Nothing Special   »   [go: up one dir, main page]

CN117392834B - Intelligent route planning method and system - Google Patents

Intelligent route planning method and system Download PDF

Info

Publication number
CN117392834B
CN117392834B CN202311227186.XA CN202311227186A CN117392834B CN 117392834 B CN117392834 B CN 117392834B CN 202311227186 A CN202311227186 A CN 202311227186A CN 117392834 B CN117392834 B CN 117392834B
Authority
CN
China
Prior art keywords
data
model
user
traffic
feedback
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202311227186.XA
Other languages
Chinese (zh)
Other versions
CN117392834A (en
Inventor
欧阳雅捷
陶宇
王书聿
曹子锘
黄立婷
赵阔
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jinan University
Original Assignee
Jinan University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Jinan University filed Critical Jinan University
Priority to CN202311227186.XA priority Critical patent/CN117392834B/en
Publication of CN117392834A publication Critical patent/CN117392834A/en
Application granted granted Critical
Publication of CN117392834B publication Critical patent/CN117392834B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/044—Recurrent networks, e.g. Hopfield networks
    • G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/096—Transfer learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G06Q10/047—Optimisation of routes or paths, e.g. travelling salesman problem
    • G—PHYSICS
    • G08—SIGNALLING
    • G08G—TRAFFIC CONTROL SYSTEMS
    • G08G1/00—Traffic control systems for road vehicles
    • G08G1/01—Detecting movement of traffic to be counted or controlled
    • G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
    • G08G1/0125—Traffic data processing
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • General Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Molecular Biology (AREA)
  • Economics (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Analytical Chemistry (AREA)
  • Computing Systems (AREA)
  • Chemical & Material Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Strategic Management (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Quality & Reliability (AREA)
  • Marketing (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Tourism & Hospitality (AREA)
  • Development Economics (AREA)
  • Operations Research (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Traffic Control Systems (AREA)

Abstract

The invention provides an intelligent route planning method and system, wherein the method comprises the following steps: collecting road data and preprocessing; analyzing the preprocessed data construction model; training the model: and analyzing and verifying according to the probability value output by the model, and providing a route planning suggestion and an alarm. The invention predicts abnormal conditions on the road, such as traffic accidents or road maintenance, in real time and provides intelligent route planning suggestions for users. The system combines various data sources, such as speed and position data of a user, user feedback, weather data and the like, and analyzes the data through a deep learning model to realize high-accuracy road anomaly prediction.

Description

Intelligent route planning method and system
Technical Field
The invention belongs to the technical field of artificial intelligence, and particularly relates to an intelligent route planning method and system.
Background
As the complexity of urban traffic networks increases, real-time road information becomes increasingly important. In large cities, traffic accidents and road repairs often result in severe traffic congestion. To address this problem, a method is needed that can predict anomalies on the road in real time, thereby helping the driver make decisions ahead of time, bypassing the possible congestion sites.
At present, some cities adopt a vehicle patrol mode to detect abnormal conditions on roads. However, this method is costly and inefficient, and during peak traffic or accident, patrol vehicles may be affected by congestion and not arrive in time.
In recent years, with the development of intelligent traffic systems, some existing intelligent schemes attempt to predict abnormal situations on roads by analyzing traffic data. For example, traffic flow analysis based on big data, vehicle trajectory prediction, and sensor-based road condition monitoring. However, these schemes often rely on a large number of hardware devices, such as road sensors, or require extensive data processing and analysis, and may not respond in real time.
In order to more effectively acquire real-time road information, many online map applications provide real-time traffic information. But this information is typically based on the user's speed and location data and it is difficult to accurately predict future traffic conditions. In addition, these applications rarely utilize feedback data of users, such as reports of traffic accidents or road damage.
Disclosure of Invention
The invention aims to provide an intelligent route planning method and system, which can predict abnormal conditions on roads, such as traffic accidents or road maintenance, and provide intelligent route planning suggestions for users. The system combines various data sources, such as speed and position data of a user, user feedback, weather data and the like, and analyzes the data through a deep learning model to realize high-accuracy road anomaly prediction.
In order to achieve the above object, the present invention provides an intelligent route planning method, which includes:
S1, collecting road data and preprocessing;
S2, analyzing a preprocessed data construction model;
The model comprises a speed and position analysis module, a user feedback analysis module and a data fusion module;
The speed and position analysis module is used for carrying out data fusion by transmitting real-time analysis data of the speed and the position and real-time feedback supplementary traffic data of a user in the user feedback analysis module into the data fusion module, and the data fusion module is used for obtaining a prediction result of a traffic event based on a multi-mode learning method;
The speed and position analysis module is used for constructing EventPredictor a deep learning model to analyze the space and time modes in the traffic network, and the EventPredictor deep learning model specifically comprises:
inputting speed features, location features and historical events;
spatial patterns in a traffic network are captured using a graph neural network, each node representing a road segment or intersection, and each edge representing a connection between two road segments, as follows:
Wherein, Is the hidden state of node v at layer l+1, N (c) is the neighbors of node v, W (l) and b (l) are the weights and biases of layer l, sigma is the activation function,Is the hidden state of the node u at the first layer;
to capture the temporal pattern of traffic data, an LSTM layer is added after the GNN, represented as follows:
(ht,ct)＝LSTM(ht-1,ct-1,xt)
Where h t and c t are the hidden and cell states at time t, x t is the input feature at time t, and h t-1 and c t-1 are the hidden and cell states at time t-1, respectively;
Finally, a fully connected layer is used to predict future traffic event probabilities, as follows:
P(event)＝σ(WohT+bo)
Wherein W o and b o are the weights and biases of the output layer; p (event) is the original prediction of the model, h T represents the feature vector of the last hidden state of the model;
S3, training the model:
And S4, analyzing and verifying according to the probability value output by the model, and providing a route planning suggestion and an alarm.
Further, the method further comprises the step of optimizing and expanding the accuracy of the model, and specifically comprises the following steps:
s501, optimizing model prediction accuracy through data enhancement, migration learning and model integration;
s502, expanding by establishing a multifunctional system;
S503, automatically tuning the model to obtain the best prediction performance, and simultaneously ensuring that the model is in the latest state.
Further, the sources of the data include user data, road camera data, historical traffic data, and auxiliary data;
the preprocessing includes user data processing, road camera data, historical data processing and auxiliary data processing.
Further, the speed and position analysis module further includes, by analyzing data of the speed and the position in real time:
Designing a dynamic traffic flow map: setting a time window Δt, during which position data p= { P 1,p2,…,pn } of all users are collected, wherein P i represents the position of the ith user, and the traffic flow F is represented as:
where n is the number of users passing through a certain road section within a time window Δt, and a is the area of the road section;
abnormality detection is performed on the real-time speed: for each user i, its velocity v i is calculated as:
Wherein Δp i is the distance the user moves within time Δt; setting a threshold value theta, and judging whether traffic abnormality occurs according to the threshold value theta;
Analyzing the position hot spot: defining a hotspot score S as:
Where w i is the weight of the ith user, determined by speed and location.
Further, the real-time feedback of the user in the user feedback analysis module supplements traffic data, specifically, the steps of constructing FeedbackNet a model, and the steps of constructing FeedbackNet a model specifically include:
converting the text feedback of the user into a fixed-length vector through an embedding layer and an attention mechanism, wherein the vector is used for capturing key information in the user feedback and is used as input of a subsequent step;
the combination of feedback and prediction uses a weight coefficient alpha to balance the original prediction of the model and the feedback of the user, and meanwhile, the new prediction probability P' (event) obtained after the original prediction of the model and the feedback of the user are combined is specifically expressed as follows:
P'(event)＝α×P(event)+(1-α)×FeedbackNet(user_feedback)
Wherein FeedbackNet (user_feedback) is the code of the user feedback, α is a weight coefficient between 0 and 1, the decision weight coefficient should how much the original prediction of the trust model is, and P (event) is the original prediction of the EventPredictor deep learning model;
the predictions are re-evaluated based on the combined user feedback.
Further, the data fusion module obtains a prediction result of the traffic event based on the multi-mode learning method, and the method specifically comprises the following steps:
inputting continuous values of the codes and the speed and position data fed back by the user and enabling the data to be input in a multi-mode;
processing each data modality through its particular encoder for capturing a unique pattern of that modality;
And finally, data fusion is carried out, and the method specifically comprises the following steps:
First, the interaction characteristics ij between each two modalities are calculated, which is expressed as follows:
Wherein e i and e j are the coding features of the ith and jth modes respectively, Is the outer product operation of the feature;
A primary attention weight is calculated for each modality and then a secondary attention weight is calculated for the interaction feature, expressed as follows:
αi＝Softmax(W1ei+b1)
γij＝Softmax(W2interactionij+b2)
Where α i is the primary attention weight of the ith modality, γ ij is the secondary attention weight of the ith and jth modality interactions, and W 1,W2 and b 1,b2 are parameters of the attention mechanism;
finally, combining the primary and secondary attention weights to obtain a final fusion weight fused_feature of each mode and interaction feature, wherein the final fusion weight fused_feature is expressed as follows:
fused_feature＝∑iαiei+∑i,jγijinteractionij。
further, the step S3 of training the model specifically includes:
its attention weight a (t, s) is calculated based on a spatio-temporal attention mechanism, specifically expressed as:
A(t,s)＝σ(Wt·t+Ws·s+b)
Wherein W t and W s are weight matrices of time and space, b is a bias term, σ is a sigmoid activation function, each time point t and space position s, and the attention weight determines the importance of each time point and space position when predicting traffic events;
Using an Adam optimizer and combining a learning rate attenuation strategy; specifically, when the EventPredictor deep learning model has no improvement in performance in several epochs in succession, the learning rate is multiplied by an attenuation factor γ, where 0< γ <1, expressed specifically as:
new learning rate＝old learning rate×γ
Wherein NEW LEARNING RATE denotes a new learning rate, and oldlearning rate denotes an old learning rate.
Further, the analyzing and verifying according to the probability value output by the model, and proposing a route planning and alarming specifically includes:
Judging whether traffic time exists or not according to the threshold value, and verifying the prediction result of the model in real time through a user, a camera and auxiliary data; if the model predicts that the traffic accident occurs, information is immediately input into the system, the system informs a user through mobile equipment or third software and provides an optimal route suggestion for bypassing the position, and meanwhile, an alarm is given, and alarm information is automatically sent to a traffic police department closest to the position;
And feeding back the real-time verification result to the system for further optimizing and training the model.
In a second aspect of the invention, there is provided an intelligent route planning system comprising:
the data acquisition unit is used for collecting road data and preprocessing the road data;
The model construction unit is used for analyzing the preprocessed data construction model;
The model comprises a speed and position analysis module, a user feedback analysis module and a data fusion module;
The speed and position analysis module is used for carrying out data fusion by transmitting real-time analysis data of the speed and the position and real-time feedback supplementary traffic data of a user in the user feedback analysis module into the data fusion module, and the data fusion module is used for obtaining a prediction result of a traffic event based on a multi-mode learning method;
the speed and position analysis module comprises a step of constructing EventPredictor a deep learning model to analyze the space and time modes in the traffic network, and specifically comprises the following steps:
inputting speed features, location features and historical events;
spatial patterns in a traffic network are captured using a graph neural network, each node representing a road segment or intersection, and each edge representing a connection between two road segments, as follows:
Wherein, Is the hidden state of node v at layer i, N (v) is the neighbors of node v, W (l) and b (l) are the weights and biases of layer i, σ is the activation function;
to capture the temporal pattern of traffic data, an LSTM layer is added after the GNN, represented as follows:
(ht,ct)＝LSTM(ht-1,ct-1,xt)
Where h t and c t are the hidden state and the cell state at time t, x t is the input feature at time t;
Finally, a fully connected layer is used to predict future traffic event probabilities, as follows:
P(event)＝σ(WohT+bo)
Wherein W o and b o are the weights and biases of the output layer; p (event) is the original prediction of the model;
the model training unit is used for training the model;
and the model analysis unit is used for analyzing and verifying according to the probability value output by the model, and providing a route planning suggestion and giving an alarm.
In a third aspect of the invention, there is provided an electronic device comprising: a processor, a memory for storing processor-executable instructions;
wherein the processor is configured to invoke the instructions stored in the memory to perform the method of any of claims 1 to 7.
The beneficial technical effects of the invention are at least as follows:
(1) The invention provides intelligent route planning for users by detecting the abnormal condition of the road in real time;
(2) The invention integrates various data sources, wherein the data sources comprise various accurate data, and the data sources are analyzed by a deep learning model, so that the prediction of road abnormality is improved, and more accurate route planning is provided.
(3) The present invention employs a time-space attention mechanism in the model training part, because conventional models often consider only one aspect of time or space. The invention considers dynamic change of time and space at the same time, and automatically distributes weight for each time point and space position through an attention mechanism.
(4) The self-adaptive learning strategy is adopted in the aspect of training the model, and when the performance of the model is not improved in a plurality of epochs in succession, the learning rate can be automatically adjusted, so that the model is helped to jump out of local optimum.
Drawings
The invention will be further described with reference to the accompanying drawings, in which embodiments do not constitute any limitation of the invention, and other drawings can be obtained by one of ordinary skill in the art without inventive effort from the following drawings.
FIG. 1 is a flow chart of an intelligent route planning method according to the present invention.
Fig. 2 is a schematic diagram of an intelligent route planning system according to the present invention.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are illustrative only and are not to be construed as limiting the invention.
In one or more embodiments, as shown in fig. 1, an intelligent route planning method is disclosed, the method comprising:
S1, collecting road data and preprocessing.
The road data source is user data, and speed and position data of the user are collected in real time through an API of the online map application. In addition, the application has a function built therein, allowing the user to directly report an abnormal situation on the road, such as a traffic accident or road damage.
Road camera data, namely, acquiring real-time video stream data by utilizing a fixed road camera of a city. These video streams will be used for object detection and abnormal behavior analysis.
Historical traffic data, that is, past traffic flow and accident record data are obtained from traffic management departments. These data will be used to predict future traffic conditions and identify high risk areas.
Other ancillary data-weather data, time and holiday information are collected from the third party API.
The pretreatment is divided into: user data processing, camera data processing, history data processing, and auxiliary data processing:
Specifically, the user data processing is specifically:
velocity and position data, denoising the velocity data by using a median filter. For missing position data, the KNN filling method is used for filling.
User feedback data, namely performing word vectorization processing on text data by using the BERT model, and converting the text into vector representation with fixed length.
And (3) processing camera data, namely performing target detection on the video stream by using a YOLOv model, and identifying vehicles and pedestrians in real time. For detected vehicles, a start (Simple Online AND REALTIME TRACKING) algorithm is used for motion trajectory tracking to detect abnormal behavior such as sudden stop.
Historical data processing, namely performing time series analysis on the historical traffic flow by using a Prophet model. For historical incident records, a DBSCAN clustering algorithm is used to identify high risk areas.
And auxiliary data processing, namely performing single-heat encoding processing on the weather data, and converting various weather conditions into vector representations. For time and holiday information, the representation is done using a time stamp and one-hot coding.
S2, analyzing the preprocessed data construction model.
The model comprises a speed and position analysis module, a user feedback analysis module and a data fusion module.
In order to improve the accuracy of the whole model, a speed and position analysis module is firstly constructed. Traffic flow and speed are direct reflections of traffic conditions. Traffic anomalies, such as congestion or accidents, can be quickly captured through real-time analysis of speed and location. The core purpose of this module is to provide real-time data support for subsequent user feedback in order to capture traffic conditions in real-time.
The user feedback analysis module then analyzes the user's feedback in detail. Real-time feedback of users is an important source of supplemental traffic data. By deeply analyzing the user feedback, more details of the traffic condition, such as the specific location of the accident or the cause of the congestion, can be obtained. The purpose of this module is to more accurately understand traffic conditions and provide more information for data fusion.
Finally, a data fusion module is constructed, and because a single data source may have deviation or deficiency, more comprehensive and accurate traffic condition analysis can be obtained by fusing multiple data sources. The purpose of this module is to provide a comprehensive traffic condition assessment in order to integrate all available data.
The three modules work together to form a complete traffic state analysis system, and the traffic state analysis system aims at providing the most accurate and real-time traffic information for users and helping the users to make better travel decisions.
The speed and position analysis module is used for carrying out data fusion by transmitting real-time analysis data of the speed and the position and real-time feedback supplementary traffic data of a user in the user feedback analysis module into the data fusion module, and the data fusion module is used for obtaining a prediction result of a traffic event based on a multi-mode learning method.
The speed and position analysis module comprises the following specific steps:
a. Setting a time window deltat, and collecting position data P= { P 1,p2,…,pn }, wherein P i represents the position of the ith user, and the traffic flow F can be expressed as:
Where n is the number of users passing through a certain road segment within a time window Δt, and a is the area of the road segment.
B. Real-time speed anomaly detection, for each user i, its speed v i can be calculated as:
Where Δp i is the distance the user moves within time Δt. A threshold value θ is set and when v i < θ, the user can be considered to be likely to encounter a traffic abnormality.
C. location hotspot analysis a hotspot score S may be defined as:
Where w i is the weight of the ith user, which can be determined based on its speed and location. For example, a slower user or a user located at an intersection may have a higher weight.
D. Traffic event prediction: in order to predict possible traffic events, such as congestion, accidents, or road maintenance, a EventPredictor deep learning model is constructed. The model combines time series analysis with a graph neural network to better capture spatial and temporal patterns in the traffic network.
Wherein the EventPredictor deep learning model combines time series analysis with a graph neural network to better capture spatial and temporal patterns in the traffic network. The method comprises the following steps:
1. input characteristics:
speed characteristics-average speed over several time windows.
The location feature is the location data of the user, represented as nodes on the traffic network.
Historical events-traffic events that occur in the past, such as accidents or road repairs.
2. The graphic neural network is used to capture spatial patterns in the traffic network because the traffic network is essentially a graphic structure. Each node represents a road segment or intersection and each edge represents a connection between two road segments.
Wherein,Is the hidden state of node v at layer l, N (v) is the neighbors of node v, W (l) and b (l) are the weights and biases of layer l, and σ is the activation function.
3. LSTM layer:
To capture the temporal pattern of traffic data, an LSTM layer is added after the GNN.
(ht,ct)＝LSTM(ht-1,ct-1,xt)
Where h t and c t are the hidden state and the cell state at time t, x t is the input feature at time t.
4. Output layer:
finally, a fully connected layer is used to predict future traffic event probabilities.
P(event)＝σ(WohT+bo)
Where W o and b o are the weights and biases of the output layers.
The user feedback analysis module is indispensable for real-time feedback of the user in order to ensure the accuracy of traffic event prediction. The invention builds FeedbackNet model, which not only analyzes the text feedback of the user, but also combines the feedback with the prediction result of the model to improve the accuracy of the prediction. The FeedbackNet model includes:
A. Coding of user feedback:
First, the text feedback of the user is converted into a fixed length vector through an embedding layer and attention mechanism. This vector captures key information in the user feedback and will be input for subsequent steps.
B. Combination of feedback and prediction:
the main objective is to combine the user feedback with the original predictions of the model. To this end, a weight coefficient is used to balance the original prediction of the model with the feedback of the user. The concrete representation is as follows:
P'(event)＝α×P(event)+(1-α)×FeedbackNet(user_feedback)
where P (event) is the original prediction of the model, feedbackNet (user_feedback) is the coding of the user feedback, and α is a weight coefficient between 0 and 1, which determines how much we should trust the original prediction of the model.
C. Reevaluation prediction:
the invention can re-evaluate the prediction probability of the traffic event after combining the user feedback. The present invention may consider the event to be likely to occur if the combined predicted probability exceeds a predetermined threshold.
In particular, the data fusion module is critical in traffic prediction for fusion of multi-source data. The invention constructs DataFusionNet model, adopts multi-mode learning method, effectively merges speed, position, user feedback and other related data, so as to provide more accurate traffic event prediction. The DataFusionNet model specifically comprises:
(1) Multimodal input:
speed and position data are input as continuous values.
And (5) user feedback, namely carrying out FeedbackNet model processing on the coded vector.
Other data such as weather, holidays, special events, etc.
(2) Modality specific encoder:
Each data modality is processed by its particular encoder to capture the unique pattern of that modality.
For example, speed and position data may be processed by a time series encoder, while user feedback may be processed by a text encoder.
(5) Fusion strategy:
First, the interaction characteristics between every two modalities are calculated.
Wherein e i and e j are coding features of the ith and jth modes,Is the outer product operation of the feature.
A primary attention weight is then calculated for each modality, and a secondary attention weight is then calculated for the interaction feature.
αi＝Softmax(W1ei+b1)
γij＝Softmax(W2 interactionij+b2)
Where α i is the primary attention weight of the ith modality, γ ij is the secondary attention weight of the ith and jth modality interactions, and W 1,W2 and b 1,b2 are parameters of the attention mechanism.
And finally, combining the primary and secondary attention weights to obtain the final fusion weight of each mode and interaction characteristic, wherein the final fusion weight is expressed as follows:
fused_feature＝∑iαiei+∑i,jγijinteractionij。
s3, training the model.
Firstly, using a time-space attention mechanism, for each time point t and space position s, calculating attention weights as follows:
A(t,s)＝σ(Wt·t+Ws·s+b)
Where W t and W s are weight matrices in time and space, b is the bias term, and σ is the sigmoid activation function.
This attention weight determines the importance of each point in time and spatial location in predicting traffic events.
Then adopting an adaptive learning strategy, using an Adam optimizer, and combining a learning rate attenuation strategy. In particular, when the model has not improved performance in consecutive epochs, the learning rate is multiplied by an attenuation factor γ, where 0< γ <1. The expression is as follows:
new learning rate＝old learning rate×γ
and S4, analyzing and verifying according to the probability value output by the model, and providing a route planning suggestion and an alarm. The method comprises the following steps:
s401, analyzing and verifying results:
A. And setting a threshold according to the probability value output by the model, wherein the threshold is set to be 0.8, and when the probability predicted by the model exceeds the threshold, the position is considered to be possible to have traffic event.
B. And (3) verifying in real time, namely verifying the prediction result of the model in real time through user feedback, cameras and other data sources. If multiple data sources all confirm the prediction of the model, then it may be more determined that the prediction is accurate.
S402, route planning suggestion:
A. Dynamic route planning-once the model predicts that a traffic event may occur at a location, this information is entered into the route planning system. The user is provided with an optimal route suggestion to bypass the location.
B. And the user notification is that the information about traffic events is notified to the user in real time by means of a mobile phone APP, a navigation system and the like, and new route suggestions are provided for the user.
S403, notifying a related department:
A. automatically notifying traffic police if the model predicts that a traffic accident occurs at a certain position, the system can automatically send notification to the nearest traffic police department to enable the traffic police to arrive at the site for processing as soon as possible.
B. Notifying a rescue team or a road maintenance team that the system may automatically notify the road maintenance team if road damage or other condition requiring maintenance is predicted.
C. and (3) data feedback, namely feeding back a real-time verification result into the system for further optimization and training of the model.
S5, optimizing and expanding accuracy of the model, and specifically comprising the following steps:
s501, optimizing model prediction accuracy through data enhancement, migration learning and model integration.
Specifically, to improve the generalization ability of the model, data enhancement techniques such as random noise injection, slight deformation of time series, and the like may be used to increase the diversity of training data. Only the last layers are adjusted to accommodate the current task using models pre-trained on other traffic datasets, thereby improving prediction accuracy. And an integration method of a plurality of models, such as bagging or boosting, is used for combining prediction results of the plurality of models so as to improve the overall prediction accuracy.
S502, expanding by building a multifunctional system.
In particular, in addition to existing data sources, the system may further support other data sources, such as social media traffic information, weather data, etc., to provide more comprehensive traffic predictions. The system is expanded to support different cities or regions to provide services for more users. An interactive user interface was developed that allows users to customize their needs, such as setting predicted time ranges, regions of interest, etc.
S503, automatically tuning the model to obtain the best prediction performance, and simultaneously ensuring that the model is in the latest state.
In particular, by using techniques such as Bayesian optimization, the system can automatically adjust the hyper-parameters of the model to obtain the best predicted performance. When new data is available, the system can update the model in real time, ensuring that the model is always in the latest state. By integrating the real-time feedback of the user, the system can more accurately know the prediction effect of the model and optimize the model according to the prediction effect.
In a second aspect of the embodiments of the present disclosure,
An intelligent route planning system is provided, fig. 2 is a schematic structural diagram of an intelligent route planning system according to an embodiment of the disclosure, including:
the data acquisition unit is used for collecting road data and preprocessing the road data;
The model construction unit is used for analyzing the preprocessed data construction model;
The model comprises a speed and position analysis module, a user feedback analysis module and a data fusion module;
The speed and position analysis module is used for carrying out data fusion by transmitting real-time analysis data of the speed and the position and real-time feedback supplementary traffic data of a user in the user feedback analysis module into the data fusion module, and the data fusion module is used for obtaining a prediction result of a traffic event based on a multi-mode learning method;
the speed and position analysis module comprises a step of constructing EventPredictor a deep learning model to analyze the space and time modes in the traffic network, and specifically comprises the following steps:
inputting speed features, location features and historical events;
spatial patterns in a traffic network are captured using a graph neural network, each node representing a road segment or intersection, and each edge representing a connection between two road segments, as follows:
Wherein, Is the hidden state of node v at layer i, N (v) is the neighbors of node v, W (l) and b (l) are the weights and biases of layer i, σ is the activation function;
to capture the temporal pattern of traffic data, an LSTM layer is added after the GNN, represented as follows:
(ht,ct)＝LSTM(ht-1,ct-1,xt)
Where h t and c t are the hidden state and the cell state at time t, x t is the input feature at time t;
Finally, a fully connected layer is used to predict future traffic event probabilities, as follows:
P(event)＝σ(WohT+bo)
Wherein W o and b o are the weights and biases of the output layer; p (event) is the original prediction of the model;
the model training unit is used for training the model;
and the model analysis unit is used for analyzing and verifying according to the probability value output by the model, and providing a route planning suggestion and giving an alarm.
In a third aspect of the embodiments of the present disclosure,
There is provided an electronic device including:
A processor;
A memory for storing processor-executable instructions;
wherein the processor is configured to invoke the instructions stored in the memory to perform the method described previously.
In a fourth aspect of embodiments of the present disclosure,
There is provided a computer readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the method as described above.
The present invention may be a method, apparatus, system, and/or computer program product. The computer program product may include a computer readable storage medium having computer readable program instructions embodied thereon for performing various aspects of the present invention.
In the foregoing embodiments, the descriptions of the embodiments are emphasized, and in part, not described or illustrated in any particular embodiment, reference is made to the related descriptions of other embodiments.
While embodiments of the invention have been shown and described, it will be understood by those skilled in the art that: many changes, modifications, substitutions and variations may be made to the embodiments without departing from the spirit and principles of the invention, the scope of which is defined by the claims and their equivalents.
The various illustrative logical blocks, modules, and circuits described in connection with the embodiments disclosed herein may be implemented or performed with a general purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, e.g., a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration.
The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. A software module may reside in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art. An exemplary storage medium is coupled to the processor such the processor can read information from, and write information to, the storage medium. In the alternative, the storage medium may be integral to the processor. The processor and the storage medium may reside in an ASIC. The ASIC may reside in a user terminal. In the alternative, the processor and the storage medium may reside as discrete components in a user terminal.
In one or more exemplary embodiments, the functions described may be implemented in hardware, software, firmware, or any combination thereof. If implemented in software as a computer program product, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes both computer storage media and communication media including any medium that facilitates transfer of a computer program from one place to another. A storage media may be any available media that can be accessed by a computer. By way of example, and not limitation, such computer-readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. Any connection is properly termed a computer-readable medium. For example, if the software is transmitted from a web site, server, or other remote source using a coaxial cable, fiber optic cable, twisted pair, digital Subscriber Line (DSL), or wireless technologies such as infrared, radio, and microwave, then the coaxial cable, fiber optic cable, twisted pair, DSL, or wireless technologies such as infrared, radio, and microwave are included in the definition of medium. Disk (disk) and disc (disk) as used herein include Compact Disc (CD), laser disc, optical disc, digital Versatile Disc (DVD), floppy disk and blu-ray disc where disks (disk) usually reproduce data magnetically, while discs (disk) reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media.
The previous description of the disclosure is provided to enable any person skilled in the art to make or use the disclosure. Various modifications to the disclosure will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other variations without departing from the spirit or scope of the disclosure. Thus, the disclosure is not intended to be limited to the examples and designs described herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (8)

1. An intelligent route planning method, the method comprising:
S1, collecting road data and preprocessing;
S2, analyzing a preprocessed data construction model;
The model comprises a speed and position analysis module, a user feedback analysis module and a data fusion module;
The speed and position analysis module is used for carrying out data fusion by transmitting real-time analysis data of the speed and the position and real-time feedback supplementary traffic data of a user in the user feedback analysis module into the data fusion module, and the data fusion module is used for obtaining a prediction result of a traffic event based on a multi-mode learning method;
The speed and position analysis module is used for constructing EventPredictor a deep learning model to analyze the space and time modes in the traffic network, and the EventPredictor deep learning model specifically comprises:
inputting speed features, location features and historical events;
spatial patterns in a traffic network are captured using a graph neural network, each node representing a road segment or intersection, and each edge representing a connection between two road segments, as follows:
Wherein, Is the hidden state of node v at layer l+1, N (v) is the neighbors of node v, W (l) and b (l) are the weights and biases of layer l, sigma is the activation function,Is the hidden state of the node u at the first layer;
to capture the temporal pattern of traffic data, an LSTM layer is added after the GNN, represented as follows:
(ht,ct)＝LSTM(ht-1,ct-1,xt)
Where h t and c t are the hidden and cell states at time t, x t is the input feature at time death, and h t-1 and c t-1 are the hidden and cell states at time death-1, respectively;
Finally, a fully connected layer is used to predict future traffic event probabilities, as follows:
P(event)＝σ(WohT+bo)
Wherein W o and b o are the weights and biases of the output layer; p (event) is the original prediction of the model, h T represents the feature vector of the last hidden state of the model;
The real-time feedback of the user in the user feedback analysis module supplements traffic data, and the specific steps are to construct FeedbackNet a model, and the specific steps of constructing FeedbackNet model include:
converting the text feedback of the user into a fixed-length vector through an embedding layer and an attention mechanism, wherein the vector is used for capturing key information in the user feedback and is used as input of a subsequent step;
The combination of feedback and prediction uses a weight coefficient alpha to balance the original prediction of the model and the feedback of the user, and meanwhile, the new prediction probability P' (event) obtained after the original prediction of the model and the feedback of the user are combined is specifically expressed as follows:
P(event)＝α×P(event)+(1-α)×FeedbackNet(user_feedback)
Wherein FeedbackNet (user_feedback) is the code of the user feedback, α is a weight coefficient between 0 and 1, determining how much the weight coefficient should trust the original prediction of the model, and P (event) is the original prediction of the EventPredictor deep learning model;
Reevaluating the predictions based on the combined user feedback;
The data fusion module obtains a prediction result of a traffic event based on a multi-mode learning method, and specifically comprises the following steps:
inputting continuous values of the codes and the speed and position data fed back by the user and enabling the data to be input in a multi-mode;
processing each data modality through its particular encoder for capturing a unique pattern of that modality;
And finally, data fusion is carried out, and the method specifically comprises the following steps:
First, the interaction characteristics ij between each two modalities are calculated, which is expressed as follows:
Wherein e i and e j are the coding features of the ith and jth modes respectively, Is the outer product operation of the feature;
A primary attention weight is calculated for each modality and then a secondary attention weight is calculated for the interaction feature, expressed as follows:
αi＝Softmax(W1ei+b1)
γij＝Softmax(W2interactionij+b2)
Where α i is the primary attention weight of the ith modality, γ ij is the secondary attention weight of the ith and jth modality interactions, and W 1,W2 and b 1,b2 are parameters of the attention mechanism;
finally, combining the primary and secondary attention weights to obtain a final fusion weight fused_feature of each mode and interaction feature, wherein the final fusion weight fused_feature is expressed as follows:
fused_feature＝∑iαiei+∑i,jγijinteractionij；
S3, training the model:
And S4, analyzing and verifying according to the probability value output by the model, and providing a route planning suggestion and an alarm.
2. The intelligent route planning method according to claim 1, further comprising S5, optimizing and expanding accuracy of the model, specifically comprising:
s501, optimizing model prediction accuracy through data enhancement, migration learning and model integration;
s502, expanding by establishing a multifunctional system;
S503, automatically tuning the model to obtain the best prediction performance, and simultaneously ensuring that the model is in the latest state.
3. An intelligent route planning method according to claim 1, wherein the sources of data include user data, road camera data, historical traffic data and assistance data;
the preprocessing includes user data processing, road camera data, historical data processing and auxiliary data processing.
4. An intelligent route planning method according to claim 1, wherein the speed and position analysis module further comprises, by analyzing data of speed and position in real time:
Designing a dynamic traffic flow map: setting a time window delta death, and collecting position data P= { P 1,p2,...,pn } of all users in the time window, wherein P i represents the position of the ith user, and the traffic flow F is represented as:
Wherein n is the number of users passing through a certain road section within a time window delta death, and A is the area of the road section;
Abnormality detection is performed on the real-time speed: for each user i, its velocity v i is calculated as:
Wherein Δp i is the distance the user moves within time Δdeath; setting a threshold value theta, and judging whether traffic abnormality occurs according to the threshold value theta;
analyzing the position hot spot: defining a hotspot score S as:
Where w i is the weight of the ith user, determined by speed and location.
5. The intelligent route planning method according to claim 1, wherein the training of the model in S3 specifically includes:
its attention weight a (t, s) is calculated based on a spatio-temporal attention mechanism, specifically expressed as:
A(t，s)＝σ(Wt·t+Ws·s+b)
Wherein W t and W s are weight matrices of time and space, b is a bias term, σ is a sigmoid activation function, each time point t and space position s, and the attention weight determines the importance of each time point and space position when predicting traffic events;
Using an Adam optimizer and combining a learning rate attenuation strategy; specifically, when the EventPredictor deep learning model has no improvement in performance in several epochs in succession, the learning rate is multiplied by an attenuation factor γ, where 0< γ <1, expressed specifically as:
new learning rate＝old learning rate×γ
wherein NEW LEARNING RATE represents a new learning rate, old LEARNING RATE represents an old learning rate.
6. The intelligent route planning method according to claim 4, wherein the analyzing and verifying are performed according to the probability value output by the model, and the route planning suggestion and the alarm are proposed, specifically comprising:
Judging whether traffic time exists or not according to the threshold value, and verifying the prediction result of the model in real time through a user, a camera and auxiliary data; if the model predicts that the traffic accident occurs, information is immediately input into the system, the system informs a user through mobile equipment or third software and provides an optimal route suggestion for bypassing the position, and meanwhile, an alarm is given, and alarm information is automatically sent to a traffic police department closest to the position;
And feeding back the real-time verification result to the system for further optimizing and training the model.
7. An intelligent route planning system, comprising:
the data acquisition unit is used for collecting road data and preprocessing the road data;
The model construction unit is used for analyzing the preprocessed data construction model;
The model comprises a speed and position analysis module, a user feedback analysis module and a data fusion module;
The speed and position analysis module is used for carrying out data fusion by transmitting real-time analysis data of the speed and the position and real-time feedback supplementary traffic data of a user in the user feedback analysis module into the data fusion module, and the data fusion module is used for obtaining a prediction result of a traffic event based on a multi-mode learning method;
the speed and position analysis module comprises a step of constructing EventPredictor a deep learning model to analyze the space and time modes in the traffic network, and specifically comprises the following steps:
inputting speed features, location features and historical events;
spatial patterns in a traffic network are captured using a graph neural network, each node representing a road segment or intersection, and each edge representing a connection between two road segments, as follows:
Wherein, Is the hidden state of node c at layer l, N (v) is the neighbors of node c, W (l) and b (l) are the weights and biases of layer l, σ is the activation function;
to capture the temporal pattern of traffic data, an LSTM layer is added after the GNN, represented as follows:
(ht,ct)＝LSTM(ht-1,ct-1,xt)
Where h t and c t are the hidden and cell states at time of death, x t is the input feature at time of death;
Finally, a fully connected layer is used to predict future traffic event probabilities, as follows:
P(event)＝σ(WohT+bo)
wherein W o and b o are the weights and biases of the output layer; p (event) is the original prediction of EventPredictor deep learning model;
The real-time feedback of the user in the user feedback analysis module supplements traffic data, and the specific steps are to construct FeedbackNet a model, and the specific steps of constructing FeedbackNet model include:
converting the text feedback of the user into a fixed-length vector through an embedding layer and an attention mechanism, wherein the vector is used for capturing key information in the user feedback and is used as input of a subsequent step;
The combination of feedback and prediction uses a weight coefficient alpha to balance the original prediction of the model and the feedback of the user, and meanwhile, the new prediction probability P' (event) obtained after the original prediction of the model and the feedback of the user are combined is specifically expressed as follows:
P(event)＝α×P(event)+(1-α)×FeedbackNet(user_feedback)
Wherein FeedbackNet (user_feedback) is the code of the user feedback, α is a weight coefficient between 0 and 1, determining how much the weight coefficient should trust the original prediction of the model, and P (event) is the original prediction of the EventPredictor deep learning model;
Reevaluating the predictions based on the combined user feedback;
The data fusion module obtains a prediction result of a traffic event based on a multi-mode learning method, and specifically comprises the following steps:
inputting continuous values of the codes and the speed and position data fed back by the user and enabling the data to be input in a multi-mode;
processing each data modality through its particular encoder for capturing a unique pattern of that modality;
And finally, data fusion is carried out, and the method specifically comprises the following steps:
First, the interaction characteristics ij between each two modalities are calculated, which is expressed as follows:
Wherein e i and e j are the coding features of the ith and jth modes respectively, Is the outer product operation of the feature;
A primary attention weight is calculated for each modality and then a secondary attention weight is calculated for the interaction feature, expressed as follows:
αi＝Softmax(W1ei+b1)
γij＝Softmax(W2interactionij+b2)
Where α i is the primary attention weight of the ith modality, γ ij is the secondary attention weight of the ith and h-th modality interactions, and W 1,W2 and b 1,b2 are parameters of the attention mechanism;
finally, combining the primary and secondary attention weights to obtain a final fusion weight fused_feature of each mode and interaction feature, wherein the final fusion weight fused_feature is expressed as follows:
fused_feature＝∑iaiei+∑i,jγijinteractionij；
the model training unit is used for training the model;
and the model analysis unit is used for analyzing and verifying according to the probability value output by the model, and providing a route planning suggestion and giving an alarm.
8. An electronic device, comprising: a processor, a memory for storing processor-executable instructions;
Wherein the processor is configured to invoke the instructions stored in the memory to perform the method of any of claims 1 to 6.
CN202311227186.XA 2023-09-21 2023-09-21 Intelligent route planning method and system Active CN117392834B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311227186.XA CN117392834B (en) 2023-09-21 2023-09-21 Intelligent route planning method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311227186.XA CN117392834B (en) 2023-09-21 2023-09-21 Intelligent route planning method and system

Publications (2)

Publication Number Publication Date
CN117392834A CN117392834A (en) 2024-01-12
CN117392834B true CN117392834B (en) 2024-07-02

Family

ID=89463985

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311227186.XA Active CN117392834B (en) 2023-09-21 2023-09-21 Intelligent route planning method and system

Country Status (1)

Country Link
CN (1) CN117392834B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118366311B (en) * 2024-06-17 2024-08-20 中咨泰克交通工程集团有限公司 Intelligent traffic monitoring method and device based on multi-mode data fusion and graph neural network, and electronic equipment

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103278168A (en) * 2013-04-28 2013-09-04 北京航空航天大学 Path planning method for avoiding of traffic hotspots

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103236163B (en) * 2013-04-28 2015-01-07 北京航空航天大学 Traffic jam avoiding prompting system based on collective intelligence network
CN104504065B (en) * 2014-12-19 2018-01-30 百度在线网络技术（北京）有限公司 Navigation way generation method and device
CN108053673B (en) * 2017-12-08 2020-03-31 深圳壹账通智能科技有限公司 Road condition forecasting method, storage medium and server
KR102290548B1 (en) * 2021-03-25 2021-08-13 전남대학교산학협력단 Device and method for preventing traffic accident through judging road condition based on deep learning model
CN116307033A (en) * 2022-11-30 2023-06-23 上海交通大学 Intelligent traffic system anomaly prediction method, device and storage medium

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103278168A (en) * 2013-04-28 2013-09-04 北京航空航天大学 Path planning method for avoiding of traffic hotspots

Also Published As

Publication number Publication date
CN117392834A (en) 2024-01-12

Similar Documents

Publication Publication Date Title
Ata et al. Modelling smart road traffic congestion control system using machine learning techniques
Nallaperuma et al. Online incremental machine learning platform for big data-driven smart traffic management
Shin et al. Prediction of traffic congestion based on LSTM through correction of missing temporal and spatial data
CN110415516A (en) Urban traffic flow prediction technique and medium based on figure convolutional neural networks
George et al. Traffic prediction using multifaceted techniques: A survey
CN117392834B (en) Intelligent route planning method and system
CN112462774A (en) Urban road supervision method and system based on unmanned aerial vehicle navigation following and readable storage medium
CN116385970B (en) People stream aggregation prediction model based on space-time sequence data
CN118571020B (en) Urban traffic jam prediction method and system based on deep learning
CN112382097A (en) Urban road supervision method and system based on dynamic traffic flow and readable storage medium
CN116203971A (en) Unmanned obstacle avoidance method for generating countering network collaborative prediction
Kumar et al. Moving Vehicles Detection and Tracking on Highways and Transportation System for Smart Cities
CN113682302B (en) Driving state estimation method and device, electronic equipment and storage medium
US20240112044A1 (en) Methods and systems for ontology construction with ai-mediated crowdsourcing and concept mining for high-level activity understanding
Krishna et al. A Computational Data Science Based Detection of Road Traffic Anomalies
Lakshna et al. Smart traffic: traffic congestion reduction by shortest route* search algorithm
Wei Enhancing road safety in internet of vehicles using deep learning approach for real-time accident prediction and prevention
Anitha et al. Prediction of road traffic using naive bayes algorithm
Zuo Public safety risk prediction of urban rail transit based on mathematical model and algorithm simulation
Michelaraki et al. Modelling the Safety Tolerance Zone: Recommendations from the i-DREAMS project
CN118736847B (en) Edge computing gateway data processing method and gateway
CN118269968B (en) Prediction method of automatic driving collision risk fused with online map uncertainty
CN113570846B (en) Traffic warning situation analysis and judgment method, equipment and readable storage medium
Anbukkarasi et al. AI Techniques for Future Smart Transportation
Muduli et al. Predicting Pedestrian Movement in Unsignalized Crossings: A Contextual Cue-Based Approach

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant