CN114330106A - Urban public traffic planning method - Google Patents

Urban public traffic planning method Download PDF

Info

Publication number
CN114330106A
CN114330106A CN202111517873.6A CN202111517873A CN114330106A CN 114330106 A CN114330106 A CN 114330106A CN 202111517873 A CN202111517873 A CN 202111517873A CN 114330106 A CN114330106 A CN 114330106A
Authority
CN
China
Prior art keywords
data
urban
points
public transport
traffic
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202111517873.6A
Other languages
Chinese (zh)
Inventor
叶可江
苏林煜
须成忠
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.)
Shenzhen Institute of Advanced Technology of CAS
Original Assignee
Shenzhen Institute of Advanced Technology of CAS
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 Shenzhen Institute of Advanced Technology of CAS filed Critical Shenzhen Institute of Advanced Technology of CAS
Priority to CN202111517873.6A priority Critical patent/CN114330106A/en
Publication of CN114330106A publication Critical patent/CN114330106A/en
Priority to PCT/CN2022/138218 priority patent/WO2023109720A1/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/10—Geometric CAD
    • G06F30/18—Network design, e.g. design based on topological or interconnect aspects of utility systems, piping, heating ventilation air conditioning [HVAC] or cabling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/04—Constraint-based CAD
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]

Abstract

The invention relates to an urban public traffic planning method, which comprises the following steps: measuring and calculating urban carbon emission, and calculating to obtain the optimal amplitude which is required to be obtained by various transportation travel modes; collecting travel data of taxis and private cars of a city to be optimized; processing the collected travel data; map matching is carried out on the original GPS track points in the journey; clustering the starting point and the ending point in the data after map matching to obtain candidate public transport stations; calculating the probability from the current public transportation station to the next different public transportation station from the obtained candidate public transportation stations by using a BilSTM neural network model based on an attention mechanism; and comparing the obtained new public transport line with the existing public transport system line to check whether the expected target is achieved. The method and the device can optimize the public transport system, optimize and adjust the urban public transport system through a deep learning model method, reduce carbon emission and promote carbon neutralization.

Description

Urban public traffic planning method
Technical Field
The invention relates to an urban public traffic planning method.
Background
The existing urban public transport planning methods mainly comprise the following steps that firstly, an urban public transport system is planned by utilizing manual work and based on experience and resident travel willingness obtained by questionnaire survey, secondly, the urban public transport system is planned by utilizing a traditional statistical method and some machine learning methods, and thirdly, the public transport system is planned by utilizing a deep learning method.
In the existing method, for example, manual statistical planning depends on data from questionnaire survey, so that the time and labor are consumed, the efficiency is not high, and the planning result is not reasonable; the traditional statistical method is used for planning a public transportation system and analyzing data of start and end point outgoing data of residents and data of an existing ground public transportation system network by using machine learning, the development of the urban public transportation system is not considered, the taxi outgoing passenger flow with the maximum traffic flow in a city cannot be considered, and the trajectory data of taxis containing a large amount of information is wasted; meanwhile, due to self limitations of deep learning methods in the prior art, such as particle swarm optimization and ant colony optimization, problems of low convergence rate, large calculation amount, easy falling into local optimal solutions and the like easily occur.
Disclosure of Invention
In view of the above, there is a need for an urban public transportation planning method.
The invention provides an urban public transport planning method, which comprises the following steps: a. analyzing the existing urban passenger transport system, measuring and calculating urban carbon emission, minimizing the urban carbon emission from the aspects of environmental protection, traffic development and resident travel, taking the minimization of the urban carbon emission as an optimization target, and calculating to obtain the optimization amplitude which is required to be obtained by various traffic travel modes under the consideration of the constraints of meeting the urban traffic demand and the road resource allocation; b. collecting travel data of taxis and private cars of the city to be optimized according to the obtained optimization amplitude; c. processing the collected travel data; d, performing map matching on the original GPS track points in the journey aiming at the processed data; e. clustering the starting point and the ending point in the data after map matching to obtain candidate public transport stations; f. calculating the probability from the current public transportation station to the next different public transportation station from the obtained candidate public transportation stations by using a BilSTM neural network model based on an attention mechanism; g. selecting a plurality of adjacent stations with higher probability as alternative stations, connecting the starting points to the lines of the adjacent stations, then connecting the lines of the public transportation stations in all the obtained lines to obtain a new public transportation line, then comparing the new public transportation line with the existing public transportation system line, and checking whether the expected target is achieved.
Specifically, the step a specifically includes:
step S11: and (3) measuring and calculating the urban carbon emission:
Figure BDA0003407460980000031
wherein: q is the average daily resident trip amount, and the unit is ten thousand times;
li-the average transit distance of the traffic pattern i in km;
xi-mode allocation rate within i years of traffic mode;
ci-carbon emission factor of mode i;
step S12: the urban carbon emission is minimized under the constraint condition, and under the condition that the urban resident trip demand is met, traffic demand constraint and land resource constraint conditions are respectively obtained:
the traffic demand constraint conditions are as follows:
Figure BDA0003407460980000032
wherein: q is the total output of the single-day maneuvering mode, and the unit is the number of people;
xi-the proportion of the traffic pattern i in the total amount of travel,
li-the average travel distance of the traffic pattern i,
w-the daily average trip amount in non-motorized ways such as walking,
p-the city frequent population,
l is the average distance of single trip of urban residents,
u-average number of trips per day of urban residents,
the land resource constraint conditions are as follows:
Figure BDA0003407460980000041
in the formula xiThe proportion of the traffic mode i in the total travel amount,
Zithe dynamic floor space is occupied by all people in the mode i,
z-the area of the road occupied by all urban people;
step S13: and solving to obtain optimization amplitudes which should be obtained by various travel modes by using an objective planning method according to the objective function obtained in the step S11 and the two constraint conditions obtained in the step S12.
Specifically, the step b specifically includes:
step S21: collecting travel data of private cars and taxis;
step S22: the data is processed into a vector format.
Specifically, the step c specifically includes:
step S31: clearing incomplete data collected in the data;
step S32: directly removing data with default values;
step S33: and comparing the track points in the travel data with the latitude and longitude of the boundary of the area to be planned, and removing the data of which the starting and ending points are not in the urban planning range.
Specifically, the step d specifically includes:
step S41: mapping the GPS numerical value in the collected data to a map;
step S42: the GPS points are mapped into a coarse-grained space.
Specifically, the step S42 specifically includes:
divide the city into g x g grid cells, at this time every GPS point liIn a unit CjIn the method, all GPS points in a single unit are regarded as the same object, and the original GPS points are used
Figure BDA0003407460980000051
Figure BDA0003407460980000052
Represented as embedded dots
Figure BDA0003407460980000053
Specifically, the step e specifically includes:
step S51: obtaining candidate public traffic stations by utilizing the obtained various GPS points through a clustering algorithm;
step S52: the similarity between two GPS points is expressed by adopting the spherical distance, and the calculation formula is as follows:
Dist＝Δσ*R
Figure BDA0003407460980000054
wherein R represents the average radius of the earth, about 6378137m, Δ lat represents the latitude difference between two GPS points, Δ lng represents the longitude difference between two GPS points, Dist represents approximately the true spherical distance between two GPS points;
step S53: before grouping each point, excluding some groups which are far away from the current point, and the excluding process does not need to calculate the distance between each group one by one;
step S54: selecting points that are far from the location of the data-dense region as reference points;
step S55: according to the selected reference point, calculating the distance between the current point and the reference point before grouping each time, excluding groups with longer distance, merging the points of each group with all the points contained in all reachable groups, and using a clustering algorithm on a merged data domain;
step S56: calculating the center of the urban traffic hot spot area by using the following formula, and taking the center as the specific position of the public traffic candidate station;
Figure BDA0003407460980000055
wherein n is the number of all data objects in a certain cluster; dist (i, j) is the distance between data object i and data object j.
Specifically, the step f specifically includes:
and predicting the probability of different sites to the next adjacent site by using a BilSTM neural network model based on an attention mechanism.
Specifically, the BilSTM neural network model outputs the final probability through a full-connection layer and a Softmax classifier, wherein the full-connection layer relates to the following formula:
Figure BDA0003407460980000061
Figure BDA0003407460980000062
wherein, WFCAnd bFCAll are parameter matrixes which can be learnt by a full connection layer; next, the final prediction layer, using Softmax as a multi-class logistic regression classifier to obtain the probability distribution of the candidate destinations, and transferring T for the input partPThe jth candidate destination djAs TPProbability of true destination
Figure BDA0003407460980000066
Obtained by performing a Softmax classifier on the raw output, the final prediction result is the most probable candidate destination in the following equation:
Figure BDA0003407460980000063
1≤j≤|D|
Figure BDA0003407460980000064
using cross entropy as a loss function, the loss function is as follows:
Figure BDA0003407460980000065
specifically, the step g specifically includes:
step S71: taking each public transport station as a starting point, and obtaining the probability of the public transport station to different stations by using the BilSTM neural network model in the step f; then, selecting a plurality of adjacent sites with higher probability as alternative sites, and connecting the starting point to the lines of the adjacent sites;
step S72: after the calculation of all public transport stations is completed, connecting the lines in which the public transport stations are stored in all the obtained lines, then optimizing the existing public transport lines according to the optimization amplitude obtained in the step S13, comparing the optimized public transport lines with the existing public transport lines, and adjusting the existing ground conventional public transport lines and track transport lines;
step S73: and inputting the adjusted route into a digital twin city of the optimized city to be modified, and checking whether the expected target is achieved.
Compared with the prior art, the method and the device can optimize the public transport system, optimize and adjust the urban public transport system through a deep learning model method, reduce carbon emission and promote carbon neutralization. The method has the advantages that the operation efficiency is higher, the training time is shorter, meanwhile, the map embedding method is used, so that the robustness is better, and finally, the model can notice the relevance among different sites by using the attention mechanism instead of simply paying attention to the positions of the sites in the sequence.
Drawings
FIG. 1 is a flow chart of an urban public transportation planning method of the present invention;
fig. 2 is a general flowchart of a GDBSCAN algorithm according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a BilSTM neural network model provided by an embodiment of the present invention;
FIG. 4 is a schematic diagram illustrating the use of an attention mechanism in a model according to an embodiment of the present invention;
fig. 5 is a schematic diagram of a BilTM neural network model based on an attention mechanism according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
Fig. 1 is a flow chart showing the operation of the urban public transportation planning method according to the preferred embodiment of the invention.
And step S1, analyzing the existing urban passenger transport system, and measuring and calculating the urban carbon emission. The method mainly aims at minimizing the carbon emission of the traffic from three aspects of environmental protection, traffic development and resident travel, and establishes a multi-objective hybrid optimization model under the constraint of meeting urban traffic demands and road resource allocation, so as to obtain the optimization amplitude which is required to be obtained by various traffic travel modes. Specifically, the method comprises the following steps:
step S11: and (4) measuring and calculating the urban carbon emission. In this embodiment, a mature IPCC traffic carbon emission measurement method is adopted, and according to the IPCC guideline report, the formula is as follows:
Figure BDA0003407460980000081
wherein: q is the average daily resident trip amount, and the unit is ten thousand times;
li——the average passing distance of the traffic mode i is km;
xi-mode allocation rate within i years of traffic mode;
cithe carbon emission factor of the transportation mode i is expressed in kg/(man km).
In this embodiment, taking the beijing city as an example, the average carbon dioxide emission factors of several major trip modes are obtained according to the data in the yearbook for traffic development in the beijing city, as shown in the following table:
Figure BDA0003407460980000082
based on the formula, the embodiment performs accounting on the daily average carbon dioxide emission of the major urban passenger transportation mode in Beijing. Taking 2018 as an example, the total carbon emission of a conventional bus is 1374.0 tons, the total carbon emission of a rail transit is 850.6 tons, the total carbon emission of a taxi is 1355.1 tons, and the total carbon emission of a private car is 14312.3 tons. It is clear that high private car reserves are a significant cause of higher total carbon emissions for this trip mode.
Step S12: the carbon emission is minimized under the constraint condition, and the trip requirements of urban residents are met no matter how the carbon emission is optimized, so that traffic requirement constraint and land resource constraint conditions are obtained respectively. The method specifically comprises the following steps:
obtaining traffic demand constraints:
Figure BDA0003407460980000091
wherein: q is the total output of the single-day maneuvering mode, and the unit is the number of people;
xi-the proportion of traffic patterns i in the total amount of travel, in units;
li-average distance of travel in traffic mode i in km;
w is the total daily trip amount in units of times in non-motorized modes such as walking and the like;
p-city permanent population, unit person;
l is the average distance of single trip of urban residents, and the unit is km;
u-average number of trips per day of urban residents in units of times
Besides the requirement of urban resident travel, with the advance of urbanization process, traffic resources gradually become a great important factor limiting urban traffic construction, and the problem of outstanding land resource shortage becomes a key factor limiting urban traffic development. The land resource constraint condition requires that the average per-capita dynamic floor area of all traffic ways is less than the per-capita road area of a city, and specifically comprises the following steps:
Figure BDA0003407460980000092
in the formula xiThe traffic mode i in the total travel amount is obtained in unit percent;
Zithe average person dynamic floor area of the mode i is one square meter per person;
z-square meter per person occupying road area
Step S13: and solving to obtain optimization amplitudes which should be obtained by various travel modes by using an objective planning method according to the objective function obtained in the step S11 and the two constraint conditions obtained in the step S12.
The goal of this embodiment is to increase the percentage of regular bus trips by about 10% and the rail transit by about 80%.
And step S2, additionally modifying the existing public transportation line according to the obtained optimization amplitude obtained in the step S1 so as to achieve the public transportation ratio of urban residents. The existing public transportation includes: ground conventional public transportation and rail transit.
The present embodiment is intended to achieve this goal by studying travel data in existing cities. And collecting travel data of taxis and private cars of the city to be optimized. Specifically, the method comprises the following steps:
step S21: and collecting travel data of private cars and taxis.
Step S22: and the data is processed into a vector format to facilitate subsequent processing.
In step S3, the data obtained in step S2 is processed to prevent the data from affecting the analysis result.
The method comprises the steps of firstly removing incomplete data collected in data, and then removing some abnormal point data, for example, removing data of which the longitude and the latitude of a starting and finishing point exceed boundaries.
The method specifically comprises the following steps:
step S31: the data is pre-processed.
Step S32: data for which a default value exists is directly removed.
Step S33: and comparing the track points in the travel data with the latitude and longitude of the boundary of the area to be planned, and removing the data of which the starting and ending points are not in the urban planning range.
And step S4, performing map matching on the original GPS track points in the journey according to the processed data.
In order to enhance the robustness of data, the original taxi GPS track points are relocated to a traffic network on an electronic map by the map matching technology, so as to obtain more robust GPS track points.
Step S41: and mapping the GPS value in the collected data to a map for subsequent processing.
Step S42: in this case, the original GPS point has a problem that the original latitude and longitude values are easily deviated when uncertainty is generated due to a low sampling rate or noise.
The solution of this embodiment is to map the GPS points into a coarse-grained space. I.e. the city is divided into g x g grid cells, at which time each GPS point liIn a unit CjIn order to make the GPS track points more robust, all GPS points in a single cell are treated as the same object. In this way, the original GPS point
Figure BDA0003407460980000111
Can be expressed as embeddedPoint of (2)
Figure BDA0003407460980000112
And step S5, clustering the start point and the end point in the data after map matching to obtain candidate public transportation sites.
And clustering the urban hot spot area by using the designed improved DBSCAN clustering algorithm, and taking the generated result as a candidate public transport station (including a ground bus station and a subway station). The method specifically comprises the following steps:
step S51: and obtaining candidate public transport stations by utilizing the obtained various GPS points through a clustering algorithm.
Step S52: there are many calculation formulas for the distance between two points, such as Euclidean distance, Manhattan distance, etc. In this embodiment, the spherical distance is used to represent the similarity between two GPS points, and the calculation formula is as follows:
Dist＝Δσ*R
Figure BDA0003407460980000121
where R represents the average radius of the earth, about 6378137m, Δ lat represents the latitude difference between two GPS points, Δ lng represents the longitude difference between two GPS points, and Dist represents approximately the true spherical distance between two GPS points.
Step S53: the conventional DBSCAN algorithm takes too long. The GDBSCAN algorithm improved based on the DBSCAN algorithm saves part of the time, but still needs to spend a lot of time for distance calculation in the grouping process. In order to further increase the clustering speed, the present embodiment improves the grouping condition by using the characteristics of the GPS data to obtain the GDBSCAN algorithm. The general flow chart of the GDBSCAN algorithm is shown in fig. 2. In the GDBSCAN algorithm, groups which are far away from the current point are excluded before grouping the points, and the exclusion process does not need to calculate the distance between each group one by one, so that a large amount of time and resources can be saved. This step needs to be done by means of reference points, the method for selecting reference points will be described in detail in step S54.
Step S54: in theory, the reference points may be random, but considering the initial purpose of selecting the reference points to minimize the distance calculation, the selection needs to follow a certain rule. Experiments prove that when points in the data dense area position are selected as reference points, the clustering speed is low, so that positions far away from the data dense area are selected as much as possible when the reference points are selected, and the clustering speed can be improved better.
Step S55: after the reference point is selected, the distance between the current point and the reference point is calculated before grouping each time, and the farther groups are excluded by utilizing the distance. The distance between the master point and the reference point of each group is fixed, the master point and the reference point can be stored to be used for a plurality of times when other points are grouped subsequently, and the distance calculation can be greatly reduced by utilizing the two groups with longer distances to be excluded before grouping. Where the master point is the first point to join a group when the group is created.
After grouping, all data points are grouped into different circular areas with eps as radius (eps values are obtained by DBSCAN algorithm), and the reachable groups of each group are also known. The points of each group are then merged with all the points contained within all its reachable groups, and then a clustering algorithm is used on the merged data fields.
Step S56: the center of the urban traffic hot spot area is calculated by the following formula, and the center is used as the specific position of the public traffic candidate station.
Figure BDA0003407460980000131
In this formula, n is the number of all data objects in a certain cluster; DiSt (i, j) is the distance between data object i and data object j, which is calculated by step S52. For any one class cluster, the present embodiment takes the data object in the class cluster with the least sum of distances to other data objects as the center point of the class cluster, i.e. the position representing the candidate public transportation station in the city. This means that the longitude and latitude coordinates of the cluster center point calculated by the formula are the longitude and latitude coordinates of the candidate public transportation station calculated in this embodiment.
Step S6: and calculating the probability from the current public transportation station to the next different public transportation station from the candidate public transportation stations obtained by the method by using a BilSTM neural network model based on the attention mechanism. The method specifically comprises the following steps:
step S61: after the embedded method is adopted, the track space described by the GPS is represented by a series of station generations. The present embodiment employs a BilSTM neural network model based on an attention mechanism to predict the probability of a different site to the next site in the neighborhood.
Step S62: the BilSTM neural network model is shown in FIG. 3, and T is represented byiTrack data e of timeiIntroduced into the BiLSTM neural network model, the traditional LSTM structure can extract the information contained therein, however, visited sites have different discrimination capabilities in determining direction, for example, at intersections, and therefore it is considered that the contribution weight of GPS should take such capabilities into account, rather than looking at their locality in the trajectory alone. Note that the attentive mechanism can capture the weight cases of hi in the sequence at different locations, while the unidirectional LSTM lacks subsequent trajectory information, e.g., e1 → e2 → e3 lacks information of the following location, and vice versa, e.g., e5 → e4 → e3, so that the BiLSTM and attentive mechanism are mutually supportive in learning the potential sequence dependencies of the trajectory data. The present embodiment chooses to use a BilSTM neural network model based on the attention mechanism to simulate the environment before and in the future for each GPS point.
Step S63: unlike standard LSTM, BiLSTM integrates previous and future sequential features therein by performing forward and backward processes simultaneously. Where the unidirectional LSTM is denoted x ═ e per input sequence1,e2,…,eN}(e∈R3×V) Where N represents the input sequence length size, 3 represents the three day types, and V is the dimension of the input vector. Further, the forward derivation formula is as follows:
Figure BDA0003407460980000141
Figure BDA0003407460980000142
Figure BDA0003407460980000143
Figure BDA0003407460980000144
Figure BDA0003407460980000145
Figure BDA0003407460980000146
wherein->Represents direction, σ represents sigmoid activation function, in、fn、on、gnAnd hnRepresenting input gate, forgetting gate, output gate, modulation gate and hidden state, respectively, and a parameter Wi、Wf、WoAnd WcRespectively, indicate the weight matrices of the above, which indicates the product of the elements. Since the backward and forward processes are in principle the same, but the sequence order is reversed, the derivation of the backward process only needs to be done>Is changed into<-is finished. Finally, the total output H of the BilSTM cells of input sequence X is represented as follows:
H＝{h1，h2，…，hn，...hN}，
Figure BDA0003407460980000151
wherein h isnIs the BiLSTM output at n steps, representing the forward and reverse hidden states.
Attention mechanism role in the BilSTM model: it has been noted previously that mechanisms have been widely used in sequence modeling and transduction models, allowing for modeling the dependence of a location on other locations, rather than its location in the input sequence. Embodiments use an attention mechanism for the purpose of capturing the relevance between already passed locations pairs to determine the location of the next station.
Step S64: note that the force mechanism is used in this model as shown in FIG. 4. In order to make the experimental model more robust, the original GPS data that has not undergone map matching is also synchronously transmitted into the model in the present embodiment, and features are extracted together with the embedded data, which are respectively denoted as h and h'. Wherein
Figure BDA0003407460980000152
And |, represent addition and multiplication of corresponding elements, respectively. More specifically, after the input sequence is input to BilSTM, H is inputiSending to a perceptron, and obtaining m after processing by the perceptroniThen to m1，m2，...，mNAnd (4) deducing the association degree between each GPS point through Softmax processing, then giving a weight to each point regularization, and finally obtaining a final expression of the embedded sequence X. The derivation formula of each involved term is as follows:
mi＝tanh(Whhi+bh)+tanh(Whh′i+bh)
Figure BDA0003407460980000161
Figure BDA0003407460980000162
wherein, WhAnd bhAre weights in the perceptron.
Step S65: FIG. 5 shows a BilsTM neural network model based on the attention mechanism, with the present embodiment outputting the final probabilities through the fully connected layer and the Softmax classifier. The formula involved in the full connection layer is as follows:
Figure BDA0003407460980000163
Figure BDA0003407460980000164
wherein, WFCAnd bFCAre all fully connected layers learnable parameter matrices. Next, the final prediction layer, in this embodiment, Softmax is used as a multi-class logistic regression classifier to obtain the probability distribution of the candidate destination. For input part transfer TPThe jth candidate destination djAs TPOf the true destination (y)
Figure BDA0003407460980000165
Obtained by performing a Softmax classifier on the raw output. The final prediction result is the most probable candidate destination in the following equation.
Figure BDA0003407460980000166
1≤j≤|D|
Figure BDA0003407460980000167
The present embodiment uses cross entropy as a loss function, as this is typically used to calculate the distance between the predicted probability distribution and the true probability distribution in the Softmax classifier. The loss function is as follows:
Figure BDA0003407460980000168
step S7: for each site, taking it as a starting point, the probability of it to a different site is obtained. And selecting a plurality of adjacent sites with higher probability as alternative sites, and connecting the starting points to the lines of the adjacent sites. And then connecting the lines in which the public transportation stops are stored in all the obtained lines to obtain a new public transportation line, comparing the new public transportation line with the lines of the existing public transportation system, and considering the optimization change of the existing public transportation system, such as the establishment of bus stops, the establishment of subway stops and the like. And detecting the modified digital twin city of the optimized city to check whether the expected target is achieved. Specifically, the method comprises the following steps:
step S71: for each public transportation station, taking it as a starting point, the probability of it going to a different station is obtained by using the BilSTM neural network model of step S6. And then selecting a plurality of nearby sites with higher probability as alternative sites, and connecting the starting points to the lines of the nearby sites.
Step S72: after the calculation of all the public transportation stations is completed, the lines with the public transportation stations in all the obtained lines are connected, so that new lines with larger demand can be obtained. And then, optimizing the existing public transportation line according to the optimization index obtained in the step S13, comparing the optimized public transportation line with the existing public transportation line, and adjusting the existing ground conventional public transportation line and rail transportation line.
Step S73: and inputting the adjusted route into a digital twin city of the optimized city to be modified, and checking whether the expected target is achieved.
The method takes the urban passenger transport system as an object, performs travel structure optimization through a multi-objective optimization model to obtain an optimization target, and then finishes planning and adjusting the urban public transport system by using a deep learning method so as to achieve the aims of reducing carbon emission and assisting in carbon neutralization.
Although the present invention has been described with reference to the presently preferred embodiments, it will be understood by those skilled in the art that the foregoing description is only illustrative of the present invention and is not intended to limit the scope of the invention, and that any modifications, equivalents, improvements and the like made within the spirit and scope of the present invention are intended to be included within the scope of the claims.

Claims (10)

1. An urban public transport planning method is characterized by comprising the following steps:
a. analyzing the existing urban passenger transport system, measuring and calculating urban carbon emission, minimizing the urban carbon emission from the aspects of environmental protection, traffic development and resident trip as an optimization target, and calculating to obtain the optimization amplitude which is required to be obtained by various traffic trip modes under the consideration of the constraints of meeting the urban traffic demand and the road resource allocation;
b. collecting travel data of taxis and private cars of the city to be optimized according to the obtained optimization amplitude;
c. processing the collected travel data;
d. performing map matching on the original GPS track points in the journey aiming at the processed data;
e. clustering the starting point and the ending point in the data after map matching to obtain candidate public transport stations;
f. calculating the probability from the current public transportation station to the next different public transportation station from the obtained candidate public transportation stations by using a BilSTM neural network model based on an attention mechanism;
g. selecting a plurality of nearby stops with higher probability as alternative stops, connecting the starting points to the lines of the nearby stops, then connecting the lines of the public transportation stops in all the obtained lines to obtain a new public transportation line, then comparing the new public transportation line with the existing public transportation system line, and checking whether the expected target is achieved.
2. The method according to claim 1, wherein said step a specifically comprises:
step S11: and (3) measuring and calculating the urban carbon emission:
Figure FDA0003407460970000011
wherein: q is the average daily resident trip amount, and the unit is ten thousand times;
li-the average transit distance of the traffic pattern i in km;
xi-mode allocation rate within i years of traffic mode;
ci-carbon emission factor of mode i;
step S12: the urban carbon emission is minimized under the constraint condition, and under the condition that the urban resident trip demand is met, traffic demand constraint and land resource constraint conditions are respectively obtained:
the traffic demand constraint conditions are as follows:
Figure FDA0003407460970000021
wherein: q is the total output of the single-day maneuvering mode, and the unit is the number of people;
xi-the proportion of the traffic pattern i in the total amount of travel,
li-the average travel distance of the traffic pattern i,
w-the daily average trip amount in non-motorized ways such as walking,
p-the city frequent population,
l is the average distance of single trip of urban residents,
u-average number of trips per day of urban residents,
the land resource constraint conditions are as follows:
Figure FDA0003407460970000022
in the formula xiThe proportion of the traffic mode i in the total travel amount,
Zithe dynamic floor space is occupied by all people in the mode i,
z-the area of the road occupied by all urban people;
step S13: and solving and obtaining the optimization amplitude which is required to be obtained by various travel modes by using an objective planning method according to the objective function obtained in the step S11 and the two constraint conditions obtained in the step S12.
3. The method according to claim 2, wherein said step b specifically comprises:
step S21: collecting travel data of private cars and taxis;
step S22: the data is processed into a vector format.
4. The method according to claim 3, wherein said step c specifically comprises:
step S31: clearing incomplete data collected in the data;
step S32: directly removing data with default values;
step S33: and comparing the track points in the travel data with the latitude and longitude of the boundary of the area to be planned, and removing the data of which the starting and ending points are not in the urban planning range.
5. The method according to claim 4, wherein said step d specifically comprises:
step S41: mapping the GPS numerical value in the collected data to a map;
step S42: the GPS points are mapped into a coarse-grained space.
6. The method according to claim 5, wherein the step S42 specifically includes:
divide the city into g x g grid cells, at this time every GPS point liIn a unit CjIn the method, all GPS points in a single unit are regarded as the same object, and the original GPS points are used
Figure FDA0003407460970000034
Figure FDA0003407460970000032
Represented as embedded dots
Figure FDA0003407460970000033
7. The method according to claim 6, wherein said step e specifically comprises:
step S51: obtaining candidate public transport stations by utilizing the obtained various GPS points through a clustering algorithm;
step S52: the similarity between two GPS points is expressed by adopting the spherical distance, and the calculation formula is as follows:
Dist＝Δσ*R
Figure FDA0003407460970000031
wherein R represents the average radius of the earth, about 6378137m, Δ lat represents the latitude difference between two GPS points, Δ lng represents the longitude difference between two GPS points, Dist represents approximately the true spherical distance between two GPS points;
step S53: before grouping each point, excluding some groups which are far away from the current point, and the excluding process does not need to calculate the distance between each group one by one;
step S54: selecting points that are far from the location of the data-dense region as reference points;
step S55: according to the selected reference point, calculating the distance between the current point and the reference point before grouping each time, excluding groups with longer distance, merging the points of each group with all the points contained in all reachable groups, and using a clustering algorithm on a merged data domain;
step S56: calculating the center of the urban traffic hot spot area by using the following formula, and taking the center as the specific position of the public traffic candidate station;
Figure FDA0003407460970000041
wherein n is the number of all data objects in a certain cluster; dist (i, j) is the distance between data object i and data object j.
8. The method according to claim 7, wherein said step f specifically comprises:
and predicting the probability of different sites to the next adjacent site by using a BilSTM neural network model based on an attention mechanism.
9. The method of claim 8, wherein the BilsTM neural network model outputs the final probability by a fully-connected layer and a Softmax classifier, the fully-connected layer involving the formula:
Figure FDA0003407460970000042
Figure FDA0003407460970000043
wherein, WFCAnd bFCAll are parameter matrixes which can be learnt by a full connection layer; next, the final prediction layer, using Softmax as a multi-class logistic regression classifier to obtain the probability distribution of the candidate destinations, and for the input partial transition TPThe jth candidate destination djAs TPProbability of true destination
Figure FDA0003407460970000054
Obtained by performing a Softmax classifier on the raw output, the final prediction result is the most probable candidate destination in the following equation:
Figure FDA0003407460970000051
Figure FDA0003407460970000052
using cross entropy as a loss function, the loss function is as follows:
Figure FDA0003407460970000053
10. the method according to claim 9, wherein said step g comprises in particular:
step S71: taking each public transport station as a starting point, and obtaining the probability of the public transport station to different stations by using the BilSTM neural network model in the step f; then, selecting a plurality of adjacent sites with higher probability as alternative sites, and connecting the starting point to the lines of the adjacent sites;
step S72: after the calculation of all public transport stations is completed, connecting the obtained lines in which the public transport stations are stored in all the lines, then optimizing the existing public transport lines according to the optimization amplitude obtained in the step S13, comparing the optimized public transport lines with the existing public transport lines, and adjusting the existing ground conventional public transport lines and track transport lines;
step S73: and inputting the adjusted route into a digital twin city of the optimized city to be modified, and checking whether the expected target is achieved.
CN202111517873.6A 2021-12-13 2021-12-13 Urban public traffic planning method Pending CN114330106A (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN202111517873.6A CN114330106A (en) 2021-12-13 2021-12-13 Urban public traffic planning method
PCT/CN2022/138218 WO2023109720A1 (en) 2021-12-13 2022-12-09 Urban public transportation planning method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202111517873.6A CN114330106A (en) 2021-12-13 2021-12-13 Urban public traffic planning method

Publications (1)

Publication Number Publication Date
CN114330106A true CN114330106A (en) 2022-04-12

Family

ID=81050942

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202111517873.6A Pending CN114330106A (en) 2021-12-13 2021-12-13 Urban public traffic planning method

Country Status (2)

Country Link
CN (1) CN114330106A (en)
WO (1) WO2023109720A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023109720A1 (en) * 2021-12-13 2023-06-22 深圳先进技术研究院 Urban public transportation planning method

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117272848B (en) * 2023-11-22 2024-02-02 上海随申行智慧交通科技有限公司 Subway passenger flow prediction method and model training method based on space-time influence
CN117593167B (en) * 2024-01-18 2024-04-12 山东国建土地房地产评估测绘有限公司 Intelligent city planning management method and system based on big data

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105489006B (en) * 2015-12-15 2018-01-12 浙江工业大学 Multiple dimensioned vehicle flow visual analysis method based on GPS data from taxi
JP6543215B2 (en) * 2016-05-12 2019-07-10 日本電信電話株式会社 Destination prediction apparatus, destination prediction method, and destination prediction program
CN107657330A (en) * 2017-08-16 2018-02-02 深圳先进技术研究院 A kind of candidate's public bus network computational methods, system and electronic equipment
CN107481511A (en) * 2017-08-16 2017-12-15 深圳先进技术研究院 A kind of method and system for calculating candidate bus station
CN110909434B (en) * 2019-10-11 2023-03-14 东南大学 Urban public transport trunk line network design method under low-carbon guidance
CN112561146B (en) * 2020-12-08 2023-04-18 哈尔滨工程大学 Large-scale real-time traffic flow prediction method based on fuzzy logic and depth LSTM
CN113486719A (en) * 2021-06-08 2021-10-08 南京邮电大学 Vehicle destination prediction method, vehicle destination prediction device, computer equipment and storage medium
CN114330106A (en) * 2021-12-13 2022-04-12 深圳先进技术研究院 Urban public traffic planning method

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023109720A1 (en) * 2021-12-13 2023-06-22 深圳先进技术研究院 Urban public transportation planning method

Also Published As

Publication number Publication date
WO2023109720A1 (en) 2023-06-22

Similar Documents

Publication Publication Date Title
CN114330106A (en) Urban public traffic planning method
Huang et al. A novel bus-dispatching model based on passenger flow and arrival time prediction
CN106910199B (en) Car networking crowdsourcing method towards city space information collection
WO2023029234A1 (en) Method for bus arrival time prediction when lacking data
KR101742043B1 (en) Apparatus and method for travel mode choice prediction
CN110909434B (en) Urban public transport trunk line network design method under low-carbon guidance
CN107656987A (en) A kind of subway station function method for digging based on LDA models
CN112347596B (en) Urban public transport network optimization method
CN109325614B (en) Bus stop site selection method based on GIS
Zhang et al. Using street view images to identify road noise barriers with ensemble classification model and geospatial analysis
CN112785077A (en) Travel demand prediction method and system based on space-time data
CN113159371B (en) Unknown target feature modeling and demand prediction method based on cross-modal data fusion
Yang et al. Dynamic origin-destination matrix estimation based on urban Rail transit AFC data: deep optimization framework with forward passing and backpropagation techniques
Xu et al. A taxi dispatch system based on prediction of demand and destination
Yang et al. Spatiotemporal DeepWalk gated recurrent neural network: a deep learning framework for traffic learning and forecasting
Wang et al. Segmented trajectory clustering-based destination prediction in IoVs
ZHAO et al. Big data-driven residents’ travel mode choice: a research overview
Ali et al. A novel transfer learning approach to detect the location of transformers in distribution network
Lu et al. Optimization of vehicle automatic navigation path based on remote sensing and GIS
Zhou et al. Refined taxi demand prediction with ST-Vec
Bi et al. Mining taxi pick-up hotspots based on grid information entropy clustering algorithm
Díaz-Parra et al. Editorial for Volume 11 Number 2: Electric School Bus Routing Problem for Smart Cities
CN114923512A (en) Urban pollutant monitoring optimization method based on shared bicycle and taxi movement tracks
Liu et al. A genetic-algorithm-based neural network approach for short-term traffic flow forecasting
Almansoub et al. Exploring the Effects of Transportation Supply on Mixed Land-Use at the Parcel Level. Land 2022, 11, 797

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