CN111402576A - Urban road traffic state prediction system based on deep learning - Google Patents

Urban road traffic state prediction system based on deep learning Download PDF

Info

Publication number
CN111402576A
CN111402576A CN202010024882.0A CN202010024882A CN111402576A CN 111402576 A CN111402576 A CN 111402576A CN 202010024882 A CN202010024882 A CN 202010024882A CN 111402576 A CN111402576 A CN 111402576A
Authority
CN
China
Prior art keywords
module
data
traffic
prediction
network
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.)
Pending
Application number
CN202010024882.0A
Other languages
Chinese (zh)
Inventor
郝威
易可夫
高志波
张兆磊
戎栋磊
王杰
王正武
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.)
Changsha University of Science and Technology
Original Assignee
Changsha University of Science and Technology
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 Changsha University of Science and Technology filed Critical Changsha University of Science and Technology
Priority to CN202010024882.0A priority Critical patent/CN111402576A/en
Publication of CN111402576A publication Critical patent/CN111402576A/en
Pending legal-status Critical Current

Links

Images

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
    • G08G1/0125—Traffic data processing
    • 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
    • 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
    • G06N3/045—Combinations of networks
    • 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
    • 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"
    • 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
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/40—Business processes related to the transportation industry
    • 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
    • G08G1/0129—Traffic data processing for creating historical data or processing based on historical data
    • 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
    • G08G1/0137—Measuring and analyzing of parameters relative to traffic conditions for specific applications

Landscapes

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

Abstract

A deep learning-based urban road traffic state prediction system realizes prediction by utilizing the following modules: the system comprises a traffic environment module, a memory library module, a neural network module, a training and promoting network module, a visualization module and an interaction module; the traffic environment module comprises an acquisition module and a preprocessing module, wherein the acquisition module acquires urban road position information, peak-hour average vehicle speed information, air temperature information, precipitation probability information and congestion length information; the preprocessing module is used for preprocessing traffic data based on a Lagrange interpolation method and a normalization method, obtaining reliable data for traffic prediction, storing the obtained data into a memory base module, and constructing a neural network according to the reliable data, so that a deep cycle learning network of a traffic state is constructed.

Description

Urban road traffic state prediction system based on deep learning
Technical Field
The invention belongs to the field of urban traffic system analysis and traffic condition prediction, and particularly relates to an urban road traffic state prediction system based on deep learning.
Background
With the development of urbanization, the contradiction between traffic infrastructure and automobile holding capacity is more severe, the congestion problem is more serious, and economic loss, travel time consumption and environmental pollution are inevitably caused. The treatment of traffic jam is firstly prevented, the traffic state change trend in a short time is predicted according to the existing traffic state of a road, and the possible jam phenomenon is early warned; and then, information platforms such as traffic broadcast and microblog are used for sending out early warning, leading vehicles to reasonably select running routes and strengthening order management so as to avoid congestion or relieve congestion degree. Therefore, how to establish a long-acting model to perform timely early warning on traffic jam is a research hotspot of optimizing an urban intelligent traffic system.
Many research achievements for traffic jam prediction at home and abroad mainly include methods such as prediction analysis, machine learning prediction and multi-classification combined prediction based on time series correlation, but all have defects of different degrees. For example, the non-parametric regression method is based on a large amount of historical data and constructs many hypothesis conditions, and thus is difficult to be applied to traffic flow having non-linear characteristics; the effectiveness of the model based on the multilayer perception neural network and the back propagation neural network is improved, but the training process is long in time and is easy to fall into local optimization; most of prediction methods based on machine learning lack robustness in processing big data, so that the model generally lacks long-term effectiveness and expansion capability; the learning method for recognizing the data mode by simulating the multilayer perception structure of the human brain based on deep learning accelerates the data processing speed on one hand and does not consider the dimension disaster problem caused by the high-dimensional state of the traffic flow parameters on the other hand.
The model is based on an improved time cycle Neural Network (RNN), and long-term dependence information of a time sequence can be learned by long-term memory (L on short-term memory, L STM), the depth of the long-term dependence information is reflected between input and output and is also reflected on a cross-time step, and the long-term dependence information is suitable for capturing space-time evolution rules in traffic state parameters such as traffic flow and speed.
Disclosure of Invention
The invention aims to solve the technical problem of providing an urban road traffic state prediction system based on deep learning, and provides a method for predicting future traffic states and delay indexes by using historical data, so that the problems of overlarge data volume, gradient explosion, gradient disappearance and the like are solved, the application of traffic multidimensional data is comprehensively considered, and the efficiency of traffic delay prediction is further improved.
In order to achieve the purpose, the technical scheme of the invention is as follows: a deep learning-based urban road traffic state prediction system realizes prediction by utilizing the following modules: the system comprises a traffic environment module, a memory library module, a neural network module, a training and promoting network module, a visualization module and an interaction module;
the traffic environment module comprises an acquisition module and a preprocessing module, wherein the acquisition module acquires urban road position information, peak-hour average vehicle speed information, air temperature information, precipitation probability information and congestion length information; the preprocessing module is used for preprocessing traffic data based on a Lagrange interpolation method and a normalization method, obtaining reliable data for traffic prediction, storing the obtained data into a memory base module, and constructing a neural network according to the reliable data so as to construct a deep cycle learning network of a traffic state; the specific pretreatment method comprises the following steps:
performing abnormal value interpolation by a Lagrange interpolation method, deleting partial invalid data and abnormal data so as to improve the value of the data, converting the data into a list to form a matrix, and converting tensor into 3 dimensions to be used as L STM cell for input, wherein the Lagrange interpolation function formula is as follows:
Figure BDA0002362098450000021
in the formula, yiPolynomial expression of degree i-1, xiRepresenting the parameters corresponding to the i point.
The method realizes the precise processing of data through a normalization function, is used for avoiding small data among different attributes of the data in the analysis process from being greatly influenced by big data, and therefore ensures the accuracy and reliability of a test result, and the expression of the normalization function is as follows:
X.normalize＝(X-X.mean)/X.std
in the formula, x.normaize represents normalized data, x.mean represents mean, and x.std represents standard deviation.
Selecting a training set and a test set, determining output and input variables and the number of network layers, determining an initial weight, a threshold value, a learning rate, an activation function, a training function and a training neural network model, stopping network training when feedback reaches the optimal state of a Q value table, if the feedback does not meet the optimal state of the Q value table, correcting and adjusting parameter values, adjusting the parameters to enable prediction and input test set data to obtain the optimal prediction result, and analyzing the prediction result to obtain the final prediction result;
the interaction module comprises: and selecting the first three optimal prediction results to be interacted with the visualization module, so that road travel information is provided for the user.
Further, the method further comprises: and operating the prediction result to obtain a visualized loss function curve graph and a Q-value graph.
Further, the method further comprises: the prediction result is associated with the set parameter sensitivity, and the three factors, namely learning _ rate, reward _ delay and e _ greedy, are evaluated by using a three-factor three-level orthogonal test, are marked as A/B/C, have corresponding levels of 0.01/0.03/0.05, 0.9/0.8/0.7 and 0.9/0.8/0.7, and take the degradation degree of the loss curve as an optimal judgment basis; fix A, B at A1、B1On level, matching three levels of C A1B1C1、A1B1C2、A1B1C3If A is1B1C3Optimally, then, take C3Horizontal, let A1And C3Fixing, and matching with two levels of factor B1B2C3、A1B3C3If A after the test1B2C3Optimally, fetch B2、C3Two levels, two tests A2B2C3、A3B2C3If A is3B2C3Optimal, i.e., the best horizontal combination.
The algorithm step of carrying out traffic delay index prediction analysis by using L STM algorithm and Q-L earning comprises the following steps:
1): initializing a DRQN network structure, wherein the parameter is q, initializing a target network, and the parameter q' is q;
2): initializing greedy parameters, a learning rate, rewards, attenuation coefficients, iteration times, the number of each iteration turn T, training and a neural network parameter alternation period;
3)：forepisodeinEpisodesdo；
4): initial traffic state St＝S0；
5): fort from 0 to T;
6): selecting act (output an integer in the range of (0 to 2^ n _ features-1): select a with a probability of 1-epsilont＝argmaxaProbability of Q (s, a, θ), epsilon randomly selects behavior at；
7): after the behavior is determined, all states s _ all which accord with the behavior are found in the data table, and then one state is randomly selected from the s _ all to be used as st+1(if it is not found eligible in s _ all, then the behavior is re-determined), then according to stAnd st+1To calculate the prize rt；
8): will experience(s)t,at,rt,st+1) Putting the mixture into a memory pool;
9): randomly taking out batch size data, and respectively calculating q _ eval and q _ next;
10): structure y-r1+gmaxa+1Q(st+1,a|q)→q_target；
11): reversely propagating and lifting the network q according to the q _ eval and the q _ target;
12): if the iteration number is an integral multiple of the transmitter _ cycle, updating q' ═ q;
13): current state ═ st+1
14): stopping the training of the single round when the maximum number of times of the game iteration T of the round is reached, and returning to the traffic state initialization again;
15)：end for；
16)：end for。
compared with the prior art, the invention has the following beneficial effects: the traffic state prediction method based on deep Q learning is provided for traffic researchers and traffic managers; reliable prediction data is provided for traffic planners and government departments, route adjustment and traffic management are facilitated to be performed in advance, and waste of manpower and material resources is reduced; real-time data are provided for road users and drivers, so that the route planning can be adjusted in time conveniently, and the road congestion and practice waste are reduced; the method is not only limited to traffic jam prediction, but also can be used for predicting various traffic states such as traffic volume, distribution conditions and the like, and has strong practicability.
Drawings
FIG. 1 is an overall flow of prediction execution in an example of the present invention, including an input layer, a hidden layer, and an output layer.
Fig. 2 is a flow chart of traffic state prediction based on deep learning in an example of the invention.
Fig. 3 is a graph of loss predicted by the method proposed by the present invention in an example of the present invention.
Fig. 4 is a Q-value graph generated by prediction using the method proposed by the present invention in the example of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
The method is basically characterized in that normalization and Lagrange interpolation preprocessing are performed on data according to the high-dimensional characteristics of traffic data, a deep reinforcement learning network is constructed on the basis of a Python language platform, the pressure of Q value table data storage and analysis is relieved by means of an L STM algorithm, the sensitivity of the network to traffic data processing is improved through continuous training, learning, memory and reinforcement of a training set, a future traffic delay index and a traffic state are finally predicted, the sensitivity degree of a prediction result and parameters is researched through adjustment of a learning rate, a greedy coefficient and a delay index, and an optimal result is selected.
The application example of the invention is shown in fig. 2, based on a lagrange interpolation method, null data and abnormal data are eliminated, 1650 pieces of effective data are finally screened out from all traffic data groups for training and prediction, and based on a normalization method, the degree of influence of small data with different properties by large data is reduced, so that all traffic data intervals are within [0,1], and the final prediction accuracy is improved. The formula is as follows:
Figure BDA0002362098450000031
lagrange interpolation function formula
Normalized function formula (X-x.mean)/x.std
As shown in fig. 1 and fig. 2, the embodiment of the present invention constructs a deep-loop Q learning network based on multidimensional traffic data, and implements a DRQN training method by using L STM algorithm and Q-L earning, and the implementation steps are as follows:
the table is a sample of the data of the analysis
Figure BDA0002362098450000032
Figure BDA0002362098450000041
(1) According to the data analysis, visualization and platform building requirements of the embodiment, introducing application libraries of Numpy, Pandas, Matplotlib and Tensorflow, and constructing a basic library.
(2) As shown in fig. 2, in the construction of the network, a traffic environment module, a deep reinforcement learning module, a memory base module and a neural network module are required to be used for supporting a training process of data, and a training and promoting network module, a loss function module and a test data module are all used for the whole training and predicting process.
(3) And (3) initializing a DRQN network structure, initializing a target network, and setting a parameter q' ═ q.
(4) And (5) initializing parameters. Let learning rate be 0.01, delay reward coefficient be 0.9, greedy coefficient be 0.9, batchsize be 15, initial weight and threshold interval [0,1], Sigmoid function for activation function, Adam for training function, maximum iteration number T.
(5) Initial traffic state St＝So。
(6-1) selecting behavior. Selecting with a probability of 1-epsilon, wherein the probability of epsilon randomly selects the behavior a
(6-2) after the behaviors are determined, finding all states s _ all which accord with the behaviors in a training set data table, and randomly selecting one behavior from the states s _ all as st+1(if it is not found in s _ all, the behavior is re-determined), then according to s1And st+1To calculate the prize rt
(7) Randomly fetching the batch size data, and respectively calculating q _ eval, q _ next and q _ target
(8) When the iteration number is integral multiple of the transmitter _ cycle, updating q' ═ q
(9) The round of training is stopped when the maximum number of iterations is reached.
The embodiment of the invention extracts the prediction data to execute the operation, and obtains and visualizes the prediction index.
After the steps are completed, the acquisition and visualization of the prediction result mainly comprise two parts, wherein the first part is the visualization of the loss function, and the second part is the visualization of Q-value. The loss function curve graph reflects the value degree and loss consumption of the training process, and the Q-value graph reflects the high benefit of the prediction result. As shown in fig. 3 and 4.
Embodiments of the present invention analyze the sensitivity association of the predicted results with the set parameters.
And (3) carrying out evaluation by utilizing a three-factor three-level orthogonal test, wherein the three factors are learning _ rate, reward _ delay and e _ greedy, are marked as A/B/C, the corresponding levels are 0.01/0.03/0.05, 0.9/0.8/0.7 and 0.9/0.8/0.7, and the degradation degree of the loss curve is used as an optimal solution judgment basis.
Selection of parameters and optimal results for the first trial (first group)
Figure BDA0002362098450000042
Selection of parameters for the second trial and optimal results (first group)
Figure BDA0002362098450000043
Selection of parameters and optimal results for the third trial (second group)
Figure BDA0002362098450000051
It is thus obtained that the optimal prediction result, which is the final prediction value of the present embodiment, can be obtained when the learning rate is 0.03, the delay reward coefficient is 0.9, and the greedy coefficient is 0.9.
The above description is only an embodiment of the present invention, and not intended to limit the scope of the present invention, and all equivalent structures or equivalent processes performed by the present invention and the contents of the appended drawings, or directly or indirectly applied in other related arts, are encompassed in the scope of the present invention.

Claims (4)

1. The urban road traffic state prediction system based on deep learning is characterized in that: the prediction is implemented using the following modules: the traffic environment module, the memory library module, the neural network module, the training and promoting network module, the visualization module and the interaction module;
the traffic environment module comprises an acquisition module and a preprocessing module, wherein the acquisition module acquires urban road position information, peak-hour average vehicle speed information, air temperature information, precipitation probability information and congestion length information; the preprocessing module is used for preprocessing traffic data based on a Lagrange interpolation method and a normalization method, obtaining reliable data for traffic prediction, storing the obtained data into a memory base module, and constructing a neural network according to the reliable data so as to construct a deep cycle learning network of a traffic state; the specific pretreatment method comprises the following steps:
performing abnormal value interpolation by a Lagrange interpolation method, deleting partial invalid data and abnormal data so as to improve the value of the data, converting the data into a list to form a matrix, and converting tensor into 3 dimensions to be used as L STM cell for input, wherein the Lagrange interpolation function formula is as follows:
Figure 395957DEST_PATH_IMAGE001
in the formula (I), the compound is shown in the specification,
Figure 429772DEST_PATH_IMAGE002
is shown as
Figure 672534DEST_PATH_IMAGE003
The polynomial of the degree of a is,
Figure 193514DEST_PATH_IMAGE004
express correspondence
Figure 863530DEST_PATH_IMAGE005
A parameter of the point;
the method realizes the precise processing of data through a normalization function, is used for avoiding small data among different attributes of the data in the analysis process from being influenced by big data, and therefore guarantees the accuracy and reliability of a test result, and the expression of the normalization function is as follows:
Figure 853483DEST_PATH_IMAGE006
in the formula (I), the compound is shown in the specification,
Figure 899936DEST_PATH_IMAGE007
which represents the normalized data, is the normalized data,
Figure 275423DEST_PATH_IMAGE008
the mean value is represented by the average value,
Figure 585181DEST_PATH_IMAGE009
represents the standard deviation;
selecting a training set and a test set, determining output and input variables and the number of network layers, determining an initial weight, a threshold value, a learning rate, an activation function, a training function and a training neural network model, stopping network training when feedback reaches the optimal state of a Q value table, if the feedback does not meet the optimal state of the Q value table, correcting and adjusting parameter values, adjusting the parameters to enable prediction and input test set data to obtain the optimal prediction result, and analyzing the prediction result to obtain the final prediction result;
the interaction module comprises: and selecting the first three optimal prediction results to be interacted with the visualization module, so that road travel information is provided for the user.
2. The traffic state prediction method according to claim 1; the method is characterized in that: the method further comprises the following steps: and operating the prediction result to obtain a visualized loss function curve graph and a Q-value graph.
3. The traffic state prediction method according to claim 1, characterized in that: the method further comprises the following steps: the prediction result is associated with the set parameter sensitivity, and the method is characterized in that:
evaluating by using a three-factor three-level orthogonal test, wherein the three factors are learning _ rate, reward _ delay and e _ greedy, are marked as A/B/C, the corresponding levels are 0.01/0.03/0.05, 0.9/0.8/0.7 and 0.9/0.8/0.7, and the degradation degree of the loss curve is used as an optimal solution judgment basis; fix A, B at A1、B1On level, matching three levels of C A1B1C1、A1B1C2、A1B1C3If A is1B1C3Optimally, then, take C3Horizontal, let A1And C3Fixing, and matching with two levels of factor B1B2C3、A1B3C3If A after the test1B2C3Optimally, fetch B2、C3Two levels, two tests A2B2C3、A3B2C3If A is3B2C3Optimal, i.e., the best horizontal combination.
4. The traffic state prediction method according to claim 1, characterized in that:
the algorithm step of carrying out traffic delay index prediction analysis by using L STM algorithm and Q-L earning comprises the following steps:
1): initializing a DRQN network structure, wherein the parameter is q, initializing a target network, and the parameter q' is q;
2): initializing greedy parameters, a learning rate, rewards, attenuation coefficients, iteration times, the number of each iteration turn T, training and a neural network parameter alternation period;
3)：forepisodeinEpisodesdo；
4): initial traffic state St＝S0；
5): fort from 0 to T;
6): selection behavior (output one)Integer in the range of (0 to 2^ n _ features-1) and selecting a with a probability of 1-epsilont＝argmaxaProbability of Q (s, a, θ), epsilon randomly selects behavior at；
7): after the behavior is determined, all states s _ all which accord with the behavior are found in the data table, and then one state is randomly selected from the s _ all to be used as st+1(if it is not found eligible in s _ all, then the behavior is re-determined), then according to stAnd st+1To calculate the prize rt；
8): will experience(s)t,at,rt,st+1) Putting the mixture into a memory pool;
9): randomly taking out batch size data, and respectively calculating q _ eval and q _ next;
10): structure y-r1+gmaxa+1Q(st+1,a|q)→q_target；
11): reversely propagating and lifting the network q according to the q _ eval and the q _ target;
12): if the iteration number is an integral multiple of the transmitter _ cycle, updating q' ═ q;
13): current state ═ st+1
14): stopping the training of the single round when the maximum number of times of the game iteration T of the round is reached, and returning to the traffic state initialization again;
15)：end for；
16)：end for。
CN202010024882.0A 2020-01-10 2020-01-10 Urban road traffic state prediction system based on deep learning Pending CN111402576A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010024882.0A CN111402576A (en) 2020-01-10 2020-01-10 Urban road traffic state prediction system based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010024882.0A CN111402576A (en) 2020-01-10 2020-01-10 Urban road traffic state prediction system based on deep learning

Publications (1)

Publication Number Publication Date
CN111402576A true CN111402576A (en) 2020-07-10

Family

ID=71431992

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010024882.0A Pending CN111402576A (en) 2020-01-10 2020-01-10 Urban road traffic state prediction system based on deep learning

Country Status (1)

Country Link
CN (1) CN111402576A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113506440A (en) * 2021-09-08 2021-10-15 四川国蓝中天环境科技集团有限公司 Traffic state estimation method for multi-source data fusion under Lagrange coordinate system

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101742042B1 (en) * 2016-11-15 2017-05-31 한국과학기술정보연구원 Apparatus and method for traffic flow prediction
CN107180530A (en) * 2017-05-22 2017-09-19 北京航空航天大学 A kind of road network trend prediction method based on depth space-time convolution loop network
CN109859469A (en) * 2019-02-15 2019-06-07 重庆邮电大学 A kind of vehicle flowrate prediction technique based on integrated LSTM neural network
CN110383292A (en) * 2017-04-07 2019-10-25 英特尔公司 The method and system through budget and simplified training for deep neural network
CN110472587A (en) * 2019-08-19 2019-11-19 四川大学 Vibrating motor defect identification method and device based on CNN and sound time-frequency characteristics figure
CN110533932A (en) * 2019-08-06 2019-12-03 银江股份有限公司 A kind of City-level universe traffic signals recommended method and system based on rolling optimization

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR101742042B1 (en) * 2016-11-15 2017-05-31 한국과학기술정보연구원 Apparatus and method for traffic flow prediction
CN110383292A (en) * 2017-04-07 2019-10-25 英特尔公司 The method and system through budget and simplified training for deep neural network
CN107180530A (en) * 2017-05-22 2017-09-19 北京航空航天大学 A kind of road network trend prediction method based on depth space-time convolution loop network
CN109859469A (en) * 2019-02-15 2019-06-07 重庆邮电大学 A kind of vehicle flowrate prediction technique based on integrated LSTM neural network
CN110533932A (en) * 2019-08-06 2019-12-03 银江股份有限公司 A kind of City-level universe traffic signals recommended method and system based on rolling optimization
CN110472587A (en) * 2019-08-19 2019-11-19 四川大学 Vibrating motor defect identification method and device based on CNN and sound time-frequency characteristics figure

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
张春露等: "基于TensorFlow 的LSTM 模型在太原空气质量AQI 指数预测中的应用", 《重庆理工大学学报( 自然科学)》 *
张熙德: "基于正交试验的深基坑变形及预测研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *
施赛江: "交叉口信号优化控制的深度循环Q学习方法", 《信息技术与网络安全》 *
曹国清: "基于PCA-LSTM的多变量矿山排土场滑坡预警研究", 《计算机应用》 *
王钰: "结合深度学习的短时车流量预测优化方法", 《计算机工程与应用》 *
陈韫: "基于LSTM深度网络的城市道路短时交通状态", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113506440A (en) * 2021-09-08 2021-10-15 四川国蓝中天环境科技集团有限公司 Traffic state estimation method for multi-source data fusion under Lagrange coordinate system

Similar Documents

Publication Publication Date Title
CN110110862A (en) A kind of hyperparameter optimization method based on adaptability model
CN106453293A (en) Network security situation prediction method based on improved BPNN (back propagation neural network)
CN109547431A (en) A kind of network security situation evaluating method based on CS and improved BP
CN110458181A (en) A kind of syntax dependency model, training method and analysis method based on width random forest
Huang Intersection traffic flow forecasting based on ν-GSVR with a new hybrid evolutionary algorithm
CN111860989B (en) LSTM neural network short-time traffic flow prediction method based on ant colony optimization
CN107222333A (en) A kind of network node safety situation evaluation method based on BP neural network
CN113205233B (en) Lithium battery life prediction method based on wolf algorithm and multi-core support vector regression
CN111416797A (en) Intrusion detection method for optimizing regularization extreme learning machine by improving longicorn herd algorithm
CN106570250A (en) Power big data oriented microgrid short-period load prediction method
CN113887748B (en) Online federal learning task allocation method and device, and federal learning method and system
CN115829081B (en) Urban traffic carbon emission prediction method based on support vector regression model
Zhao et al. Adaptive swarm intelligent offloading based on digital twin-assisted prediction in VEC
CN111967308A (en) Online road surface unevenness identification method and system
CN117634693A (en) Carbon emission prediction method based on energy data
CN111402576A (en) Urban road traffic state prediction system based on deep learning
Liang et al. RLTD: A Reinforcement Learning-based Truth Data Discovery scheme for decision support systems under sustainable environments
CN115481784A (en) Traffic flow prediction method based on improved combination model and application
Cheng et al. Efficient scenario analysis for optimal adaptation of bridge networks under deep uncertainties through knowledge transfer
CN118193978A (en) Automobile roadblock avoiding method based on DQN deep reinforcement learning algorithm
Hu et al. Dynamic traffic signal control using mean field multi‐agent reinforcement learning in large scale road‐networks
CN117077870B (en) Water resource digital management method based on artificial intelligence
CN111507499A (en) Construction method, test method, device and system of model for prediction
CN116861256A (en) Furnace temperature prediction method, system, equipment and medium for solid waste incineration process
CN116524316A (en) Scene graph skeleton construction method under reinforcement learning framework

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
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20200710

WD01 Invention patent application deemed withdrawn after publication