time periods;

and S504, correspondingly fusing the first traffic light influence queue and the second traffic light influence queue in each time period to obtain the traffic light influence queues corresponding to different time periods of the traffic light influence road sections.

Optionally, in the example of the present disclosure, the target historical parking information may indicate a historical parking position and parking time of the vehicle on the traffic light affected road section, and is matched with the traffic light status information, so as to provide accurate data for determining the first traffic light affected queue. For example, if the traffic light status information shows that the red light time in a certain period is 10 seconds, the parking information with the parking time greater than or equal to 10 seconds in the certain period can be screened out and used as the target historical parking information, so that the parking information with other parking time less than 10 seconds is excluded to avoid interference.

Optionally, in the example of the present disclosure, the first traffic light influence queue may represent a vehicle distribution and a ranking sequence on a traffic light influence road section, and is matched with traffic light state information, so as to provide an intuitive index for determining the traffic light influence queue. For example, if the first traffic light influencing queue shows that 10 vehicles are arranged in a certain order within a certain period of time and are matched with the red light state information, it may be determined that the vehicles on the road section are influenced by the red light, so that the queue is included in the range of the traffic light influencing queue.

Optionally, in the example of the present disclosure, the traffic light influence queue may represent a vehicle distribution and arrangement sequence on a traffic light influence road section, and is matched with traffic light state information and a traffic flow change rule, so as to provide the most accurate index for road condition prediction. For example, if the traffic light influence queue shows that 10 vehicles are arranged according to a certain sequence within a certain period and are matched with traffic light state information and a traffic flow change rule, it can be judged that the vehicles on the road section are comprehensively influenced by traffic lights and traffic flows, so that the road condition on the road section is predicted to be worst, and serious congestion can occur.

Optionally, in the example of the present disclosure, the second traffic light influence queue may represent a vehicle distribution and arrangement sequence on the traffic light influence road section, and is matched with a traffic flow change rule, so as to provide valuable reference information for determining the traffic light influence queue. For example, if the second traffic light influencing queue shows that 10 vehicles are arranged in a certain order within a certain period of time and match with the traffic flow change rule, it can be inferred that the vehicles on the road section are influenced by the traffic flow, so that the queue is included in the range of the traffic light influencing queue.

Through the disclosed example, the disclosed example screens out the target historical parking information influenced by the traffic light in each period according to the traffic light state information in different periods, can effectively filter out irrelevant parking information, and provides necessary information for determining the first traffic light influence queue. According to the target historical parking information in each time period, a first traffic light influence queue of a traffic light influence road section in each time period is determined, the vehicle queuing situation on the traffic light influence road section can be accurately determined, and important information is provided for determining the traffic light influence queue. According to the traffic flow change rules of different time periods, the second traffic light influence queues of the traffic light influence road sections in each time period are determined, the traffic flow dynamic conditions on the traffic light influence road sections can be effectively analyzed, and important information is provided for determining the traffic light influence queues. And correspondingly fusing the first traffic light influence queue and the second traffic light influence queue in each time period to obtain the traffic light influence queues corresponding to the traffic light influence road sections in different time periods, so that the influence of the traffic light and the traffic flow can be effectively and comprehensively considered, and an optimal basis is provided for road condition prediction.

As an alternative example, as shown in fig. 6, determining a first traffic light impact queue of the traffic light impact road section in each period according to the target historical parking information in each period includes:

s601, acquiring a plurality of historical parking positions in the target historical parking information in each period;

s602, generating the first traffic light influence queue of the traffic light influence road section in each period according to the plurality of historical parking positions.

Optionally, the first traffic light influence queue may represent a vehicle distribution and arrangement sequence on the traffic light influence road section, and is matched with the historical parking position, so as to provide an intuitive index for determining the traffic light influence queue. For example, if the first traffic light influencing queue shows that 10 vehicles are arranged in a certain order within a certain period of time and are matched with the historical parking positions, it may be determined that the vehicles on the road section are influenced by the red light, so that the queue is included in the range of the traffic light influencing queue.

Alternatively, the historical parking location may represent the parking coordinates of the vehicle on the traffic light affecting road segment, thereby providing accurate data for generating the first traffic light affecting queue. For example, if the historical parking location shows that a certain vehicle is parked at a certain coordinate point of a certain period, it may be determined that the vehicle is a parking spot at the certain coordinate point of the period, thereby bringing the parking spot into range of the first traffic light influence queue.

In the above disclosed example, by acquiring a plurality of historical parking positions in the target historical parking information in each period, parking points on the traffic light influence road section can be effectively determined, and necessary information is provided for generating the first traffic light influence queue. And then, generating a first traffic light influence queue of the traffic light influence road section in each period according to a plurality of historical parking positions, so that the vehicle queuing situation on the traffic light influence road section can be accurately generated, and important information is provided for determining the traffic light influence queue.

By adopting the disclosed example, more than half map navigation users can meet the traffic light in the navigation process, each user can pass through the traffic light more than 20 times every day on average, the map navigation users can accurately judge the road condition under the traffic light scene by passing through the traffic light more than 5 hundred million times every day, more comfortable user experience can be brought to the users, and the overall public praise of the users to the map is improved.

In the technical scheme of the disclosure, the related processes of collecting, storing, using, processing, transmitting, providing, disclosing and the like of the personal information of the user accord with the regulations of related laws and regulations, and the public order colloquial is not violated.

Fig. 7 is a schematic frame diagram of a road condition prediction device according to an embodiment of the present disclosure, and as shown in fig. 7, the present disclosure further provides a road condition prediction device 700, including:

an obtaining unit 701, configured to obtain historical vehicle track information and traffic light state information of a traffic light affected road section within a predetermined period in response to a vehicle entering the traffic light affected road section;

a determining unit 702, configured to determine queue length information of a traffic light influence queue corresponding to the traffic light influence road section according to the historical vehicle track information and the traffic light state information;

a prediction unit 703, configured to predict and obtain the road condition information of the traffic light affected road section according to the queue length information of the traffic light affected queue in response to the parking operation of the vehicle.

According to one or more optional examples of the disclosure, the determining unit includes:

According to one or more optional examples of the disclosure, the acquiring subunit includes:

the detection module is used for carrying out parking detection on each group of historical vehicle track information so as to determine parking operation of the traffic light influence road section in each period;

and the first determining module is used for determining historical vehicle parking information corresponding to each parking operation according to the parking operation of the traffic light influence road section in each period.

According to one or more optional examples of the disclosure, the determining subunit includes:

According to one or more optional examples of the disclosure, the third determining module further includes:

and the fusion submodule is used for correspondingly fusing the first traffic light influence queue and the second traffic light influence queue in each time period to obtain the traffic light influence queues corresponding to different time periods of the traffic light influence road sections.

According to one or more optional examples of the disclosure, the first determining submodule is specifically further configured to:

According to one or more optional examples of the disclosure, the prediction unit includes:

the first prediction subunit is used for inputting the queue length information of the traffic light influence queue into a road condition prediction model obtained through training in advance so as to obtain the road condition information of the traffic light influence road section output by the road condition prediction model; the road condition prediction model is obtained by training an initial neural network model in advance, and each pair of enhanced training data for training comprises: the sample traffic light influences the queue length information of the queue and the corresponding traffic light influences the sample road condition information of the road section.

According to one or more optional examples of the disclosure, the road condition prediction model learns time sequence information of each pair of labeled training data in a training process, and adopts a plurality of classifiers to estimate road condition information in different driving directions based on each pair of labeled training data; and in the training process of the road condition prediction model, predicting and scoring each pair of unlabeled training data, and fusing each pair of unlabeled training data meeting a scoring threshold with road condition information in different driving directions to obtain each pair of enhanced training data.

and the second prediction subunit is used for predicting the road condition information of the traffic light influence road section according to the historical vehicle track information and the queue length information of the traffic light influence queue.

According to one or more optional examples of the disclosure, the above second prediction subunit is specifically further configured to:

the pre-training model is pre-built based on a mask prediction task and a time sequence prediction task, the mask prediction task is used for randomly masking part of contents in each set of historical vehicle track information in the historical vehicle track information, so that the pre-training model learns local dependency relations in each set of historical vehicle track information by predicting the masked part of contents, and the part of contents comprise: each node in the traffic light influence queue and the corresponding time of each node in the queue is influenced by the traffic light; the time sequence prediction task is used for predicting the time corresponding to each node in the traffic light influence queue when the historical vehicle track of the individual group passes through the traffic light influence queue by adopting the historical vehicle track information of the individual group and the corresponding traffic light influence queue.

According to one or more optional examples of the disclosure, the pre-training model marks historical road condition information corresponding to the historical vehicle track information according to historical parking information in the historical vehicle track information, the traffic light state information and time corresponding to each node in the traffic light influence queue.

According to one or more optional examples of the disclosure, the training data of the pre-training model includes: the historical vehicle track information and the corresponding queue length information of the traffic light influence queue;

According to embodiments of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

According to an embodiment of the present disclosure, there is provided a non-transitory computer-readable storage medium storing computer instructions for causing a computer to perform a method according to any one of the above.

According to an embodiment of the present disclosure, there is provided a computer program product comprising: a computer program stored in a readable storage medium, from which at least one processor of an electronic device can read, the at least one processor executing the computer program causing the electronic device to perform the solution provided by any one of the embodiments described above.

The present disclosure also provides an electronic device, according to an embodiment of the present disclosure, fig. 8 shows a schematic block diagram of an example electronic device 800 that may be used to implement an embodiment of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 8, the apparatus 800 includes a computing unit 801 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) 802 or a computer program loaded from a storage unit 808 into a Random Access Memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other by a bus 804. An input/output (I/O) interface 805 is also connected to the bus 804.

Various components in device 800 are connected to I/O interface 805, including: an input unit 806 such as a keyboard, mouse, etc.; an output unit 807 such as various types of displays, speakers, and the like; a storage unit 808, such as a magnetic disk, optical disk, etc.; and a communication unit 809, such as a network card, modem, wireless communication transceiver, or the like. The communication unit 809 allows the device 800 to exchange information/data with other devices via a computer network such as the internet and/or various telecommunication networks.

The computing unit 801 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of computing unit 801 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, etc. The calculation unit 801 performs the respective methods and processes described above, such as the road condition prediction method. For example, in some embodiments, the road condition prediction method may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program may be loaded and/or installed onto device 800 via ROM 802 and/or communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the road condition prediction method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform the road condition prediction method in any other suitable way (e.g. by means of firmware).

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), complex Programmable Logic Devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus such that the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user; and a keyboard and pointing device (e.g., a mouse or trackball) by which a user can provide input to the computer. Other kinds of devices may also be used to provide for interaction with a user; for example, feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form, including acoustic input, speech input, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), and the internet.

The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, also called a cloud computing server or a cloud host, and is a host product in a cloud computing service system, so that the defects of high management difficulty and weak service expansibility in the traditional physical hosts and VPS service ("Virtual Private Server" or simply "VPS") are overcome. The server may also be a server of a distributed system or a server that incorporates a blockchain.

It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps recited in the present disclosure may be performed in parallel or sequentially or in a different order, provided that the desired results of the technical solutions of the present disclosure are achieved, and are not limited herein.

The above detailed description should not be taken as limiting the scope of the present disclosure. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present disclosure are intended to be included within the scope of the present disclosure.

Claims

1. A road condition prediction method, the method comprising:

2. The method of claim 1, wherein determining the queue length information of the traffic light impact queue corresponding to the traffic light impact road segment according to the historical vehicle track information and the traffic light status information comprises:

3. The method of claim 2, wherein obtaining historical parking information in the historical vehicle track information comprises:

4. The method of claim 3, wherein determining a traffic light impact queue corresponding to the traffic light impact road segment according to the historical parking information and the traffic light status information comprises:

5. The method of claim 4, wherein determining the traffic light impact queues corresponding to the traffic light impact road segments in different time periods according to the traffic flow change law, the historical parking information and the traffic light status information of different time periods comprises:

According to the traffic light state information of different time periods, screening target historical parking information influenced by traffic lights in each time period from the historical parking information of different time periods;

6. The method of claim 5, wherein determining a first traffic light impact queue for the traffic light impact road segment for each time period based on the target historical parking information for each time period, comprises:

7. The method according to any one of claims 1 to 6, wherein predicting the road condition information of the traffic light influencing road section according to the queue length information of the traffic light influencing queue comprises:

8. The method of claim 7, wherein,

and in the training process of the road condition prediction model, learning the time sequence information of each pair of labeled training data, and adopting a plurality of classifiers to estimate the road condition information in different driving directions based on each pair of labeled training data.

9. The method of claim 7, wherein,

and in the training process of the road condition prediction model, predicting and scoring each pair of unlabeled training data, and fusing each pair of unlabeled training data meeting a scoring threshold with road condition information in different driving directions to obtain each pair of enhanced training data.

10. The method of claim 1, wherein predicting the traffic information of the traffic light affected section according to the queue length information of the traffic light affected queue comprises:

11. The method of claim 10, wherein predicting the road condition information of the traffic light-affected road segment based on the historical vehicle track information and the queue length information of the traffic light-affected queue comprises:

12. The method of claim 11, wherein,

and marking the historical road condition information corresponding to the historical vehicle track information by the pre-training model according to the historical parking information in the historical vehicle track information, the traffic light state information and the time corresponding to each node in the traffic light influence queue.

13. The method of claim 11, wherein the training data of the pre-training model comprises: the historical vehicle track information and the corresponding queue length information of the traffic light influence queue;

14. A road condition prediction apparatus, the apparatus comprising:

15. The apparatus of claim 14, wherein the determining unit comprises:

16. The apparatus of claim 15, wherein the acquisition subunit comprises:

17. The apparatus of claim 16, wherein the determination subunit comprises:

18. The apparatus of claim 17, wherein the third determination module further comprises:

the screening sub-module is used for screening target historical parking information influenced by traffic lights in each time period from the historical parking information in different time periods according to the traffic light state information in different time periods;

19. The apparatus of claim 18, wherein the first determination submodule is further configured to:

20. The apparatus of any of claims 14 to 19, wherein the prediction unit comprises:

21. The apparatus of claim 20, wherein the road condition prediction model learns timing information of each pair of tagged training data during training, and obtains road condition information in different driving directions based on each pair of tagged training data by using a plurality of classifiers; and in the training process of the road condition prediction model, predicting and scoring each pair of unlabeled training data, and fusing each pair of unlabeled training data meeting a scoring threshold with road condition information in different driving directions to obtain each pair of enhanced training data.

22. The apparatus of claim 14, wherein the prediction unit comprises:

23. The apparatus of claim 22, wherein the second prediction subunit is further specifically configured to:

24. The apparatus of claim 23, wherein,

25. The apparatus of claim 23, wherein the training data of the pre-training model comprises: the historical vehicle track information and the corresponding queue length information of the traffic light influence queue;

26. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-13.

27. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1-13.

28. A computer program product comprising a computer program which, when executed by a processor, implements the steps of the method of any of claims 1-13.