CN109272746B - MFD estimation method based on BP neural network data fusion - Google Patents

MFD estimation method based on BP neural network data fusion Download PDF

Info

Publication number
CN109272746B
CN109272746B CN201810950491.4A CN201810950491A CN109272746B CN 109272746 B CN109272746 B CN 109272746B CN 201810950491 A CN201810950491 A CN 201810950491A CN 109272746 B CN109272746 B CN 109272746B
Authority
CN
China
Prior art keywords
road network
road
network
weighted
mfd
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
CN201810950491.4A
Other languages
Chinese (zh)
Other versions
CN109272746A (en
Inventor
黄�良
林晓辉
曹成涛
李彩红
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.)
Guangdong Communications Polytechnic
Original Assignee
Guangdong Communications Polytechnic
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 Guangdong Communications Polytechnic filed Critical Guangdong Communications Polytechnic
Priority to CN201810950491.4A priority Critical patent/CN109272746B/en
Publication of CN109272746A publication Critical patent/CN109272746A/en
Application granted granted Critical
Publication of CN109272746B publication Critical patent/CN109272746B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

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
    • G08G1/0108—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

Abstract

The invention relates to the technical field of neural networks, in particular to an MFD estimation method based on BP neural network data fusion, which combines traffic data obtained by an LDD estimation method and an FCD estimation method and considers the traffic data by 100 percent of networking vehicle data (C and C) under the Internet of vehiclesNetwork Car Data，NCD) The estimated traffic parameters are inspection data, and BP neural network data fusion models of the road network weighted traffic flow and the road network weighted traffic density are respectively established, so that more accurate road network weighted traffic flow and road network weighted traffic density can be obtained, and the road network MFD can be estimated more accurately. Carrying out BP neural network data fusion on traffic data collected by a fixed detector and a floating car on a road section with the fixed detector to obtain road section weighted traffic flow and weighted traffic density; and extracting the weighted traffic flow and the weighted traffic density of the road section by using the traffic data collected by the floating car on the road section without the fixed detector. And finally estimating the MFD of the road network according to the weighted traffic flow and the weighted traffic density of the road network with data fusion.

Description

MFD estimation method based on BP neural network data fusion
Technical Field
The invention relates to the technical field of neural network methods, in particular to an MFD estimation method based on BP neural network data fusion.
Background
Urban traffic congestion presents a significant challenge to urban traffic. How to relieve the problem of urban traffic congestion becomes a key research direction of numerous scholars, and the scholars propose various traffic control strategies, so that the urban traffic congestion is effectively relieved to a certain extent, but with the continuous increase of traffic flow, the traffic congestion becomes more severe, and various traffic control strategies are not applicable. Recently, two scholars, namely Daganzo and gerroliminis, have disclosed the objective existence of a Macroscopic Fundamental Map (MFD), and they consider that MFD can not only describe an urban road network from a Macroscopic level, but also monitor and predict the traffic operation state of the road network, so as to provide a new idea for implementing a traffic control strategy on an oversaturated road network from the Macroscopic level, however, how to obtain the MFD of the urban road network becomes a big difficulty. The road network MFD can be estimated by using traffic data collected by stationary detectors (e.g., toroidal induction coils, microwave, video detectors, etc.) or GPS float cars. The fixed Detector Data estimation method (Loop Detector Data, LDD estimation method) is a method of collecting traffic Data in real time by fixed detectors installed on a road section, and estimating the road network MFD by using the MFD correlation theory. The Floating Car Data estimation method (FCD estimation method) is a method for estimating the MFD of a road network by acquiring traffic Data of Floating cars in the road network in real time through vehicles (such as taxis and buses) equipped with GPS vehicle terminals and obtaining a weighted traffic flow and a weighted traffic density of the road network by using a driving trajectory estimation method proposed by Edie (1963). Some scholars research two estimation methods, for example (Courbon et al, 2011) compare and analyze three road network MFD estimation methods, such as a theoretical analysis method, an LDD estimation method, and an FCD estimation method, and research the influence of the fixed detector position and the floating car coverage balance on the MFD estimation. (Nagle et al, 2013) propose that when the coverage of a floating car is at least 15%, a more accurate road network MFD can be obtained by using an FCD estimation method, but the premise of using this method is that the coverage of the floating car must be known and the distribution of the floating cars in the road network is uniform. (Lu et al, 2013) estimate the road network MFD using actual intersection video detection data and taxi float data, and find that the data processing time interval affects the estimation result of the road network MFD. (Leclercq et al, 2014) compared the difference between the two road network MFD estimation methods, such as the LDD estimation method and the FCD estimation method, using the traffic data obtained from the actual road network, and investigated the application range of the two methods. (Du et al, 2016) estimate road network MFD by estimating road network traffic flow and traffic density using a small number of floating car data, assuming that the proportion of floating cars in a particular start-to-end point is known, and estimating the equivalent proportion of floating cars needed to estimate road network traffic flow, for the actual case where floating car coverage is not uniform in the road network. But in fact, the fixed detector can only collect traffic data of partial road sections, the road sections without the fixed detector cannot obtain the traffic data, the coverage rate of the GPS floating car is low, the traffic data amount is insufficient, and the estimated road network MFD has a large error. (Ambuhl L et al, 2016) proposes to perform data fusion on the two methods to estimate more accurate road network MFD aiming at the situation that most of documents only adopt one of the methods to estimate road network MFD and rarely combine the two methods together, but the data fusion algorithm is obtained through a large amount of empirical experiments and has no universal applicability.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide an MFD estimation method based on BP neural Network data fusion, traffic data obtained by an LDD estimation method and an FCD estimation method are considered in a combined mode, traffic parameters estimated by 100% networking vehicle data (Network data, NCD) under the Internet of vehicles are used as inspection data, BP neural Network data fusion models of road Network weighted traffic flow and road Network weighted traffic density are respectively established, so that more accurate road Network weighted traffic flow and road Network weighted traffic density can be obtained, and the MFD road Network is estimated more accurately. Carrying out BP neural network data fusion on traffic data collected by the fixed detector and the floating car on a road section with the fixed detector to obtain road section weighted traffic flow and weighted traffic density; and extracting the weighted traffic flow and the weighted traffic density of the road section by using the traffic data collected by the floating car on the road section without the fixed detector. And finally, estimating the MFD of the road network according to the road network weighted traffic flow and the weighted traffic density of the data fusion.
In order to solve the technical problems, the invention adopts the technical scheme that:
the MFD estimation method based on BP neural network data fusion comprises the following specific steps:
(1) firstly, selecting an experimental area, establishing a traffic simulation model by using Vissim traffic simulation software according to the basic data, setting 5 percent of vehicles as floating vehicles, arranging a detector at the middle position of each road section, setting 100 percent of vehicles as networking vehicles (special floating vehicles), and establishing a vehicle networking environment traffic model for obtaining check data of a neural network model; in order to simulate the whole process of a road network from low peak to high peak to congestion, in the road network simulation model, a simulated traffic flow starts from low peak, the driving traffic volume of each road section on the boundary of the road network is increased by 100pcu/h every 900s until reaching the oversaturation state of the high peak, the simulation is carried out for 30000s, 1 time of data is collected every 300s, and 100 times of data are collected;
(2) after the step (1), setting data such as the number of vehicles, running time, running distance and the like uploaded by the floating vehicles and the networked vehicles every 5 seconds, and calculating the weighted traffic flow q of the floating vehicle road network every 300 seconds according to an FCD estimation methodFCDWeighted traffic density k of floating car road networkFCDWeighted traffic flow q of network-connected vehicle road networkNCDWeighted traffic density k of road network of networked vehiclesNCD(ii) a Similarly, the road section traffic density and the road section traffic flow of each road section detector are collected every 300 seconds, and the weighted traffic flow q of the road network every 300 seconds is calculated according to an LDD (light detection diode) estimation methodLDDAnd road network weighted traffic density kLDD。
(3) After step (2), q for the first 50 cyclesFCD、qLDD、nFCDAs a training sample of a road network weighted traffic flow fusion model, q of the first 50 periodsNCDAs a test sample, training a road network weighted traffic flow fusion model; k of the first 50 periodsFCD、kLDD、nFCDAs a training sample of a road network weighted traffic density fusion model, k of the first 50 periodsNCDAs a test sample, training a road network weighted traffic density fusion model; finally, performing data fusion on the data of 100 periods by using the trained neural network model;
(4) after step (3), road network MFDs based on the FCD estimation are generated respectivelyFRoad network MFD based on LDD estimationLMFD estimated based on BP neural network data fusionBPRoad network standard MFD based on the track of networked vehiclesNAnd comparing the average relative error of the determined road network MFD and the road network standard MFD.
The invention provides an MFD estimation method based on BP neural Network Data fusion, which combines traffic Data obtained by an LDD estimation method and an FCD estimation method for consideration, takes traffic parameters estimated by 100% networking vehicle Data (Network Car Data, NCD) under the Internet of vehicles as inspection Data, and respectively establishes BP neural Network Data fusion models of road Network weighted traffic flow and road Network weighted traffic density so as to obtain more accurate road Network weighted traffic flow and road Network weighted traffic density, thereby more accurately estimating the road Network MFD. Carrying out BP neural network data fusion on traffic data collected by a fixed detector and a floating car on a road section with the fixed detector to obtain road section weighted traffic flow and weighted traffic density; and extracting the weighted traffic flow and the weighted traffic density of the road section by using the traffic data collected by the floating car on the road section without the fixed detector. And finally estimating the MFD of the road network according to the weighted traffic flow and the weighted traffic density of the road network with data fusion.
Preferably, in step (2), the LDD estimation method includes the following specific steps:
(1) firstly, each road section in the road network is provided with a fixed detector (such as a loop coil detector, a video acquisition detector and the like), so that the road section traffic flow and the traffic density acquired by the fixed detector can be directly used for estimating the road network MFD,
(2) after step (1), according to the MFD-related theory proposed by Geroliiminis and Daganzo,2008[1～6]Thus, it can be seen that:
Figure BDA0001771360150000041
in the formula: n is the number of moving vehicles (veh) of the road network;
qw、kw、ow-road network weighted traffic flow (veh/h), road network weighted traffic density (veh/km), road network weighted time occupancy;
i、li-road section i and the length of the road section (km);
qi、ki、oi-traffic (veh/h), density (vehk/km) and time occupancy of the stretch i;
s-average vehicle length of the vehicle.
Preferably, in step (2), the FCD estimation method comprises the following specific steps:
when the tracks of all vehicles in the road network are known, the traffic flow and the traffic density of the road network can be calculated according to the tracks of the vehicles, and the formula is as follows:
Figure BDA0001771360150000042
Figure BDA0001771360150000051
in the formula: k is road network traffic density, veh/km;
q-road network traffic flow, veh/h;
m is the number of vehicles recorded in the acquisition period T;
n-total number of road sections in road network
tj-collecting the travel time, s, of the jth vehicle within the period T;
li-length of the ith road segment, m;
t-acquisition period, s;
dj is the running distance m of the jth vehicle in the acquisition period T;
Tm-collecting the sum of the driving times, s, of all vehicles of the road network within the period T.
Dm-collecting the sum of the driving distances s of all vehicles of the road network within the period T.
If it is difficult to actually acquire the running states (running distance and running time) of all vehicles, acquiring the running state of a part of floating vehicles; nagle (2014) proposes that assuming that the proportion ρ of the floating cars in the road network is known and is uniformly distributed in each region of the road network, the traffic flow and the traffic density of the road network can be estimated according to the formula:
Figure BDA0001771360150000052
Figure BDA0001771360150000053
in the formula:
Figure BDA0001771360150000061
-road network traffic density, veh/km, estimated using floating car data;
Figure BDA0001771360150000062
-using the floating car data to estimate road network traffic flow, veh/h;
m' -collecting the number of the floating vehicles recorded in the period T;
n-total number of road sections in road network
tj′-collecting the travel time, s, of the jth floating car within the period T;
li-length of the ith road segment, m;
t-acquisition period, s;
dj′-collecting the distance, m, traveled by the jth vehicle during the period T; .
Preferably, the BP neural network consists of an input layer, a hidden layer and an output layer, and the learning algorithm is a global approximation method; and performing data fusion on the road network weighted traffic flow and the road network weighted traffic density calculated by the LDD estimation method and the FCD estimation method by using a BP neural network model to obtain the road network weighted traffic flow and the road network weighted traffic density so as to estimate the MFD of the road network.
Preferably, the two key parameters estimated by the road network MFD are a road network weighted traffic flow and a road network weighted traffic density, and a road network weighted traffic flow fusion model and a road network weighted traffic density fusion model based on the BP neural network are designed respectively; the method comprises the following specific steps:
(1) firstly, inputting data and outputting data;
for the road network weighted traffic flow fusion model, the input data mainly comprises road network weighted traffic flow qLDD obtained by LDD estimation method, road network weighted traffic flow qFCD obtained by FCD estimation method, road network floating vehicle sample number nFCCD, namely neural network inputThe input layer has 3 parameters, and the output layer is the weighted traffic flow of the fused road network
Figure BDA0001771360150000063
For the road network weighted traffic density fusion model, the input data mainly comprises road network weighted traffic density kLDD obtained by an LDD estimation method, road network weighted traffic density kFCD obtained by an FCD estimation method, the number nFCCD of road network floating car samples, namely, 3 parameters are arranged on the input layer of a neural network, and the output layer is the fused road network weighted traffic density
Figure BDA0001771360150000071
(2) After the step (1), determining the number of network layers;
the number of layers of the BP neural network at least comprises 3 layers (an input layer, a hidden layer and an output layer), the hidden layer can be more than 1 layer, but the more the number of hidden layers is, the more complex the network is, and the longer the training time is. In general, a three-layer network including 1 hidden layer can meet most application requirements, and the number of network layers of two models is set to be 3;
(3) after step (2), determination of the number of neurons;
the number of neurons in the cryptic layer was calculated according to the following empirical formula:
Figure BDA0001771360150000072
in the formula, NBP-the number of hidden layer neurons;
nBP-input layer neuron number;
mBP-output layer neuron number;
c is an empirical constant with the value of 0-10.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides an MFD estimation method based on BP neural Network Data fusion, which combines traffic Data obtained by an LDD estimation method and an FCD estimation method for consideration, takes traffic parameters estimated by 100% networking vehicle Data (Network Car Data, NCD) under the Internet of vehicles as inspection Data, and respectively establishes BP neural Network Data fusion models of road Network weighted traffic flow and road Network weighted traffic density so as to obtain more accurate road Network weighted traffic flow and road Network weighted traffic density, thereby more accurately estimating the road Network MFD. Carrying out BP neural network data fusion on traffic data collected by a fixed detector and a floating car on a road section with the fixed detector to obtain road section weighted traffic flow and weighted traffic density; and extracting the weighted traffic flow and the weighted traffic density of the road section by using the traffic data collected by the floating car on the road section without the fixed detector. And finally estimating the MFD of the road network according to the weighted traffic flow and the weighted traffic density of the road network with data fusion.
Drawings
Fig. 1 is a schematic diagram of a network structure of a road network weighted traffic flow fusion model according to an embodiment.
Fig. 2 is a schematic diagram of a network structure of a road network weighted traffic density fusion model.
FIG. 3 is a schematic diagram of a layout of a core network in a river.
FIG. 4 is a road network weighted traffic flow comparison graph for various estimation methods.
FIG. 5 is a road network weighted traffic density comparison graph for various estimation methods.
Fig. 6 is a table diagram of the average of absolute values of phase errors relative to qNCD.
Fig. 7 is a table of averages of absolute values of phase errors relative to kBP.
FIG. 8 is a schematic diagram of a road network two-dimensional MFD for various estimation methods.
Fig. 9 is a schematic diagram of the resulting three-dimensional MFD of the road network for various estimation methods.
FIG. 10 is a diagram of a fitting function for each MFD.
Detailed Description
The present invention will be further described with reference to the following embodiments. Wherein the showings are for the purpose of illustration only and are shown by way of illustration only and not in actual form, and are not to be construed as limiting the present patent; to better illustrate the embodiments of the present invention, some components of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product; it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted.
The same or similar reference numerals in the drawings of the embodiments of the present invention correspond to the same or similar components; in the description of the present invention, it should be understood that if there is an orientation or positional relationship indicated by the terms "upper", "lower", "left", "right", etc. based on the orientation or positional relationship shown in the drawings, it is only for convenience of description and simplification of description, but it is not intended to indicate or imply that the referred device or element must have a specific orientation, be constructed in a specific orientation, and be operated, and therefore, the terms describing the positional relationship in the drawings are only for illustrative purposes and are not to be construed as limitations of the present patent, and specific meanings of the above terms may be understood by those skilled in the art according to specific situations.
Examples
Fig. 1 to 10 show an embodiment of an MFD estimation method based on BP neural network data fusion according to the present invention, which includes the following specific steps:
(1) firstly, selecting an experimental area, establishing a traffic simulation model by using Vissim traffic simulation software according to the basic data, setting 5 percent of vehicles as floating vehicles, arranging a detector at the middle position of each road section, setting 100 percent of vehicles as networking vehicles (special floating vehicles), and establishing a vehicle networking environment traffic model for obtaining check data of a neural network model; in order to simulate the whole process of a road network from low peak to high peak to congestion, in the road network simulation model, a simulated traffic flow starts from low peak, the driving traffic volume of each road section on the boundary of the road network is increased by 100pcu/h every 900s until reaching the oversaturation state of the high peak, the simulation is carried out for 30000s, 1 time of data is collected every 300s, and 100 times of data are collected;
(2) after the step (1), setting data such as the number of vehicles, running time, running distance and the like uploaded by the floating vehicles and the networked vehicles every 5 seconds, and calculating the weighted traffic flow of the floating vehicle network every 300 seconds according to an FCD estimation methodqFCDWeighted traffic density k of floating car road networkFCDWeighted traffic flow q of network-connected vehicle road networkNCDWeighted traffic density k of road network of networked vehiclesNCD(ii) a Similarly, the road section traffic density and the road section traffic flow of each road section detector are collected every 300 seconds, and the weighted traffic flow q of the road network every 300 seconds is calculated according to an LDD (light detection diode) estimation methodLDDAnd road network weighted traffic density kLDD。
(3) After step (2), q for the first 50 cyclesFCD、qLDD、nFCDAs a training sample of a road network weighted traffic flow fusion model, q of the first 50 periodsNCDAs a test sample, training a road network weighted traffic flow fusion model; k of the first 50 periodsFCD、kLDD、nFCDAs a training sample of a road network weighted traffic density fusion model, k of the first 50 periodsNCDAs a test sample, training a road network weighted traffic density fusion model; finally, performing data fusion on the data of 100 periods by using the trained neural network model;
(4) after step (3), road network MFDs based on the FCD estimation are generated respectivelyFRoad network MFD based on LDD estimationLMFD estimated based on BP neural network data fusionBPRoad network standard MFD based on the track of networked vehiclesNAnd comparing the average relative error of the determined road network MFD and the road network standard MFD.
In the step (2), the LDD estimation method specifically includes the following steps:
(1) firstly, each road section in the road network is provided with a fixed detector (such as a loop coil detector, a video acquisition detector and the like), so that the road section traffic flow and the traffic density acquired by the fixed detector can be directly used for estimating the road network MFD,
(2) after step (1), according to the MFD-related theory proposed by Geroliiminis and Daganzo,2008[1～6]Thus, it can be seen that:
Figure BDA0001771360150000101
in the formula: n is the number of moving vehicles (veh) of the road network;
qw、kw、ow-road network weighted traffic flow (veh/h), road network weighted traffic density (veh/km), road network weighted time occupancy;
i、li-road section i and the length of the road section (km);
qi、ki、oi-traffic (veh/h), density (vehk/km) and time occupancy of the stretch i;
s-average vehicle length of the vehicle.
In step (2), the FCD estimation method includes the following specific steps:
when the tracks of all vehicles in the road network are known, the traffic flow and the traffic density of the road network can be calculated according to the tracks of the vehicles, and the formula is as follows:
Figure BDA0001771360150000102
Figure BDA0001771360150000111
in the formula: k is road network traffic density, veh/km;
q-road network traffic flow, veh/h;
m is the number of vehicles recorded in the acquisition period T;
n-total number of road sections in road network
tj-collecting the travel time, s, of the jth vehicle within the period T;
li-length of the ith road segment, m;
t-acquisition period, s;
dj is the running distance m of the jth vehicle in the acquisition period T;
Tm-collecting the sum of the driving times, s, of all vehicles of the road network within the period T.
Dm-collecting all vehicles of the road network within a period TSum of driving distances, s.
If it is difficult to actually acquire the running states (running distance and running time) of all vehicles, acquiring the running state of a part of floating vehicles; nagle (2014) proposes that assuming that the proportion ρ of the floating cars in the road network is known and is uniformly distributed in each region of the road network, the traffic flow and the traffic density of the road network can be estimated according to the formula:
Figure BDA0001771360150000112
Figure BDA0001771360150000113
in the formula:
Figure BDA0001771360150000121
-road network traffic density, veh/km, estimated using floating car data;
Figure BDA0001771360150000122
-using the floating car data to estimate road network traffic flow, veh/h;
m' -collecting the number of the floating vehicles recorded in the period T;
n-total number of road sections in road network
tj'-collecting the travel time, s, of the jth floating car within the period T;
li-length of the ith road segment, m;
t-acquisition period, s;
dj'-collecting the distance, m, traveled by the jth vehicle during the period T; .
The BP neural network consists of an input layer, a hidden layer and an output layer, and the learning algorithm of the BP neural network is a global approximation method; and performing data fusion on the road network weighted traffic flow and the road network weighted traffic density calculated by the LDD estimation method and the FCD estimation method by using a BP neural network model to obtain the road network weighted traffic flow and the road network weighted traffic density so as to estimate the MFD (road volume density).
The two key parameters estimated by the road network MFD are a road network weighted traffic flow and a road network weighted traffic density, and a road network weighted traffic flow fusion model and a road network weighted traffic density fusion model based on a BP neural network are designed respectively; the method comprises the following specific steps:
(1) firstly, inputting data and outputting data;
for the road network weighted traffic flow fusion model, the input data mainly comprises road network weighted traffic flow qLDD obtained by an LDD estimation method, road network weighted traffic flow qFCD obtained by an FCD estimation method, road network floating car sample number nFCCD, namely a neural network input layer has 3 parameters, and the output layer is fused road network weighted traffic flow
Figure BDA0001771360150000123
For the road network weighted traffic density fusion model, the input data mainly comprises road network weighted traffic density kLDD obtained by an LDD estimation method, road network weighted traffic density kFCD obtained by an FCD estimation method, the number nFCCD of road network floating car samples, namely, 3 parameters are arranged on the input layer of a neural network, and the output layer is the fused road network weighted traffic density
Figure BDA0001771360150000131
(2) After the step (1), determining the number of network layers;
the number of layers of the BP neural network at least comprises 3 layers (an input layer, a hidden layer and an output layer), the hidden layer can be more than 1 layer, but the more the number of hidden layers is, the more complex the network is, and the longer the training time is. In general, a three-layer network including 1 hidden layer can meet most application requirements, and the number of network layers of two models is set to be 3;
(3) after step (2), determination of the number of neurons;
the number of neurons in the cryptic layer was calculated according to the following empirical formula:
Figure BDA0001771360150000132
in the formula, NBP-the number of hidden layer neurons;
nBP-input layer neuron number;
mBP-output layer neuron number;
c is an empirical constant with the value of 0-10.
N of the above two modelsBP＝3，mBPC takes 7, so the number of hidden layer neurons for both models is 9. The structures of the BP neural networks of the two models are shown in fig. 1 and fig. 2.
In this embodiment, a core road network intersection group in the Tianhequn area of Guangzhou city is selected as a research object, and the road network is composed of main roads such as a sky river road, a sky river east road, a sky river north road, a sports west road, a sports east road and the like, and partial branches, and includes more than 7 plane intersections and more than 20 entrances and exits, as shown in fig. 3. The traffic data is based on the data detected by the SCATS traffic number control system in 2017, 8, month, 6 and 6 peak hours (18: 00-19: 00), as shown in FIG. 3.
For q of 100 cyclesBP、qFCD、qLDD、qNCDData comparison was performed, as shown in FIG. 4, for k of 100 cyclesBP、kFCD、kLDD、kNCDData comparison was performed as shown in fig. 5:
as can be seen from fig. 4-5, the magnitude of the change of the road network weighted traffic flow and the road network weighted traffic density obtained by the FCD estimation method is large because the number of floating cars is small; the change of the road network weighted traffic flow and the road network weighted traffic density obtained by the LDD estimation method and the vehicle networking estimation is stable and the change trend is consistent, the road network weighted traffic flow and the road network weighted traffic density are gradually increased along with the passing of the simulation time, the stable value is maintained in a period of time, and then the stable value is rapidly reduced, but the road network weighted traffic flow and the road network weighted traffic density obtained by the LDD estimation method are both smaller than the road network weighted traffic flow and the road network weighted traffic density estimated by the vehicle networking, because when the data acquisition interval is reached, a small number of vehicles do not reach the fixed detector.
Analyzing the simulation data to obtain qLDD、qFCD、qBPAnd q isNCDAnd k is the sum of the average of the absolute values of the relative errorsLDD、kFCD、kBPAnd k isNCDThe average value of the absolute values of the relative errors in (2) is shown in fig. 6 and 7.
As is clear from FIGS. 6 and 7, q isFCDAnd kFCDThe average of absolute values of the phase errors of (a) is the largest, 11.52% and 12.26%, respectively; q. q.sLDDAnd kLDDThe average absolute value of the phase error of (a) is 8.22% and 11.54%, respectively; after data fusion by BP neural network, qBPAnd kBPHas an average value of 6.2% and 7.2%, respectively, which is closest to the standard value qNCDAnd kNCD。
Generating a road network MFD based on the FCD estimation method by using various estimation dataFRoad network MFD based on LDD estimation methodLMFD estimated based on BP neural network data fusionBPRoad network standard MFD based on the track of networked vehiclesNAs shown in fig. 8.
As can be seen from FIG. 8, MFDFExhibit a large dispersion, MFDL、MFDBP、MFDNThe points of the traffic flow are concentrated, the weighted traffic flow of the road network and the weighted traffic density of the road network are gradually increased along with the simulation time, the weighted traffic density of the road network is started from 70veh/km, the road network is maintained at a high weighted traffic flow, the weighted traffic flow of the road network is sharply reduced along with the increase of the weighted traffic density of the road network, and the road network is in a saturated state. Meanwhile, the road network MFD also has a hysteresis phenomenon, and accords with the characteristics of the road network MFD. Performing data fitting on scattered points of each MFD to obtain a fitting function, and calculating the optimal weighted traffic density k of each fitting formula0And maximum weighted traffic flow qmaxAnd solving for k for each MFD0And q ismaxK from the standard value0(NCD)And q ismax(NCD)Phase error, as shown in fig. 9.
As can be seen from FIG. 9, MFDFK of (a)0And q ismaxThe relative error was maximal at 18.43% and 5.32%, respectively. MFDBPAnd MFDLK of (a)0And q ismaxRelative error is small, but MFDBPK of (a)0And q ismaxCloser to the standard MFDNK of (a)0(NCD)And q ismax(NCD)0.86% and-4.05%, respectively. Therefore, the MFD of the road network after the BP neural network data fusion is more accurate.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. It will be apparent to those skilled in the art that other variations and modifications can be made on the basis of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.

Claims (5)

1. An MFD estimation method based on BP neural network data fusion is characterized by comprising the following specific steps:
(1) firstly, selecting an experimental area, establishing a traffic simulation model by using Vissim traffic simulation software, setting 5 percent of vehicles as floating vehicles, setting a detector at the middle position of each road section, setting 100 percent of vehicles as networking vehicles, and establishing a traffic model of a vehicle networking environment for obtaining check data of a neural network model; in order to simulate the whole process of a road network from low peak to high peak to jam, in a traffic simulation model of the road network, the simulated traffic flow starts from the low peak, the driving traffic volume of each road section at the boundary of the road network is increased by 100pcu/h every 900s until reaching the oversaturation state of the peak, the simulation is carried out for 30000s, 1 time of data is collected every 300s, and 100 times of data are collected;
(2) after the step (1), the number of vehicles uploaded by the floating vehicles and the networked vehicles every 5 seconds, the running time and the running distance are set, and the weighted traffic flow q of the floating vehicle road network every 300 seconds is calculated according to the FCD estimation methodFCDFloating car road network weightingTraffic density kFCDWeighted traffic flow q of network-connected vehicle road networkNCDAnd weighted traffic density k of networked vehicle road networkNCD(ii) a Similarly, the road section traffic density and the road section traffic flow of each road section detector are collected every 300 seconds, and the weighted traffic flow q of the road network every 300 seconds is calculated according to an LDD (light detection diode) estimation methodLDDAnd road network weighted traffic density kLDD；
(3) After step (2), q for the first 50 cyclesFCD、qLDD、nFCDTraining samples as a road network weighted traffic flow fusion model, nFCDQ of the first 50 periods for the number of road network floating car samplesNCDAs a test sample, training a road network weighted traffic flow fusion model; k of the first 50 periodsFCD、kLDD、nFCDAs a training sample of a road network weighted traffic density fusion model, k of the first 50 periodsNCDAs a test sample, training a road network weighted traffic density fusion model; finally, performing data fusion on the data of 100 periods by using the trained neural network model;
(4) after step (3), road network MFDs based on the FCD estimation are generated respectivelyFRoad network MFD based on LDD estimationLMFD estimated based on BP neural network data fusionBPRoad network standard MFD based on the track of networked vehiclesNAnd comparing the average relative error of the determined road network MFD and the road network standard MFD.
2. The MFD estimation method based on BP neural network data fusion as claimed in claim 1, wherein in step (2), the LDD estimation method comprises the following specific steps:
(1) firstly, each road section in the road network is provided with a fixed detector, so that road traffic flow and traffic density acquired by the fixed detectors can be directly used for estimating road network MFD;
(2) after step (1), according to the MFD-related theory, it can be seen that:
Figure FDA0002991413220000021
in the formula: n is the number veh of moving vehicles in the road network;
qw、kw、ow-road network weighted traffic flow veh/h, road network weighted traffic density veh/km, road network weighted time occupancy;
i、li-road section i and the length km of this road section;
qi、ki、oi-traffic veh/h, density vehk/km and time occupancy of road section i;
s-average vehicle length of the vehicle.
3. The MFD estimation method based on BP neural network data fusion according to claim 2, wherein in step (2), the FCD estimation method comprises the following specific steps:
when the tracks of all vehicles in the road network are known, the traffic flow and the traffic density of the road network can be calculated according to the tracks of the vehicles, and the formula is as follows:
Figure FDA0002991413220000022
Figure FDA0002991413220000031
in the formula: k is road network traffic density, veh/km;
q-road network traffic flow, veh/h;
m is the number of vehicles recorded in the acquisition period T;
n is the total number of road sections in the road network;
tj-collecting the travel time, s, of the jth vehicle within the period T;
li-length of the ith road segment, m;
t-acquisition period, s;
dj is the running distance m of the jth vehicle in the acquisition period T;
Tmcollecting the sum of the driving time s of all vehicles in the road network in the period T;
Dmcollecting the sum of the driving distances of all vehicles in the road network within the period T, s;
if the running states of all vehicles are difficult to obtain actually, the running states of part of floating vehicles are obtained; assuming that the proportion ρ of the floating cars in the road network is known and evenly distributed in each region of the road network, the traffic flow and the traffic density of the road network can be estimated according to the above formula (2) and formula (3), which are as follows:
Figure FDA0002991413220000032
Figure FDA0002991413220000033
in the formula:
Figure FDA0002991413220000034
-road network traffic density, veh/km, estimated using floating car data;
Figure FDA0002991413220000041
-using the floating car data to estimate road network traffic flow, veh/h;
m' -collecting the number of the floating vehicles recorded in the period T;
n-total number of road sections in road network
tj'-collecting the travel time, s, of the jth floating car within the period T;
li-length of the ith road segment, m;
t-acquisition period, s;
dj'-collecting the distance traveled, m, by the jth vehicle within the period T.
4. The MFD estimation method based on BP neural network data fusion as claimed in claim 1, wherein the BP neural network is composed of three layers of input layer, hidden layer and output layer, and its learning algorithm is a global approximation method; and performing data fusion on the road network weighted traffic flow and the road network weighted traffic density calculated by the LDD estimation method and the FCD estimation method by using a BP neural network model to obtain the road network weighted traffic flow and the road network weighted traffic density so as to estimate the MFD of the road network.
5. The MFD estimation method based on BP neural network data fusion as claimed in claim 1, wherein two key parameters of road network MFD estimation are road network weighted traffic flow and road network weighted traffic density, and a road network weighted traffic flow fusion model and a road network weighted traffic density fusion model based on BP neural network are respectively designed; the method comprises the following specific steps:
(1) firstly, inputting data and outputting data;
for the road network weighted traffic flow fusion model, the input data mainly comprises the road network weighted traffic flow q obtained by an LDD estimation methodLDDRoad network weighted traffic flow q obtained by FCD estimation methodFCDNumber n of floating car samples in road networkFCDThat is, the input layer of the neural network has 3 parameters, and the output layer is the weighted traffic flow of the fused road network
Figure FDA0002991413220000051
For the road network weighted traffic density fusion model, the input data mainly comprises the road network weighted traffic density k obtained by the LDD estimation methodLDDRoad network weighted traffic density k obtained by FCD estimation methodFCDNumber n of floating car samples in road networkFCDThat is, the input layer of the neural network has 3 parameters, and the output layer is the weighted traffic density of the fused road network
Figure FDA0002991413220000052
(2) After the step (1), determining the number of network layers;
the number of layers of the BP neural network at least comprises 3 layers of an input layer, a hidden layer and an output layer, wherein the number of the hidden layers is one or more than one, and the more the number of the hidden layers is, the more complex the network is and the longer the training time is; the three-layer network comprising 1 hidden layer can meet the application requirement, and the number of network layers of the two models is set to be 3 preliminarily;
(3) after step (2), determination of the number of neurons;
the number of neurons in the cryptic layer was calculated according to the following empirical formula:
Figure FDA0002991413220000053
in the formula, NBP-the number of hidden layer neurons;
nBP-input layer neuron number;
mBP-output layer neuron number;
c is an empirical constant with the value of 0-10.
CN201810950491.4A 2018-08-20 2018-08-20 MFD estimation method based on BP neural network data fusion Active CN109272746B (en)

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 CN109272746A (en) 2019-01-25
CN109272746B true 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)

Families Citing this family (8)

* Cited by examiner, † Cited by third party
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
CN113537555B (en) * 2021-06-03 2023-04-11 太原理工大学 Traffic sub-region model prediction sliding mode boundary control method considering disturbance
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
CN113947905B (en) * 2021-10-19 2022-09-06 交通运输部公路科学研究所 Traffic operation situation sensing method, module and system
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

Citations (4)

* Cited by examiner, † Cited by third party
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

Patent Citations (4)

* Cited by examiner, † Cited by third party
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)

* Cited by examiner, † Cited by third party
Title
Data fusion algorithm for macroscopic fundamental diagram estimation;Lukas Ambühl 等;《Transportation Research》;20160805;全文 *
基于大数据的宏观基本图研究;莫尚华;《中国优秀硕士学位论文全文数据库》;20180115;全文 *

Also Published As

Publication number Publication date
CN109272746A (en) 2019-01-25

Similar Documents

Publication Publication Date Title
CN109272746B (en) MFD estimation method based on BP neural network data fusion
Essa et al. Simulated traffic conflicts: do they accurately represent field-measured conflicts?
CN104658252B (en) Method for evaluating traffic operational conditions of highway based on multisource data fusion
Orfila et al. An android based ecodriving assistance system to improve safety and efficiency of internal combustion engine passenger cars
CN108960426A (en) Road grade Synthesize estimation system based on BP neural network
Anand et al. Traffic density estimation under heterogeneous traffic conditions using data fusion
CN106710215B (en) Bottleneck upstream lane grade traffic status prediction system and implementation method
CN107346460A (en) Following operating mode Forecasting Methodology based on the lower front truck operation information of intelligent network contact system
Fernandes et al. Impacts of roundabouts in suburban areas on congestion-specific vehicle speed profiles, pollutant and noise emissions: An empirical analysis
Li et al. Fundamental diagrams of commercial adaptive cruise control: Worldwide experimental evidence
CN109308805A (en) A kind of road network MFD estimating and measuring method based on self-adaptive weighted average data fusion
Loulizi et al. Steady-state car-following time gaps: an empirical study using naturalistic driving data
CN109272745A (en) A kind of track of vehicle prediction technique based on deep neural network
CN105869402B (en) Express highway section speed modification method based on polymorphic type floating car data
Anya et al. Application of AIMSUN microsimulation model to estimate emissions on signalized arterial corridors
Aljamal et al. Real-time estimation of vehicle counts on signalized intersection approaches using probe vehicle data
Zhu et al. An automated vehicle fuel economy benefits evaluation framework using real-world travel and traffic data
CN109118787A (en) A kind of car speed prediction technique based on deep neural network
Gebhard et al. Range prediction for electric bicycles
Zhang et al. Driver curve speed model and its application to ACC speed control in curved roads
Patnaik et al. Development of roundabout entry capacity model using INAGA method for heterogeneous traffic flow conditions
Prontri et al. Traffic signal control using fuzzy logic
Zhai et al. Comparative analysis of drive-cycles, speed limit violations, and emissions in two cities: Toronto and Beijing
Li et al. Calibrating VISSIM roundabout model using a critical gap and follow-up headway approach
Adamec et al. Modelling of traffic load by the datafromsky system in the smart city concept

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