CN109272746A - A kind of MFD estimating and measuring method based on BP neural network data fusion - Google Patents
A kind of MFD estimating and measuring method based on BP neural network data fusion Download PDFInfo
- Publication number
- CN109272746A CN109272746A CN201810950491.4A CN201810950491A CN109272746A CN 109272746 A CN109272746 A CN 109272746A CN 201810950491 A CN201810950491 A CN 201810950491A CN 109272746 A CN109272746 A CN 109272746A
- Authority
- CN
- China
- Prior art keywords
- network
- traffic
- data
- mfd
- road
- 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.)
- Granted
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
- G08G1/0108—Measuring and analyzing of parameters relative to traffic conditions based on the source of data
- G08G1/0112—Measuring and analyzing of parameters relative to traffic conditions based on the source of data from the vehicle, e.g. floating car data [FCD]
-
- 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
- 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
-
- 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
Landscapes
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Traffic Control Systems (AREA)
Abstract
The present invention relates to nerual network technique method field, more particularly, to a kind of MFD estimating and measuring method based on BP neural network data fusion, LDD is estimated into method and the FCD estimation resulting traffic data of method combines considerations, under car networking 100% network car data (Network Car Data, NCD) estimation traffic parameter be inspection data, the BP neural network data fusion model of the network power magnitude of traffic flow and network power traffic density is established respectively, to can get the more accurate network power magnitude of traffic flow and network power traffic density, to more accurately estimate road network MFD.In the section for having fixed detector, BP neural network data fusion is carried out to fixed detector and Floating Car traffic data collected, obtains the section weighting magnitude of traffic flow and weighting traffic density；In the section of not fixed detector, the section weighting magnitude of traffic flow and weighting traffic density are extracted with the traffic data that Floating Car acquires.The last network power magnitude of traffic flow and weighting traffic density according to data fusion, estimates road network MFD.
Description
Technical field
The present invention relates to nerual network technique method fields, are melted more particularly, to one kind based on BP neural network data
The MFD estimating and measuring method of conjunction.
Background technique
Urban traffic blocking brings huge challenge to urban transportation.How to alleviate Urban Traffic Jam Based to have become
For the focus on research direction of numerous scholars, scholars propose various traffic control strategies, are effectively relieved to a certain extent
Urban traffic blocking, but as the continuous increase of wagon flow, traffic congestion grow in intensity, various traffic control strategies will be uncomfortable
With.In the recent period, two scholars of Daganzo and Geroliminis disclose macroscopical parent map (Macroscopic Fundamental
Diagrams, MFD) outness, they think that MFD city road network can be not only described from macroscopic aspect, and
And can monitor and predict road grid traffic operating status, it is mentioned to implement traffic control strategy to supersaturated road network from macroscopic aspect
New approaches are supplied, however the MFD for how obtaining city road network becomes a big difficulty again.Road network MFD can pass through fixed test at present
Device (such as toroidal inductor, microwave, video detector) or the traffic data of GPS Floating Car acquisition are estimated.Fixed inspection
It is real by being mounted on the fixed detector in section for surveying device data estimation method (Loop Detector Data, LDD estimate method)
When acquire traffic data, then utilize MFD correlation theory, estimate road network MFD.Floating car data estimates method (Floating Car
Data, FCD estimate method) it is to acquire road network in real time by being equipped with the vehicle of GPS car-mounted terminal (such as taxi, public transport) and float
Vehicle traffic data estimates method using the wheelpath that Edie (1963) propose, obtains the network power magnitude of traffic flow and weighting traffic
Density, to estimate road network MFD.Some scholars study two kinds of estimating and measuring methods, and such as (Courbon et al., 2011) is right
Three kinds of road network MFD estimating and measuring methods such as theoretical analysis, LDD estimation method, FCD estimation method are compared analysis, study fixed inspection
Survey device position and the harmonious influence to MFD estimation of Floating Car covering.(Nagle et al., 2013) proposes Floating Car coverage rate extremely
When few 15%, method is estimated using FCD, can get accurate road network MFD, but premise in this way is must to know
Road Floating Car coverage rate, and Floating Car is uniform in the distribution of road network.(Lu et al., 2013) utilizes practical intersection video
Detection data and taxi floating data estimate road network MFD, and find that data processing time interval will affect estimating for road network MFD
Survey result.The traffic data that (Leclercq et al., 2014) is obtained using practical road network, compares LDD estimation method and FCD estimates
Two kinds of road network MFD estimating and measuring method differences such as survey method, and inquired into the scope of application of two methods.(Du et al., 2016) is directed to
Floating Car coverage rate non-uniform actual conditions in road network, it is assumed that Floating Car ratio of some specific starting point into terminal be
It is known, Floating Car ratio of equal value needed for estimating road grid traffic flow, and road grid traffic is estimated using a small number of floating car datas
Flow and traffic density, to estimate road network MFD.But actually fixed detector can only collect the traffic number of part way
According to the section for not installing fixed detector can not then obtain traffic data, and the coverage rate of GPS Floating Car is low, traffic data amount
Deficiency, there are biggish errors by the road network MFD estimated.(Amb ü hl L et al., 2016) for most of documents all only with
It is above-mentioned one of which method estimate road network MFD, the case where seldom combining both, propose by above two method into
Row data fusion, thus the more accurate road network MFD of estimation, but its data anastomosing algorithm is obtained by a large amount of empirical experimentations
, and do not have general applicability.
Summary of the invention
It is an object of the invention to overcome the deficiencies of the prior art and provide a kind of based on BP neural network data fusion
LDD is estimated method and the FCD estimation resulting traffic data of method combines consideration, under car networking 100% by MFD estimating and measuring method
The traffic parameter that car data (Network CarData, NCD) estimates of networking is inspection data, establishes network power traffic respectively
The BP neural network data fusion model of flow and network power traffic density, to can get more accurate network power
The magnitude of traffic flow and network power traffic density, to more accurately estimate road network MFD.It is right in the section for having fixed detector
Fixed detector and Floating Car traffic data collected carry out BP neural network data fusion, obtain section weighting traffic flow
Amount and weighting traffic density；In the section of not fixed detector, section weighting is extracted with the traffic data that Floating Car acquires and is handed over
Through-current capacity and weighting traffic density.The last network power magnitude of traffic flow and weighting traffic density according to data fusion, estimation
Road network MFD.
In order to solve the above technical problems, the technical solution adopted by the present invention is that:
A kind of MFD estimating and measuring method based on BP neural network data fusion is provided, the specific steps are as follows:
(1) it is imitative to establish traffic using Vissim traffic simulation software according to above-mentioned basic data for test block selected first
5% vehicle is set as Floating Car by true mode, and detector is arranged in each section middle position, 100% vehicle is set as joining
Net vehicle (special Floating Car) constructs car networking Environmental Communication model, for obtaining the verification data of neural network model；For
Simulation road network is from ebb-peak-congestion whole process, and in the road network simulation model, simulation traffic flow is opened from ebb
Begin, each section in road network boundary drives into the volume of traffic and increases 100pcu/h every 900s, until emulating altogether to the hypersaturated state on peak
30000s acquires 1 data every 300s, acquires 100 data altogether；
(2) after step (1), setting Floating Car and networking vehicle every 5 seconds upload vehicle numbers, running time, traveling away from
From etc. data, estimate method according to FCD, calculate every 300 seconds Floating Car network power magnitude of traffic flow qFCD, Floating Car road network adds
Weigh traffic density kFCD, networking vehicle network power magnitude of traffic flow qNCD, networking vehicle network power traffic density kNCD；Equally every
Road section traffic volume density, the road section traffic volume flow of each section detector are acquired within 300 seconds, method is estimated according to LDD, calculates every 300 seconds
Network power magnitude of traffic flow qLDDAnd network power traffic density kLDD。
(3) after step (2), by the q in preceding 50 periodsFCD、qLDD、nFCDMould is merged as the network power magnitude of traffic flow
The training sample of type, the q in preceding 50 periodsNCDAs test samples, network power magnitude of traffic flow Fusion Model is trained；
By the k in preceding 50 periodsFCD、kLDD、nFCDAs the training sample of network power traffic density Fusion Model, preceding 50 periods
kNCDAs test samples, network power traffic density Fusion Model is trained；Finally with trained neural network mould
Type carries out data fusion to the data in 100 periods；
(4) after step (3), the road network MFD based on FCD estimation is generated respectivelyF, the road network MFD based on LDD estimationL,
MFD based on the estimation of BP neural network data fusionBP, the road network standard MFD based on networking wheel pathsN, the road network of comparative determination
The average relative error of MFD and road network standard MFD.
The present invention provides a kind of MFD estimating and measuring method based on BP neural network data fusion, and LDD is estimated method and FCD estimates
The resulting traffic data of survey method combines consideration, under car networking 100% networking car data (Network Car Data,
NCD) traffic parameter estimated is inspection data, establishes the BP of the network power magnitude of traffic flow and network power traffic density respectively
Neural Network Data Fusion model, to can get the more accurate network power magnitude of traffic flow and network power traffic density,
To more accurately estimate road network MFD.It is collected to fixed detector and Floating Car in the section for having fixed detector
Traffic data carries out BP neural network data fusion, obtains the section weighting magnitude of traffic flow and weighting traffic density；Do not fixing
The section weighting magnitude of traffic flow and weighting traffic density are extracted with the traffic data that Floating Car acquires in the section of detector.Finally according to
According to the network power magnitude of traffic flow and weighting traffic density of data fusion, road network MFD is estimated.
Preferably, in step (2), LDD estimates method, and specific step is as follows:
(1) each section is mounted on fixed detector (such as ring coil detector, video acquisition inspection first in road network
Survey device etc.), then the road section traffic volume flow and traffic density that can be directly acquired by fixed detector estimate road network MFD,
(2) after step (1), the MFD correlation reason that is proposed according to (Geroliminis and Daganzo, 2008)
By[1~6], it is known that:
In formula: N --- road network move vehicle number (veh)；
qw、kw、ow--- the network power magnitude of traffic flow (veh/h), network power traffic density (veh/km), network power
Time occupancy；
i、li--- the length (km) of section i and the section；
qi、ki、oi--- flow (veh/h), density (vehk/km) and the time occupancy of section i；
The average vehicle commander of s --- vehicle.
Preferably, in step (2), FCD estimates method, and specific step is as follows:
When known to the track of all vehicles of road network, the magnitude of traffic flow and traffic density of road network can be calculated according to wheel paths,
Formula is as follows:
In formula: k --- road grid traffic density, veh/km；
Q --- road grid traffic flow, veh/h；
The vehicle number recorded in m --- collection period T；
N --- section sum in road network
tj--- the running time of jth vehicle, s in collection period T；
li--- the length in the i-th section, m；
T --- collection period, s；
Dj --- the operating range of jth vehicle, m in collection period T；
Tm--- the sum of the running time of all vehicles of road network, s in collection period T.
Dm--- the sum of the operating range of all vehicles of road network, s in collection period T.
If being actually difficult to obtain the driving status (operating range and running time) of all vehicles, fetching portion is floated
The driving status of vehicle；Nagle (2014) puts forward a hypothesis known to ratio ρ of the Floating Car in road network, and equal in each region of road network
Even distribution, then can estimate the magnitude of traffic flow and traffic density of road network, formula is as follows according to above-mentioned formula:
In formula:--- the road grid traffic density estimated using floating car data, veh/km；
--- the road grid traffic flow estimated using floating car data, veh/h；
The Floating Car number recorded in m ' --- collection period T；
N --- section sum in road network
tj′--- the running time of jth in collection period T ' Floating Car, s；
li--- the length in the i-th section, m；
T --- collection period, s；
dj′--- the operating range of jth in collection period T ' vehicle, m；.
Preferably, the BP neural network is made of input layer, hidden layer, three layers of output layer, and learning algorithm is complete
Office's approach method；Using BP neural network model, method is estimated to LDD and FCD estimation method calculates resulting network power traffic flow
Amount and network power traffic density carry out data fusion, obtain the network power magnitude of traffic flow and network power traffic density, thus
Estimate road network MFD.
Preferably, two key parameters of road network MFD estimation are that the network power magnitude of traffic flow and network power traffic are close
Degree will design network power magnitude of traffic flow Fusion Model and the fusion of network power traffic density based on BP neural network respectively
Model；Specific step is as follows:
(1) firstly, input data and output data；
For network power magnitude of traffic flow Fusion Model, input data mainly includes the LDD estimation resulting network power of method
The magnitude of traffic flow qLDD, FCD estimate the resulting network power magnitude of traffic flow qFCD of method, road network Floating Car sample size nFCD, i.e., refreshing
There are 3 parameters through network input layer, output layer is the fused network power magnitude of traffic flow
For network power traffic density Fusion Model, input data mainly includes the LDD estimation resulting network power of method
Traffic density kLDD, FCD estimate the resulting network power traffic density kFCD of method, road network Floating Car sample size nFCD, i.e., refreshing
There are 3 parameters through network input layer, output layer is fused network power traffic density
(2) it after step (1), carries out the network number of plies and determines；
The number of plies of BP neural network include at least 3 layers (input layer, hidden layer, output layer), hidden layer can with 1 layer or more,
But the implicit number of plies is more, and network is more complicated, and the training time is also longer.Under normal conditions, the three-layer network comprising 1 hidden layer
Most of application requirements can be met, the network number of plies for setting two models is 3 layers；
(3) after step (2), the determination of neuronal quantity；
According to following empirical formula calculate can hidden layer neuronal quantity:
In formula, NBP--- hidden layer neuron quantity；
nBP--- input layer quantity；
mBP--- output layer neuron quantity；
C --- empirical, value 0-10.
Compared with prior art, the beneficial effects of the present invention are:
The present invention provides a kind of MFD estimating and measuring method based on BP neural network data fusion, and LDD is estimated method and FCD estimates
The resulting traffic data of survey method combines consideration, under car networking 100% networking car data (Network Car Data,
NCD) traffic parameter estimated is inspection data, establishes the BP of the network power magnitude of traffic flow and network power traffic density respectively
Neural Network Data Fusion model, to can get the more accurate network power magnitude of traffic flow and network power traffic density,
To more accurately estimate road network MFD.It is collected to fixed detector and Floating Car in the section for having fixed detector
Traffic data carries out BP neural network data fusion, obtains the section weighting magnitude of traffic flow and weighting traffic density；Do not fixing
The section weighting magnitude of traffic flow and weighting traffic density are extracted with the traffic data that Floating Car acquires in the section of detector.Finally according to
According to the network power magnitude of traffic flow and weighting traffic density of data fusion, road network MFD is estimated.
Detailed description of the invention
Fig. 1 is the schematic diagram of the network structure of embodiment network power magnitude of traffic flow Fusion Model.
Fig. 2 is the schematic diagram of the network structure of network power traffic density Fusion Model.
Fig. 3 is the schematic diagram of Tianhe District core road network layout.
Fig. 4 is the network power magnitude of traffic flow comparison diagram of various estimating and measuring methods.
Fig. 5 is the network power traffic density comparison diagram of various estimating and measuring methods.
Fig. 6 is the tabular drawing of the average value of the phase error absolute value relative to qNCD.
Fig. 7 is the tabular drawing of the average value of the phase error absolute value relative to kBP.
Fig. 8 is the schematic diagram of the road network two dimension MFD of various estimating and measuring methods.
Fig. 9 is the schematic diagram of the gained road network three-dimensional MFD of various estimating and measuring methods.
Figure 10 is the schematic diagram of the fitting function of each MFD.
Specific embodiment
The present invention is further illustrated With reference to embodiment.Wherein, being given for example only property of attached drawing is said
Bright, expression is only schematic diagram, rather than pictorial diagram, should not be understood as the limitation to this patent；In order to better illustrate the present invention
Embodiment, the certain components of attached drawing have omission, zoom in or out, and do not represent the size of actual product；To art technology
For personnel, the omitting of some known structures and their instructions in the attached drawings are understandable.
The same or similar label correspond to the same or similar components in the attached drawing of the embodiment of the present invention；Of the invention
In description, it is to be understood that if having the instructions such as term " on ", "lower", "left", "right" orientation or positional relationship be based on
Orientation or positional relationship shown in the drawings, is merely for convenience of description of the present invention and simplification of the description, rather than indication or suggestion institute
The device or element of finger must have a particular orientation, be constructed and operated in a specific orientation, therefore position pass is described in attached drawing
The term of system only for illustration, should not be understood as the limitation to this patent, for those of ordinary skill in the art and
Speech, can understand the concrete meaning of above-mentioned term as the case may be.
Embodiment
As Fig. 1 to 10 show a kind of implementation of the MFD estimating and measuring method based on BP neural network data fusion of the present invention
Example, the specific steps are as follows:
(1) it is imitative to establish traffic using Vissim traffic simulation software according to above-mentioned basic data for test block selected first
5% vehicle is set as Floating Car by true mode, and detector is arranged in each section middle position, 100% vehicle is set as joining
Net vehicle (special Floating Car) constructs car networking Environmental Communication model, for obtaining the verification data of neural network model；For
Simulation road network is from ebb-peak-congestion whole process, and in the road network simulation model, simulation traffic flow is opened from ebb
To begin, each section in road network boundary drives into the volume of traffic and increases 100pcu/h every 900s, until to the hypersaturated state on peak, it is imitative altogether
True 30000s acquires 1 data every 300s, acquires 100 data altogether；
(2) after step (1), setting Floating Car and networking vehicle every 5 seconds upload vehicle numbers, running time, traveling away from
From etc. data, estimate method according to FCD, calculate every 300 seconds Floating Car network power magnitude of traffic flow qFCD, Floating Car road network adds
Weigh traffic density kFCD, networking vehicle network power magnitude of traffic flow qNCD, networking vehicle network power traffic density kNCD；Equally every
Road section traffic volume density, the road section traffic volume flow of each section detector are acquired within 300 seconds, method is estimated according to LDD, calculates every 300 seconds
Network power magnitude of traffic flow qLDDAnd network power traffic density kLDD。
(3) after step (2), by the q in preceding 50 periodsFCD、qLDD、nFCDMould is merged as the network power magnitude of traffic flow
The training sample of type, the q in preceding 50 periodsNCDAs test samples, network power magnitude of traffic flow Fusion Model is trained；
By the k in preceding 50 periodsFCD、kLDD、nFCDAs the training sample of network power traffic density Fusion Model, preceding 50 periods
kNCDAs test samples, network power traffic density Fusion Model is trained；Finally with trained neural network mould
Type carries out data fusion to the data in 100 periods；
(4) after step (3), the road network MFD based on FCD estimation is generated respectivelyF, the road network MFD based on LDD estimationL,
MFD based on the estimation of BP neural network data fusionBP, the road network standard MFD based on networking wheel pathsN, the road network of comparative determination
The average relative error of MFD and road network standard MFD.
Wherein, in step (2), LDD estimates method, and specific step is as follows:
(1) each section is mounted on fixed detector (such as ring coil detector, video acquisition inspection first in road network
Survey device etc.), then the road section traffic volume flow and traffic density that can be directly acquired by fixed detector estimate road network MFD,
(2) after step (1), the MFD correlation reason that is proposed according to (Geroliminis and Daganzo, 2008)
By[1~6], it is known that:
In formula: N --- road network move vehicle number (veh)；
qw、kw、ow--- the network power magnitude of traffic flow (veh/h), network power traffic density (veh/km), network power
Time occupancy；
i、li--- the length (km) of section i and the section；
qi、ki、oi--- flow (veh/h), density (vehk/km) and the time occupancy of section i；
The average vehicle commander of s --- vehicle.
In addition, FCD estimates method, and specific step is as follows in step (2):
When known to the track of all vehicles of road network, the magnitude of traffic flow and traffic density of road network can be calculated according to wheel paths,
Formula is as follows:
In formula: k --- road grid traffic density, veh/km；
Q --- road grid traffic flow, veh/h；
The vehicle number recorded in m --- collection period T；
N --- section sum in road network
tj--- the running time of jth vehicle, s in collection period T；
li--- the length in the i-th section, m；
T --- collection period, s；
Dj --- the operating range of jth vehicle, m in collection period T；
Tm--- the sum of the running time of all vehicles of road network, s in collection period T.
Dm--- the sum of the operating range of all vehicles of road network, s in collection period T.
If being actually difficult to obtain the driving status (operating range and running time) of all vehicles, fetching portion is floated
The driving status of vehicle；Nagle (2014) puts forward a hypothesis known to ratio ρ of the Floating Car in road network, and equal in each region of road network
Even distribution, then can estimate the magnitude of traffic flow and traffic density of road network, formula is as follows according to above-mentioned formula:
In formula:--- the road grid traffic density estimated using floating car data, veh/km；
--- the road grid traffic flow estimated using floating car data, veh/h；
The Floating Car number recorded in m' --- collection period T；
N --- section sum in road network
tj'--- the running time of jth in collection period T ' Floating Car, s；
li--- the length in the i-th section, m；
T --- collection period, s；
dj'--- the operating range of jth in collection period T ' vehicle, m；.
Wherein, the BP neural network is made of input layer, hidden layer, three layers of output layer, and learning algorithm is the overall situation
Approach method；Using BP neural network model, method is estimated to LDD and FCD estimation method calculates the resulting network power magnitude of traffic flow
Data fusion is carried out with network power traffic density, the network power magnitude of traffic flow and network power traffic density are obtained, to estimate
Survey road network MFD.
Wherein, road network MFD estimation two key parameters be the network power magnitude of traffic flow and network power traffic density,
Network power magnitude of traffic flow Fusion Model and network power traffic density fusion mould based on BP neural network are designed respectively
Type；Specific step is as follows:
(1) firstly, input data and output data；
For network power magnitude of traffic flow Fusion Model, input data mainly includes the LDD estimation resulting network power of method
The magnitude of traffic flow qLDD, FCD estimate the resulting network power magnitude of traffic flow qFCD of method, road network Floating Car sample size nFCD, i.e., refreshing
There are 3 parameters through network input layer, output layer is the fused network power magnitude of traffic flow
For network power traffic density Fusion Model, input data mainly includes the LDD estimation resulting network power of method
Traffic density kLDD, FCD estimate the resulting network power traffic density kFCD of method, road network Floating Car sample size nFCD, i.e., refreshing
There are 3 parameters through network input layer, output layer is fused network power traffic density
(2) it after step (1), carries out the network number of plies and determines；
The number of plies of BP neural network include at least 3 layers (input layer, hidden layer, output layer), hidden layer can with 1 layer or more,
But the implicit number of plies is more, and network is more complicated, and the training time is also longer.Under normal conditions, the three-layer network comprising 1 hidden layer
Most of application requirements can be met, the network number of plies for setting two models is 3 layers；
(3) after step (2), the determination of neuronal quantity；
According to following empirical formula calculate can hidden layer neuronal quantity:
In formula, NBP--- hidden layer neuron quantity；
nBP--- input layer quantity；
mBP--- output layer neuron quantity；
C --- empirical, value 0-10.
The n of above-mentioned two modelBP=3, mBP=1, C take 7, therefore the hidden layer neuron quantity of above-mentioned two model is equal
It is 9.The BP neural network structure chart of above-mentioned two model is as shown in Figure 1, Figure 2.
The present embodiment chooses reported in Tianhe district of Guangzhou core road network intersection group as research object, the road network by Milky Way road,
The trunk roads such as Milky Way East Road, Milky Way North Road, sport West Road, sport East Road and partial branch composition, including more than 7 level-crossings
Mouthful, more than 20 entrances, as shown in Figure 3.Traffic flow data is with SCATS traffic signal control system on August 6th, 2017
Based on peak hour (18:00-19:00) institute detection data, as shown in Figure 3.
To the q in 100 periodsBP、qFCD、qLDD、qNCDData comparison is carried out, as shown in figure 4, to the k in 100 periodsBP、
kFCD、kLDD、kNCDData comparison is carried out, as shown in Figure 5:
From Fig. 4-5 it is found that the resulting network power magnitude of traffic flow of FCD estimation method and network power traffic density change width
Spend it is larger, this is because Floating Car negligible amounts；LDD estimate method and car networking estimate the resulting network power magnitude of traffic flow and
The variation of network power traffic density is relatively stable, and variation tendency is more consistent, with the passage of simulation time, network power
The magnitude of traffic flow and network power traffic density are gradually increased, and then maintain more stable value whithin a period of time, next and anxious
Play decline, but the resulting network power magnitude of traffic flow of LDD estimation method and network power traffic density are respectively less than car networking estimation
The network power magnitude of traffic flow and network power traffic density, this is because small part vehicle is still when reaching data acquisition intervals
Fixed detector is not reached.
Analysis emulation data, obtain qLDD、qFCD、qBPWith qNCDAbsolute relative error average value and kLDD、kFCD、
kBPWith kNCDAbsolute relative error average value, as shown in Figure 6,7.
From shown in Fig. 6,7 it is found that qFCDAnd kFCDPhase error absolute value average value it is maximum, respectively 11.52% He
12.26%；qLDDAnd kLDDThe average value of phase error absolute value take second place, respectively 8.22% and 11.54%；Through BP nerve
After network data fusion, qBPAnd kBPThe average value of phase error absolute value be respectively 6.2% and 7.2%, closest to standard
Value qNCDAnd kNCD。
Using various estimation data, the road network MFD based on FCD estimation method is generatedF, the road network MFD based on LDD estimation methodL,
MFD based on the estimation of BP neural network data fusionBP, the road network standard MFD based on networking wheel pathsN, as shown in Figure 8.
As it can be observed in the picture that MFDFScatterplot biggish discreteness, MFD is presentedL、MFDBP、MFDNScatterplot more concentrate, and
And as the passage of simulation time, the network power magnitude of traffic flow and network power traffic density are gradually increased, network power is handed over
For flux density since 70veh/km, road network maintains the higher weighting magnitude of traffic flow, as network power traffic density increases,
The network power magnitude of traffic flow sharply declines, and hypersaturated state occurs in road network.Road network MFD has also appeared " hysteresis phenomenon " simultaneously, accords with
The characteristic of combining net MFD.Data fitting is carried out to the scatterplot of each MFD, fitting function is obtained, calculates the best of each fitting formula
Weight traffic density k0With maximum weighted magnitude of traffic flow qmax, and solve the k of each MFD0And qmaxWith the k of standard value0(NCD)With
qmax(NCD)Phase error, as shown in Figure 9.
As can be seen from Figure 9, MFDFK0And qmaxRelative error is maximum, and respectively 18.43% and 5.32%. MFDBPWith MFDL
K0And qmaxRelative error is smaller, but MFDBPK0And qmaxCloser to standard MFDNK0(NCD)And qmax(NCD), respectively
0.86% and -4.05%.It can be seen that the road network MFD after BP neural network data fusion is more accurate.
Obviously, the above embodiment of the present invention be only to clearly illustrate example of the present invention, and not be pair
The restriction of embodiments of the present invention.For those of ordinary skill in the art, may be used also on the basis of the above description
To make other variations or changes in different ways.There is no necessity and possibility to exhaust all the enbodiments.It is all this
Made any modifications, equivalent replacements, and improvements etc., should be included in the claims in the present invention within the spirit and principle of invention
Protection scope within.
Claims (5)
1. a kind of MFD estimating and measuring method based on BP neural network data fusion, which is characterized in that specific step is as follows:
(1) traffic simulation mould is established using Vissim traffic simulation software according to above-mentioned basic data in test block selected first
5% vehicle is set as Floating Car by type, and detector is arranged in each section middle position, 100% vehicle is set as vehicle of networking
(special Floating Car) constructs car networking Environmental Communication model, for obtaining the verification data of neural network model；In order to simulate
Road network is from ebb-peak-congestion whole process, in the road network simulation model, simulates traffic flow since ebb, road network side
Boundary drives into each section the volume of traffic and increases 100pcu/h every 900s, until emulating 30000s altogether, often to the hypersaturated state on peak
1 data is acquired every 300s, acquires 100 data altogether；
(2) after step (1), setting Floating Car and networking vehicle are every 5 seconds uploads vehicle number, running time, operating ranges etc.
Data estimate method according to FCD, calculate the Floating Car network power magnitude of traffic flow q every 300 secondsFCD, Floating Car network power hand over
Flux density kFCD, networking vehicle network power magnitude of traffic flow qNCD, networking vehicle network power traffic density kNCD；Equally adopted every 300 seconds
Collect road section traffic volume density, the road section traffic volume flow of each section detector, estimate method according to LDD, the road network calculated every 300 seconds adds
Weigh magnitude of traffic flow qLDDAnd network power traffic density kLDD。
(3) after step (2), by the q in preceding 50 periodsFCD、qLDD、nFCDAs network power magnitude of traffic flow Fusion Model
Training sample, the q in preceding 50 periodsNCDAs test samples, network power magnitude of traffic flow Fusion Model is trained；Will before
The k in 50 periodsFCD、kLDD、nFCDAs the training sample of network power traffic density Fusion Model, the k in preceding 50 periodsNCDMake
For test samples, network power traffic density Fusion Model is trained；Finally with trained neural network model to 100
The data in a period carry out data fusion；
(4) after step (3), the road network MFD based on FCD estimation is generated respectivelyF, the road network MFD based on LDD estimationL, it is based on
The MFD of BP neural network data fusion estimationBP, the road network standard MFD based on networking wheel pathsN, the road network MFD of comparative determination with
The average relative error of road network standard MFD.
2. the MFD estimating and measuring method according to claim 1 based on BP neural network data fusion, which is characterized in that in step
Suddenly in (2), LDD estimates method, and specific step is as follows:
(1) each section is mounted on fixed detector (such as ring coil detector, video acquisition detector first in road network
Deng), then the road section traffic volume flow and traffic density that can be directly acquired by fixed detector estimate road network MFD,
(2) after step (1), the MFD correlation reason that is proposed according to (Geroliminis and Daganzo, 2008)
By[1~6], it is known that:
In formula: N --- road network move vehicle number (veh)；
qw、kw、ow--- the network power magnitude of traffic flow (veh/h), network power traffic density (veh/km), network power time
Occupation rate；
i、li--- the length (km) of section i and the section；
qi、ki、oi--- flow (veh/h), density (vehk/km) and the time occupancy of section i；
The average vehicle commander of s --- vehicle.
3. the MFD estimating and measuring method according to claim 2 based on BP neural network data fusion, which is characterized in that in step
Suddenly in (2), FCD estimates method, and specific step is as follows:
When known to the track of all vehicles of road network, the magnitude of traffic flow and traffic density of road network, formula can be calculated according to wheel paths
It is as follows:
In formula: k --- road grid traffic density, veh/km；
Q --- road grid traffic flow, veh/h；
The vehicle number recorded in m --- collection period T；
N --- section sum in road network；
tj--- the running time of jth vehicle, s in collection period T；
li--- the length in the i-th section, m；
T --- collection period, s；
Dj --- the operating range of jth vehicle, m in collection period T；
Tm--- the sum of the running time of all vehicles of road network, s in collection period T.
Dm--- the sum of the operating range of all vehicles of road network, s in collection period T.
If being actually difficult to obtain the driving status (operating range and running time) of all vehicles, with regard to the row of fetching portion Floating Car
Sail state；Nagle (2014) puts forward a hypothesis known to ratio ρ of the Floating Car in road network, and is uniformly distributed in each region of road network,
So according to above-mentioned formula, the magnitude of traffic flow and traffic density of road network can be estimated, formula is as follows:
In formula:--- the road grid traffic density estimated using floating car data, veh/km；
--- the road grid traffic flow estimated using floating car data, veh/h；
The Floating Car number recorded in m' --- collection period T；
N --- section sum in road network
tj'--- the running time of jth in collection period T ' Floating Car, s；
li--- the length in the i-th section, m；
T --- collection period, s；
dj'--- the operating range of jth in collection period T ' vehicle, m；.
4. the MFD estimating and measuring method according to claim 1 based on BP neural network data fusion, which is characterized in that described
BP neural network is made of input layer, hidden layer, three layers of output layer, and learning algorithm is global approach method；Utilize BP mind
Through network model, method is estimated to LDD and FCD estimation method calculates the resulting network power magnitude of traffic flow and network power traffic density
Data fusion is carried out, the network power magnitude of traffic flow and network power traffic density are obtained, to estimate road network MFD.
5. the MFD estimating and measuring method according to claim 1 based on BP neural network data fusion, which is characterized in that road network
Two key parameters of MFD estimation are the network power magnitude of traffic flow and network power traffic density, are designed respectively based on BP mind
Network power magnitude of traffic flow Fusion Model and network power traffic density Fusion Model through network；Specific step is as follows:
(1) firstly, input data and output data；
For network power magnitude of traffic flow Fusion Model, input data mainly includes the LDD estimation resulting network power traffic of method
Flow qLDD, FCD estimate the resulting network power magnitude of traffic flow qFCD of method, road network Floating Car sample size nFCD, i.e. nerve net
Network input layer has 3 parameters, and output layer is the fused network power magnitude of traffic flow
For network power traffic density Fusion Model, input data mainly includes the LDD estimation resulting network power traffic of method
Density kLDD, FCD estimate the resulting network power traffic density kFCD of method, road network Floating Car sample size nFCD, i.e. nerve net
Network input layer has 3 parameters, and output layer is fused network power traffic density
(2) it after step (1), carries out the network number of plies and determines；
The number of plies of BP neural network includes at least 3 layers (input layer, hidden layer, output layer), and hidden layer can be but hidden with 1 layer or more
More containing the number of plies, network is more complicated, and the training time is also longer.Under normal conditions, the three-layer network comprising 1 hidden layer can meet
Most of application requirements, the network number of plies for setting two models is 3 layers；
(3) after step (2), the determination of neuronal quantity；
According to following empirical formula calculate can hidden layer neuronal quantity:
In formula, NBP--- hidden layer neuron quantity；
nBP--- input layer quantity；
mBP--- output layer neuron quantity；
C --- empirical, value 0-10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810950491.4A CN109272746B (en) | 2018-08-20 | 2018-08-20 | MFD estimation method based on BP neural network data fusion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810950491.4A CN109272746B (en) | 2018-08-20 | 2018-08-20 | MFD estimation method based on BP neural network data fusion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109272746A true CN109272746A (en) | 2019-01-25 |
| CN109272746B CN109272746B (en) | 2021-06-08 |
Family
ID=65153669
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810950491.4A Active CN109272746B (en) | 2018-08-20 | 2018-08-20 | MFD estimation method based on BP neural network data fusion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109272746B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110210509A (en) * | 2019-03-04 | 2019-09-06 | 广东交通职业技术学院 | A kind of road net traffic state method of discrimination based on MFD+ spectral clustering+SVM |
| CN110610186A (en) * | 2019-04-24 | 2019-12-24 | 广东交通职业技术学院 | Road network partition based on ACO-FCM clustering algorithm and evaluation method thereof |
| CN111739293A (en) * | 2020-06-10 | 2020-10-02 | 广东世纪高通科技有限公司 | Data fusion method and device |
| CN113379754A (en) * | 2020-12-02 | 2021-09-10 | 哈尔滨理工大学 | Road center line extraction method based on vehicle-mounted GPS data and neural network |
| 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 |
| CN113537555A (en) * | 2021-06-03 | 2021-10-22 | 太原理工大学 | Traffic sub-region model prediction sliding mode boundary control method considering disturbance |
| CN113947905A (en) * | 2021-10-19 | 2022-01-18 | 交通运输部公路科学研究所 | Traffic operation situation sensing method, module and system |
| CN114358416A (en) * | 2021-12-31 | 2022-04-15 | 广东工业大学 | Public transport road network partitioning method, system, equipment and medium based on multi-source traffic data |
| CN115457764A (en) * | 2022-08-24 | 2022-12-09 | 华南理工大学 | Road section traffic density estimation method, device and medium based on vehicle track data |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101719315A (en) * | 2009-12-23 | 2010-06-02 | 山东大学 | Method for acquiring dynamic traffic information based on middleware |
| CN102509454A (en) * | 2011-11-03 | 2012-06-20 | 安徽科力信息产业有限责任公司 | Road state merging method based on floating car data (FCD) and earth magnetism detector |
| DE102016007035A1 (en) * | 2016-06-08 | 2017-12-14 | Audi Ag | Method for operating a device of a motor vehicle and associated motor vehicle |
| CN107591004A (en) * | 2017-11-01 | 2018-01-16 | 中原智慧城市设计研究院有限公司 | A kind of intelligent traffic guidance method based on bus or train route collaboration |
-
2018
- 2018-08-20 CN CN201810950491.4A patent/CN109272746B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101719315A (en) * | 2009-12-23 | 2010-06-02 | 山东大学 | Method for acquiring dynamic traffic information based on middleware |
| CN102509454A (en) * | 2011-11-03 | 2012-06-20 | 安徽科力信息产业有限责任公司 | Road state merging method based on floating car data (FCD) and earth magnetism detector |
| DE102016007035A1 (en) * | 2016-06-08 | 2017-12-14 | Audi Ag | Method for operating a device of a motor vehicle and associated motor vehicle |
| CN107591004A (en) * | 2017-11-01 | 2018-01-16 | 中原智慧城市设计研究院有限公司 | A kind of intelligent traffic guidance method based on bus or train route collaboration |
Non-Patent Citations (2)
| Title |
|---|
| LUKAS AMBÜHL 等: "Data fusion algorithm for macroscopic fundamental diagram estimation", 《TRANSPORTATION RESEARCH》 * |
| 莫尚华: "基于大数据的宏观基本图研究", 《中国优秀硕士学位论文全文数据库》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110210509A (en) * | 2019-03-04 | 2019-09-06 | 广东交通职业技术学院 | A kind of road net traffic state method of discrimination based on MFD+ spectral clustering+SVM |
| CN110610186A (en) * | 2019-04-24 | 2019-12-24 | 广东交通职业技术学院 | Road network partition based on ACO-FCM clustering algorithm and evaluation method thereof |
| CN111739293A (en) * | 2020-06-10 | 2020-10-02 | 广东世纪高通科技有限公司 | Data fusion method and device |
| CN113379754A (en) * | 2020-12-02 | 2021-09-10 | 哈尔滨理工大学 | Road center line extraction method based on vehicle-mounted GPS data and neural network |
| CN113537555A (en) * | 2021-06-03 | 2021-10-22 | 太原理工大学 | Traffic sub-region model prediction sliding mode boundary control method considering disturbance |
| 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 |
| CN113506440B (en) * | 2021-09-08 | 2021-11-30 | 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 |
| CN113947905A (en) * | 2021-10-19 | 2022-01-18 | 交通运输部公路科学研究所 | Traffic operation situation sensing method, module and system |
| CN114358416A (en) * | 2021-12-31 | 2022-04-15 | 广东工业大学 | Public transport road network partitioning method, system, equipment and medium based on multi-source traffic data |
| CN115457764A (en) * | 2022-08-24 | 2022-12-09 | 华南理工大学 | Road section traffic density estimation method, device and medium based on vehicle track data |
| CN115457764B (en) * | 2022-08-24 | 2023-07-18 | 华南理工大学 | Road section traffic density estimation method, device and medium based on vehicle track data |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109272746B (en) | 2021-06-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109272746A (en) | A kind of MFD estimating and measuring method based on BP neural network data fusion | |
| CN109308805A (en) | A kind of road network MFD estimating and measuring method based on self-adaptive weighted average data fusion | |
| CN108198415B (en) | A kind of city expressway accident forecast method based on deep learning | |
| CN105788252B (en) | Arterial street track of vehicle reconstructing method based on fixed point detector and signal timing dial data fusion | |
| CN101639871B (en) | Vehicle-borne dynamic traffic information induction system analog design method facing behavior research | |
| CN106710215B (en) | Bottleneck upstream lane grade traffic status prediction system and implementation method | |
| CN103700274B (en) | A kind of traffic flow detection induction method | |
| CN104298540B (en) | A kind of underlying model parameter correcting method of traffic simulation software | |
| CN104464310A (en) | Signal collaborative optimization control method and system of multiple intersections of urban region | |
| CN110517492A (en) | Based on the traffic route recommended method of parallel integrated study, system, device | |
| CN106530749B (en) | Signal-control crossing queue length estimation method based on single section low frequency detection data | |
| Song et al. | Delay correction model for estimating bus emissions at signalized intersections based on vehicle specific power distributions | |
| CN104778834A (en) | Urban road traffic jam judging method based on vehicle GPS data | |
| CN102176283A (en) | Traffic network simplifying model and navigating method based on same | |
| CN109348423A (en) | A kind of arterial road coordinate control optimization method based on sample path data | |
| Anya et al. | Application of AIMSUN microsimulation model to estimate emissions on signalized arterial corridors | |
| CN103500511B (en) | A kind of Intersections split control method based on car networking | |
| Zhi-Peng et al. | An improved adaptive exponential smoothing model for short-term travel time forecasting of urban arterial street | |
| CN109816983A (en) | A kind of short-term traffic flow forecast method based on depth residual error network | |
| CN106297296A (en) | A kind of fine granularity distribution method hourage based on sparse tracing point data | |
| CN106846808B (en) | A kind of vehicle parking based on license plate data time number calculating method | |
| CN108986464A (en) | Regional traffic signal control effect appraisal procedure based on Weighted Complex Networks | |
| Zhu et al. | Simulated analysis of exclusive bus lanes on expressways: case study in Beijing, China | |
| CN107092988A (en) | Method for predicting station-parking time of bus on special lane | |
| CN102074112A (en) | Time sequence multiple linear regression-based virtual speed sensor design method |
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 | ||
| CB03 | Change of inventor or designer information |
Inventor after: Huang Liang Inventor after: Lin Xiaohui Inventor after: Cao Chengtao Inventor after: Li Caihong Inventor before: Lin Xiaohui Inventor before: Cao Chengtao Inventor before: Li Caihong Inventor before: Huang Liang |
|
| CB03 | Change of inventor or designer information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |