Trailer real-time traffic information acquisition and processing system based on artificial intelligence

The trailer real-time traffic information system uses a neural network model to integrate multi-modal data for autonomous learning, addressing the limitations of rule-based systems by enhancing adaptability and safety in vehicle control.

CN119058729AActive Publication Date: 2024-12-03SHANDONG LIANGSHAN TONGYA AUTOMOBILE MFG CO LTD

Patent Information

Application Number
CN202411570391.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-11-06
Publication Date
2024-12-03
Estimated Expiration
2044-11-06

AI Technical Summary

Technical Problem

Existing vehicle control systems, particularly rule-based methods, struggle to adapt to complex and variable road and traffic conditions, leading to inadequate decision-making and control effectiveness.

Method used

A trailer real-time traffic information collection and processing system utilizing a feature-fused neural network model that integrates multi-modal inputs such as vehicle driving information, historical traffic data, and road images to predict future trajectories and accident probabilities, autonomously learning driving strategies without manual rule sets.

Benefits of technology

Enhances adaptability and safety in decision-making by accurately predicting future vehicle behaviors and optimizing driving strategies, effectively handling complex scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119058729A_ABST
    Figure CN119058729A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of intelligent traffic, and discloses a trailer real-time traffic information acquisition and processing system based on artificial intelligence, comprising the following steps: acquiring trailer driving information, acquiring trailer historical traffic data information, and acquiring observed road image information; the trailer driving information is coded into graph structure data, the graph structure data comprises nodes and edges connected with the nodes, and the trailer driving information of each node is coded into first sequence data; the historical traffic data information of the trailer is coded into second sequence data; and inputting the first sequence data, the second sequence data and the observed road image into the trailer prediction model, and outputting a trailer driving strategy. According to the invention, the path prediction of the trailer can be realized, the probability of abnormal events of the trailer can be judged according to historical trailer traffic information, and the driving efficiency and safety of the trailer are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and more specifically, to a trailer real-time traffic information collection and processing system based on artificial intelligence. Background Art

[0002] Existing technologies usually use rule-based methods to control vehicle driving. This method usually relies on a pre-defined set of rules and selects appropriate rules to guide vehicle driving according to the current road and traffic conditions. However, due to the complexity and variability of actual road and traffic conditions, rule-based methods often cannot cover all possible situations, resulting in poor control decision-making effects.

[0003] In order to solve the above problems, a data-driven approach has emerged in recent years, which uses historical data to train deep neural network models to predict and control vehicle driving. This method can autonomously learn the laws of vehicle driving from massive data without manually designing complex rule sets. By inputting data such as the current vehicle status, road conditions, and traffic information, the neural network model can directly output a reasonable driving strategy, showing stronger adaptability and decision-making quality. Summary of the invention

[0004] The present invention provides a trailer real-time traffic information collection and processing system based on artificial intelligence, which can comprehensively utilize multi-modal inputs such as vehicle driving information, historical traffic data and road images, accurately predict future vehicle driving trajectories and accident probabilities through a feature-fused neural network model, and output the optimal driving strategy, thereby improving the adaptability and safety of decision-making, and solving the technical problem of insufficient decision-making ability of rule-based methods in the prior art in complex scenarios.

[0005] The present invention provides a trailer real-time traffic information collection and processing system based on artificial intelligence, comprising the following steps:

[0006] The information collection module collects trailer driving information, including trailer information, other vehicle information, weather information, and road information; obtains trailer historical traffic data information; and obtains observed road image information;

[0007] The first encoding module encodes the trailer driving information into graph structure data, which includes nodes and edges connecting the nodes; encodes the trailer driving information of each node into a first sequence data, which includes n sequence units, and the tth sequence unit represents the trailer driving information of all nodes at the tth moment;

[0008] The second encoding module encodes the trailer historical traffic data information into second sequence data, the second sequence data includes m sequence units, and the sth sequence unit represents the trailer historical traffic data information at the sth moment;

[0009] The trailer prediction module inputs the No. 1 sequence data, the No. 2 sequence data, and the observed road image into the trailer prediction model. The trailer prediction model includes a first hidden layer, a second hidden layer, a first feature fusion layer, a first output layer, a third hidden layer, a second output layer, a fourth hidden layer, and a third output layer. The first hidden layer inputs the No. 1 sequence data, outputs the first hidden state to the second hidden layer, the second hidden layer outputs the second hidden state to the first feature fusion layer, the first feature fusion layer outputs the first fusion state to the first output layer, and outputs the trailer driving strategy; the third hidden layer inputs the No. 2 sequence data, outputs the third hidden state to the first feature fusion layer and the second output layer, and the second output layer outputs the probability of an abnormal event occurring to the trailer; the fourth hidden layer inputs the observed road image, outputs the fourth hidden state to the first feature fusion layer and the third output layer, and the third output layer outputs the positions of all vehicles at the next moment.

[0010] In a preferred embodiment, the calculation formula of the first hidden layer is as follows:

[0011] ;

[0012] in Represents the first hidden state of the vth node of the tth sequence unit of the first sequence data, is the set of nodes that have edges with the node v of the graph structure data. Represents the node feature of the u-th node of the t-th sequence unit of the first sequence data, is a normalizing constant, represents the graph weight parameter, represents the graph bias parameter, It is a S-shaped function.

[0013] In a preferred embodiment, the calculation formula of the second hidden layer is as follows:

[0014] ;

[0015] in Indicates The second hidden state, 1, Indicates the sequence number of the first sequence data, Indicates The second hidden state, , Represents the set of all nodes in the graph structure data, and are the first and second sequence weight parameters, is the first sequence bias parameter, and tanh is the hyperbolic tangent function.

[0016] In a preferred embodiment, the calculation formula of the third hidden layer is as follows:

[0017] ;

[0018] in Indicates A third hidden state, 1, Indicates the sequence number of the second sequence data, Indicates A third hidden state, Indicates the first Sequence data, and are the weight parameters of the third and fourth sequences, is the bias parameter of the second sequence, and tanh is the hyperbolic tangent function.

[0019] In a preferred embodiment, the calculation formula of the fourth hidden layer is as follows:

[0020] ;

[0021] represents the fourth hidden state, represents the input observed road image, Represents the convolution function.

[0022] In a preferred embodiment, the calculation formula of the first feature fusion layer is as follows:

[0023] ;

[0024] in, represents the first fusion state, represents the concatenation function, represents the sum weight matrix, Indicates The second hidden state, Indicates A third hidden state, represents the fourth hidden state, represents the summation bias parameter, It is a S-shaped function.

[0025] In a preferred embodiment, the calculation formula of the first output layer is as follows:

[0026] ;

[0027] in represents the first output vector, and the i-th component of the output vector represents the probability value of the i-th action; the set of all actions is called the action space, which contains all possible actions that may be selected. represents the first fusion state, is the first output weight parameter, is the first output bias parameter, represents the sigmoid function;

[0028] The calculation formula of the second output layer is as follows:

[0029] ;

[0030] in represents the second output vector, and the two components of the output vector represent the driving and traveling behavior of the trailer and the probability of abnormal events caused by the driving and traveling behavior of the trailer, respectively. Indicates A third hidden state, is the second output weight parameter, is the second output bias parameter, represents the sigmoid function;

[0031] The calculation formula of the third output layer is as follows:

[0032] ;

[0033] in represents the third output vector, the fth component of the vector represents the position coordinates of the fth vehicle at the next moment, represents the fourth hidden state, is the third output weight parameter, is the third output bias parameter, Represents a sigmoid function.

[0034] In a preferred embodiment, the fourth hidden layer is trained in combination with the third output layer, and the predicted vehicle position coordinates are continuously optimized through the real vehicle position coordinates, and the training model has the ability to automatically calibrate the vehicle position; the training loss function formula is as follows:

[0035] ;

[0036] in represents the loss value, represents the total number of training samples, represents the real position coordinates of the qth vehicle, Represents the predicted position coordinates of the qth vehicle output by the model;

[0037] The third hidden layer is trained in combination with the second output layer. The training model has the ability to judge abnormal events during the driving of the trailer. The loss function formula of the training is as follows:

[0038] ;

[0039] in represents the loss value, represents the probability of abnormal events predicted by the model, Indicates the actual abnormal event that occurred. If it occurred, it is 1, and if it did not occur, it is 0.

[0040] In a preferred embodiment, only the first output layer is used when training the trailer prediction model, and the training steps include:

[0041] Step 101, constructing a target network, the target network replicates the trailer prediction model;

[0042] Step 102, randomly obtain an experience from the experience pool, the experience includes the state at time e , the action performed at time e , the state at time e+1 , perform the action Rewards ;

[0043] The states at time e and e+1 represent the trailer driving information at the e-th and e+1-th time input into the trailer prediction model, respectively;

[0044] Then calculate the state The target network output corresponds to The probability value ;

[0045] ;

[0046] express The maximum probability value of the target network output;

[0047] Step 103, according to and Calculate the loss value loss and update the trailer prediction model;

[0048] ;

[0049] express The output of the trailer prediction model corresponds to The probability value of

[0050] is the discount factor, a value between 0 and 1;

[0051] Step 104, updating the target network every fixed number of network updates so that its parameters are the same as those of the current trailer prediction model;

[0052] Step 105, until the collected experience number reaches a set threshold or the trailer prediction model converges, the step is terminated.

[0053] A storage medium storing non-temporary computer-readable instructions, when the non-temporary computer-readable instructions are executed by a computer, can execute a trailer real-time traffic information collection and processing system based on artificial intelligence as described above.

[0054] The beneficial effects of the present invention are:

[0055] Through deep learning models, the vehicle driving rules can be autonomously learned from a large amount of historical data. It no longer relies on manually designed rules, has stronger generalization and adaptability, and can effectively cope with complex and changeable actual road and traffic conditions.

[0056] By integrating vehicle driving information, historical traffic data, and road images and performing feature fusion through a neural network model, it is possible to comprehensively consider various factors and make accurate predictions on future vehicle trajectories and accident probabilities, providing strong support for making optimal driving decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a module diagram of a trailer real-time traffic information collection and processing system based on artificial intelligence of the present invention. DETAILED DESCRIPTION

[0058] The subject matter described herein will now be discussed with reference to example implementations. It should be understood that the discussion of these implementations is only to enable those skilled in the art to better understand and implement the subject matter described herein, and the functions and arrangements of the elements discussed may be changed without departing from the scope of protection of the present specification. Various examples may omit, replace, or add various processes or components as needed. In addition, the features described in some examples may also be combined in other examples.

[0059] At least one embodiment of the present invention discloses a trailer real-time traffic information collection and processing system based on artificial intelligence, such as Figure 1 As shown, the following steps are included:

[0060] The information collection module 100 collects trailer driving information, which includes trailer information, other vehicle information, weather information, and road information; obtains trailer historical traffic data information; and obtains observed road image information;

[0061] Trailer information includes: trailer location, speed, and vehicle attributes. Vehicle attributes include: trailer description text consisting of vehicle type, vehicle size, and load capacity;

[0062] Other vehicle information includes: other vehicle description text consisting of the position, speed, and vehicle type of other vehicles;

[0063] Weather information includes: weather description text consisting of rainy and snowy days, foggy days, sunny days, cloudy days, overcast days, precipitation, and visibility;

[0064] Road information includes: road description text consisting of road sections, intersections, traffic signs, and traffic lights;

[0065] The trailer historical traffic data information includes: the trailer's trajectory, road, weather during historical driving, and the trailer's historical traffic description text composed of abnormal event records;

[0066] The observed road image information includes: a picture composed of all observed information of the road around the trailer;

[0067] The first encoding module 200 encodes the trailer driving information into graph structure data, wherein the graph structure data includes nodes and edges connecting the nodes, wherein a node represents a trailer, other vehicles, weather, and road, and an edge represents the relationship between the trailer, other vehicles, weather, and road;

[0068] There is an edge between the nodes corresponding to other vehicles adjacent to the trailer and the trailer node; there is an edge between the node corresponding to the trailer affected by the weather and the weather node; there is an edge between the nodes corresponding to other vehicles affected by the weather and the weather node; there is an edge between the node corresponding to the road affected by the weather and the weather node; there is an edge between the node corresponding to the trailer affected by the road and the road node; there is an edge between the nodes corresponding to other vehicles affected by the road and the road node;

[0069] The node characteristics of the node representing the trailer are equal to the trailer characteristics, the node characteristics of the nodes representing other vehicles are equal to the other vehicle characteristics, the node characteristics of the node representing the weather are equal to the weather characteristics, and the node characteristics of the node representing the road are equal to the road characteristics;

[0070] Trailer features, other vehicle features, weather features, and road features are obtained through feature engineering;

[0071] Encode the trailer driving information of each node into a number one sequence data, the number one sequence data includes n sequence units, and the t-th sequence unit represents the trailer driving information of all nodes at the t-th moment;

[0072] In one embodiment of the present invention, the collected trailer driving information, trailer historical traffic data information, and observed road image information are processed by feature engineering to obtain corresponding features;

[0073] The trailer driving information includes trailer description text, other vehicle description text, weather description text, and road description text. The WordEmbedding algorithm is used as a feature engineering method to generate trailer features, other vehicle features, weather features, and road features.

[0074] The trailer historical traffic data information includes the trailer historical traffic description text, which is used as a feature engineering method through the WordEmbedding algorithm;

[0075] The observed road image information includes a picture composed of all the observed information of the road around the trailer, and the CNN algorithm is used as a feature engineering method;

[0076] The second encoding module 300 encodes the trailer historical traffic data information into second sequence data, the second sequence data includes m sequence units, and the sth sequence unit represents the trailer historical traffic data information at the sth moment;

[0077] The trailer prediction module 400 inputs the No. 1 sequence data, the No. 2 sequence data, and the observed road image into the trailer prediction model, and the trailer prediction model includes a first hidden layer, a second hidden layer, a first feature fusion layer, a first output layer, a third hidden layer, a second output layer, a fourth hidden layer, and a third output layer, wherein the first hidden layer inputs the No. 1 sequence data, outputs the first hidden state to the second hidden layer, the second hidden layer outputs the second hidden state to the first feature fusion layer, the first feature fusion layer outputs the first fusion state to the first output layer, and outputs the trailer driving strategy; the third hidden layer inputs the No. 2 sequence data, outputs the third hidden state to the first feature fusion layer and the second output layer, and the second output layer outputs the probability of an abnormal event occurring to the trailer; the fourth hidden layer inputs the observed road image, outputs the fourth hidden state to the first feature fusion layer and the third output layer, and the third output layer outputs the positions of all vehicles at the next moment;

[0078] In one embodiment of the present invention, the calculation formula of the first hidden layer is as follows:

[0079]

[0080] in Represents the first hidden state of the vth node of the tth sequence unit of the first sequence data, is the set of nodes that have edges with the node v of the graph structure data. Represents the node feature of the u-th node of the t-th sequence unit of the first sequence data, is a normalizing constant (such as node degree), represents the graph weight parameter, represents the graph bias parameter, It is a S-shaped function.

[0081] In one embodiment of the present invention, the node feature of a node is represented by a trailer feature, other vehicle features, weather features, and road features;

[0082] In one embodiment of the present invention, the calculation formula of the second hidden layer is as follows:

[0083]

[0084] in Indicates The second hidden state, 1, Indicates the sequence number of the first sequence data, Indicates The second hidden state, , Represents the set of all nodes in the graph structure data, and are the first and second sequence weight parameters, is the first sequence bias parameter, and tanh is the hyperbolic tangent function.

[0085] In one embodiment of the present invention, the calculation formula of the third hidden layer is as follows:

[0086]

[0087] in Indicates A third hidden state, 1, Indicates the sequence number of the second sequence data, Indicates A third hidden state, Indicates the first Sequence data, and are the weight parameters of the third and fourth sequences, is the bias parameter of the second sequence, and tanh is the hyperbolic tangent function.

[0088] In one embodiment of the present invention, the calculation formula of the fourth hidden layer is as follows:

[0089]

[0090] represents the fourth hidden state, represents the input observed road image, Represents the convolution function.

[0091] In one embodiment of the present invention, the calculation formula of the first feature fusion layer is as follows:

[0092]

[0093] in, represents the first fusion state, represents the concatenation function, represents the sum weight matrix, Indicates The second hidden state, Indicates A third hidden state, represents the fourth hidden state, represents the summation bias parameter, It is a S-shaped function.

[0094] In one embodiment of the present invention, the calculation formula of the first output layer is as follows:

[0095]

[0096] in represents the first output vector, and the i-th component of the output vector represents the probability value of the i-th action; the set of all actions is called the action space, which contains all possible actions that may be selected. represents the first fusion state, is the first output weight parameter, is the first output bias parameter, Represents a sigmoid function.

[0097] In one embodiment of the present invention, an action is represented by a matrix, the first row of the matrix represents the predicted trajectory of the trailer, the second row represents the predicted safe distance between the trailer and other vehicles, and the third row represents the prediction of abnormal events during the driving of the trailer. The abnormal events represent events that affect the normal driving of the trailer, including congestion, road construction, failure, low visibility, and traffic accidents.

[0098] The actions in the action space are obtained by random combinations of all possible values ​​of trajectories, safety distances, and abnormal events;

[0099] In one embodiment of the present invention, the calculation formula of the second output layer is as follows:

[0100]

[0101] in represents the second output vector, and the two components of the output vector represent the driving and traveling behavior of the trailer and the probability of abnormal events caused by the driving and traveling behavior of the trailer, respectively. Indicates A third hidden state, is the second output weight parameter, is the second output bias parameter, Represents a sigmoid function.

[0102] In one embodiment of the present invention, the calculation formula of the third output layer is as follows:

[0103]

[0104] in represents the third output vector, the fth component of the vector represents the position coordinates of the fth vehicle at the next moment, represents the fourth hidden state, is the third output weight parameter, is the third output bias parameter, Represents a sigmoid function.

[0105] In one embodiment of the present invention, the fourth hidden layer is trained in combination with the third output layer, and the predicted vehicle position coordinates are continuously optimized through the real vehicle position coordinates, and the training model has the ability to automatically calibrate the vehicle position; the training loss function formula is as follows:

[0106]

[0107] in represents the loss value, represents the total number of training samples, represents the real position coordinates of the qth vehicle, Represents the predicted position coordinates of the qth vehicle output by the model;

[0108] In one embodiment of the present invention, the third hidden layer is trained in combination with the second output layer, and the training model has the ability to judge abnormal events occurring during the driving of the trailer; the loss function formula of the training is as follows:

[0109]

[0110] in represents the loss value, represents the probability of abnormal events predicted by the model, Indicates the actual abnormal event that occurred, 1 if it occurred, and 0 if it did not occur;

[0111] In one embodiment of the present invention, only the first output layer is used when training the trailer prediction model, and the training steps include:

[0112] Step 101, constructing a target network, the target network replicates the trailer prediction model;

[0113] Step 102, randomly obtain an experience from the experience pool, the experience includes the state at time e , the action performed at time e , the state at time e+1 , perform the action Rewards ;

[0114] The states at time e and e+1 represent the trailer driving information at the e-th and e+1-th time input into the trailer prediction model, respectively;

[0115] Then calculate the state The target network output corresponds to The probability value ;

[0116] ;

[0117] express The maximum probability value of the target network output;

[0118] Step 103, according to and Calculate the loss value loss and update the trailer prediction model;

[0119] ;

[0120] express The output of the trailer prediction model corresponds to The probability value of

[0121] is the discount factor, a value between 0 and 1;

[0122] Step 104, updating the target network every fixed number of network updates so that its parameters are the same as those of the current trailer prediction model;

[0123] Step 105, until the collected experience number reaches a set threshold or the trailer prediction model converges, the step is terminated.

[0124] In one embodiment of the present invention, the experience is derived from the record of historical action execution.

[0125] In one embodiment of the present invention, the reward The calculation formula is as follows:

[0126]

[0127] in represents the number of safe time steps in the predicted trajectory, represents the total number of time steps, The distance of the predicted trajectory, Indicates the actual distance.

[0128] A storage medium storing non-temporary computer-readable instructions, when the non-temporary computer-readable instructions are executed by a computer, can execute a trailer real-time traffic information collection and processing system based on artificial intelligence as described above.

[0129] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation mode. The above-mentioned specific implementation mode is merely illustrative and not restrictive. Under the guidance of this embodiment, ordinary technicians in this field can also make more forms of equivalent embodiments, all of which are within the protection of this embodiment.

Claims

1. A trailer real-time traffic information collection and processing system based on artificial intelligence, characterized in that: The following steps are involved: An information collection module collects trailer driving information, which includes trailer information, other vehicle information, weather information, and road information; Get trailer historical traffic data information; Acquire observed road image information; The first encoding module encodes the trailer driving information into graph structure data, which includes nodes and edges connecting the nodes; encodes the trailer driving information of each node into a first sequence data, which includes n sequence units, and the tth sequence unit represents the trailer driving information of all nodes at the tth moment; The second encoding module encodes the trailer historical traffic data information into second sequence data, the second sequence data includes m sequence units, and the sth sequence unit represents the trailer historical traffic data information at the sth moment; The trailer prediction module inputs the No. 1 sequence data, the No. 2 sequence data, and the observed road image into the trailer prediction model. The trailer prediction model includes a first hidden layer, a second hidden layer, a first feature fusion layer, a first output layer, a third hidden layer, a second output layer, a fourth hidden layer, and a third output layer. The first hidden layer inputs the No. 1 sequence data, outputs the first hidden state to the second hidden layer, the second hidden layer outputs the second hidden state to the first feature fusion layer, the first feature fusion layer outputs the first fusion state to the first output layer, and outputs the trailer driving strategy; the third hidden layer inputs the No. 2 sequence data, outputs the third hidden state to the first feature fusion layer and the second output layer, and the second output layer outputs the probability of an abnormal event occurring to the trailer; the fourth hidden layer inputs the observed road image, outputs the fourth hidden state to the first feature fusion layer and the third output layer, and the third output layer outputs the positions of all vehicles at the next moment.

2. The trailer real-time traffic information collection and processing system based on artificial intelligence according to claim 1 is characterized in that: The calculation formula for the first hidden layer is as follows: ; in Represents the first hidden state of the vth node of the tth sequence unit of the first sequence data, is the set of nodes that have edges with the node v of the graph structure data. Represents the node feature of the u-th node of the t-th sequence unit of the first sequence data, is a normalizing constant, represents the graph weight parameter, represents the graph bias parameter, It is a S-shaped function.

3. The trailer real-time traffic information collection and processing system based on artificial intelligence according to claim 2 is characterized in that: The calculation formula for the second hidden layer is as follows: ; in Indicates The second hidden state, 1, Indicates the sequence number of the first sequence data, Indicates The second hidden state, , Represents the set of all nodes in the graph structure data, and are the first and second sequence weight parameters, is the first sequence bias parameter, and tanh is the hyperbolic tangent function.

4. The trailer real-time traffic information collection and processing system based on artificial intelligence according to claim 1 is characterized in that: The calculation formula for the third hidden layer is as follows: ; in Indicates A third hidden state, 1, Indicates the sequence number of the second sequence data, Indicates A third hidden state, Indicates the first Sequence data, and are the weight parameters of the third and fourth sequences, is the bias parameter of the second sequence, and tanh is the hyperbolic tangent function.

5. The trailer real-time traffic information collection and processing system based on artificial intelligence according to claim 1 is characterized in that: The calculation formula for the fourth hidden layer is as follows: ; represents the fourth hidden state, represents the input observed road image, Represents the convolution function.

6. The trailer real-time traffic information collection and processing system based on artificial intelligence according to claim 1 is characterized in that: The calculation formula of the first feature fusion layer is as follows: ; in, represents the first fusion state, represents the concatenation function, represents the sum weight matrix, Indicates The second hidden state, Indicates A third hidden state, represents the fourth hidden state, represents the summation bias parameter, It is a S-shaped function.

7. The trailer real-time traffic information collection and processing system based on artificial intelligence according to claim 6 is characterized in that: The calculation formula of the first output layer is as follows: ; in represents the first output vector, and the i-th component of the output vector represents the probability value of the i-th action; the set of all actions is called the action space, which contains all possible actions that may be selected. represents the first fusion state, is the first output weight parameter, is the first output bias parameter, represents the sigmoid function; The calculation formula of the second output layer is as follows: ; in represents the second output vector, and the two components of the output vector represent the driving and traveling behavior of the trailer and the probability of abnormal events caused by the driving and traveling behavior of the trailer, respectively. Indicates A third hidden state, is the second output weight parameter, is the second output bias parameter, represents the sigmoid function; The calculation formula of the third output layer is as follows: ; in represents the third output vector, the fth component of the vector represents the position coordinates of the fth vehicle at the next moment, represents the fourth hidden state, is the third output weight parameter, is the third output bias parameter, Represents a sigmoid function.

8. The trailer real-time traffic information collection and processing system based on artificial intelligence according to claim 7 is characterized in that: The fourth hidden layer is trained in combination with the third output layer. The predicted vehicle position coordinates are continuously optimized through the real vehicle position coordinates. The training model has the ability to automatically calibrate the vehicle position. The loss function formula for training is as follows: ; in represents the loss value, represents the total number of training samples, represents the real position coordinates of the qth vehicle, Represents the predicted position coordinates of the qth vehicle output by the model; The third hidden layer is trained in combination with the second output layer. The training model has the ability to judge abnormal events during the driving of the trailer. The loss function formula of the training is as follows: ; in represents the loss value, represents the probability of abnormal events predicted by the model, Indicates the actual abnormal event that occurred. If it occurred, it is 1, and if it did not occur, it is 0.

9. The trailer real-time traffic information collection and processing system based on artificial intelligence according to claim 7 is characterized in that: Only the first output layer is used when training the trailer prediction model. The training steps include: Step 101, constructing a target network, the target network replicates the trailer prediction model; Step 102, randomly obtain an experience from the experience pool, the experience includes the state at time e , the action performed at time e , the state at time e+1 , perform the action Rewards ; The states at time e and e+1 represent the trailer driving information at the e-th and e+1-th time input into the trailer prediction model, respectively; Then calculate the state The target network output corresponds to The probability value ; ; express The maximum probability value of the target network output; Step 103, according to and Calculate the loss value loss and update the trailer prediction model; ; express The output of the trailer prediction model corresponds to The probability value of is the discount factor, a value between 0 and 1; Step 104, updating the target network every fixed number of network updates so that its parameters are the same as those of the current trailer prediction model; Step 105, until the collected experience number reaches a set threshold or the trailer prediction model converges, the step is terminated.

10. A storage medium storing non-transitory computer-readable instructions, characterized in that: When the non-temporary computer-readable instructions are executed by a computer, it is capable of running an artificial intelligence-based trailer real-time traffic information collection and processing system as described in any one of claims 1-9.

Citation Information

Patent Citations

  • Multi-modal vehicle trajectory prediction method based on long and short term memory network

    CN113076599A

  • AI-based behavior prediction method and system for surrounding vehicles of intelligent driving vehicle

    CN113837297A

  • Intelligent networked vehicle lane changing decision modeling method based on space-time diagram neural network

    CN114170789A

  • Vehicle track multi-modal prediction method considering road information based on LSTM-GNN

    CN116304969A

  • Attention mechanism and target point information-based surrounding vehicle track prediction method and device

    CN116923450A

Cited By

  • Dynamic image fusion method for big data and artificial intelligence virtual human panoramic live broadcast

    CN119299746A

  • Flood prevention intelligent control method and system and monitoring well lid thereof

    CN119359087A

  • Risk control assessment method and system based on multi-dimensional perception and enterprise data quantification

    CN119379016A

  • Operation and maintenance method of intelligent traffic monitoring system

    CN119441321A

  • Preventive driving auxiliary method and system based on artificial intelligence

    CN120278505A