CN108364464B - A Probabilistic Model-Based Method for Modeling Travel Time of Public Transport Vehicles - Google Patents

Abstract

The invention discloses a method for modeling the travel time of public transport vehicles based on a probability model, which belongs to the technical field of intelligent traffic information processing. The method of the invention includes: collecting and processing the operation data of public transport vehicles; using the offset lognormal distribution to fit the travel time of road sections between platforms; A first-in, first-out queue is used to model the stopping time of bus stops based on a probabilistic model; according to the travel time of each road segment and the stopping time of each platform, the travel time of the bus route is obtained, and the distribution, expectation, variance and reliability of the travel time are analyzed. The method of the invention is suitable for predicting the travel time of public transportation vehicles, and the prediction result is accurate; the invention can analyze the reasons for the fluctuation of travel time, so as to improve the service level of public transportation.

Description

A Probabilistic Model-Based Method for Modeling Travel Time of Public Transport Vehicles

technical field

The invention belongs to the technical field of intelligent traffic information processing, in particular to a method for modeling the travel time of public transport vehicles based on a probability model.

Background technique

At present, in the face of increasing urban traffic demand, road congestion, air pollution, and limited land resources, many cities have begun to implement the travel concept of "transit city". People are reducing their private car travel in favor of urban public transport. The reliability of bus travel time is the core element of public transportation service level. On the one hand, travel time reliability is an important factor to attract travelers to choose bus travel. Travel time prediction is also an important part of the intelligent passenger service system, such as on-time rate prediction. On the other hand, travel time prediction is an important indicator of bus operation, and accurate travel time prediction can help bus companies formulate countermeasures in advance, improve the operating efficiency of the bus system, and reduce operating costs.

In terms of travel time prediction, existing technologies mainly focus on the prediction of the reliability and fluctuation of private car travel time, such as dynamic path planning, estimated time of arrival, etc., and the analysis is based on road segments or paths. However, bus travel time is significantly different from private car travel time. In addition to road segment travel time, bus travel time is also affected by passengers’ boarding behavior (i.e., platform stop time). When the platform stops, the bus needs to queue up to enter the platform. , waiting for passengers to get on and off the bus, and driving from the platform to the main road, these processes are all important parts of the bus travel time, and involve the interaction of multi-line buses on the same platform.

In addition, the existing technology mainly predicts the travel time of private cars through models such as OLS (Ordinary Least Squares), SVR (Support Vector Regression), neural network, and deep learning, and the obtained result is the predicted value; for the analysis of the travel time of public transportation vehicles, These prediction techniques cannot analyze the magnitude and cause of the fluctuation of bus travel time, cannot analyze the boarding behavior of passengers, and cannot consider the interaction behavior of multi-line buses between the same platform.

SUMMARY OF THE INVENTION

The purpose of the present invention is to overcome the above-mentioned deficiencies of the prior art, and to provide a method for modeling bus travel time based on a probability model, so as to analyze the reliability of bus travel time under different traffic states and passenger demand states.

The present invention provides a probability model-based bus travel time modeling method, the specific steps are as follows:

Step 1: Collect the operation data of all public transport vehicles on the line to be studied, including: departure time from the starting station, time to arrive at each intermediate platform, time to leave each intermediate platform, time to arrive at the terminal station, and number of passengers on each platform;

Step 2, using the offset lognormal distribution to fit the travel time between the platforms;

乘客上下车时间t^b和驶入主干道时间β之和。Step 3, model the process of entering the platform of public transport vehicles. The process is: queuing to enter the platform, waiting for passengers to get on and off the bus, and entering the main road from the platform; modeling the queuing to enter the platform as a first-in-first-out (FIFO) queue, The interaction behavior of multi-line bus vehicles on the same platform is considered; the time taken for passengers to get on and off the bus is calculated according to the number of passengers getting on and off; the normal distribution is used to fit the time for bus vehicles to enter the main road from the platform. Get the stop time of the bus stop, the stop time of the bus stop is the queuing time

The sum of the passenger getting on and off time t ^b and the time β entering the main road.

Step 4, according to the travel time of the road segment and the stop time of the platform obtained in Step 2 and Step 3, calculate the route travel time of the bus vehicle, and analyze the distribution, expectation, variance and reliability of the travel time.

The route travel time T of the bus vehicle is the sum of the travel time of all road segments and the stop time of all platforms, expressed as:

Among them, IR is all the sections of the line to be studied, _SR is all the bus stops of the line to be studied, _{t i} is the travel time of the bus on the section i, and _ts is the stop time of the bus at the platform s.

Compared with the prior art, the advantages and positive effects of the present invention are:

(1) Different from the private car travel time prediction method, the method of the present invention not only considers the road segment travel time, but also considers the bus stop time, which makes the bus travel time prediction more applicable.

(2) The present invention models the bus travel time through a probability model, and uses the probability modeling method to analyze the influence of the travel time of the road section and the boarding behavior of passengers on the travel time, the interaction behavior of multi-line bus vehicles between the same platform, and obtain the travel time. In addition to the forecasting function, it can also analyze the reasons for the fluctuation of travel time and analyze the reliability of travel time, which is very important to improve the service level and operational efficiency of public transportation.

(3) The present invention models the bus stop process, and analyzes the interaction of multi-line buses on the same platform, which cannot be considered by other prediction methods.

Description of drawings

Fig. 1 is the schematic flow chart of the bus vehicle travel time modeling method of the present invention;

Fig. 2 is the schematic diagram of the method of the present invention for the modeling of bus stop platform;

3 is a schematic diagram of a bus route in an embodiment;

FIG. 4 is a schematic diagram of a model fitting result in an embodiment.

Detailed ways

The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

The present invention provides a method for modeling the travel time of public transport vehicles based on a probability model. The flow chart is shown in FIG. 1 and includes the following steps:

Step 1) Data collection and processing. The data used includes bus IC card swiping data and bus GPS data, and the following information is extracted for all operating bus vehicles on the line to be studied: departure time from the starting station, arrival time at each intermediate platform, departure time from each intermediate platform, and arrival time at the terminal station. , The number of passengers on each platform. In order to study the interaction of multi-line buses on the same platform, the operation status of different lines of buses passing through all intermediate platforms is also required, and the information to be extracted is the same as above.

Step 2) Calculate the travel time of road sections between platforms. First, the offset lognormal distribution is used to fit the travel time of the sections between the platforms. The lognormal distribution can fit the situation that the peak of the travel time distribution of the section moves forward, and the increased offset can describe the average travel time of the bus section.

Assuming that the travel time of the link between stations obeys the offset lognormal distribution, the travel time of the link between stations can be calculated as:

t _i =λ _i +exp(μ _i +σ _i z _i ) (1)

Among them, t _i represents the travel time of the bus vehicle on the road segment i, i is a positive integer; λ _i represents the unobstructed traveling time of the road segment, which is the distribution offset; exp(μ _i +σ _{i zi} ) represents the excess travel The time part obeys the lognormal distribution, where μ _i and σ _i are the mean and variance of the excess travel time, respectively, and _zi is the standard normal distribution. According to the above formula, the travel time expectation E(t _i ) and variance Var(t _i ) of road segment i can be obtained as:

Step 3) Time modeling of queuing to enter the platform. As shown in Figure 2, when the bus enters the platform, there may be other lines of buses picking up or dropping off passengers at the platform. At this time, it is necessary to wait for the other lines of buses to complete the pick-up and drop-off and drive out of the platform before the bus can To enter a station, the process can be described as a first-in, first-out queue.

Let the station serial number be _{s∈SR, SR is} the set of bus stops of the line to be studied, and s is a positive integer; let the set of all bus lines passing the station s except the line to be studied be L _s , when the line to be studied is set as L s . The probability pls that the bus from line _l is also in the queue when the bus of the line enters platform s obeys the Bernoulli distribution as follows:

为根据运营经验得到的两线路公交车辆相遇概率，M_s为待研究线路与线路l的公交车在站台s的相遇次数，N_s为待研究线路的公交车辆驶过站台s的次数，

为参数为1,

的贝努利分布。则公交车辆排队等候的时间

为：in,

is the encounter probability of two lines of bus vehicles based on operational experience, M _s is the number of encounters between the line to be studied and the bus of line l at platform s, N _s is the number of times that the bus of the line to be studied passes through platform s,

for the parameter to be 1,

the Bernoulli distribution. the waiting time of the bus

for:

where t _ls is the sum of the time of getting on and off the bus at platform s and the time of leaving the platform.

Step 4) Calculate the time taken for passengers to get on and off the bus. Assuming that it takes time α for each person to get on and off the bus, the pick-up and drop-off time of the bus depends on the maximum number of passengers getting on and off, so the pick-up and drop-off time t ^b can be calculated as

分别为下车和上车的人数。in

The number of people getting off and getting on, respectively.

Step 5) Time fitting of bus vehicles entering the main road. According to the previous research conclusions, the present invention uses a normal distribution to fit the time β of the bus entering the main road, as follows:

where μ _m is the average time of the bus entering the main road, and σ _m is the variance of this time.

Step 6) Summarize the stopping time of the bus stops. As shown in Figure 2, the bus stop time at the platform is the sum of the queuing time, the passenger alighting time, and the time to enter the main road. Considering the behavior of passengers getting on the bus, and considering the interaction between the bus vehicle and the vehicle on the trunk road in the part of entering the main road, the total stop time t _s of the bus vehicle can be calculated as follows:

The expectation and variance of the platform stop time are

Step 7) Bus route travel time fitting. The route travel time T is the sum of the road segment travel time _TL and the platform stop time T _S. The traditional private car travel time prediction method mainly considers the road segment travel time, so the platform stop behavior of public vehicles cannot be analyzed.

Among them, IR is all the sections of the research line, and _SR is all the bus stops of the research line _.

The route travel time expectation and variance are

Among them, ρ _ij represents the travel time correlation coefficient of road segment i and road segment j. Suppose the speed vectors of road segment i and road segment j are X _i and X _j respectively, then ρ _ij can be expressed as:

Example

Below in conjunction with an example as shown in FIG. 3, the modeling method of bus travel time based on the probability model of the present invention is described, and the details are as follows:

1), data collection. As shown in Figure 3, select the bus route of Hangzhou Route 68 from 2017.05.01 to 2017.05.31 as the research object. Route 68 has a total of 11 platforms and a total length of 11.73 kilometers. Extract all IC card swiping records and GPS records of 68 routes. The IC card swiping records include fields: card number, card swiping time, line ID, vehicle ID, GPS data includes the following fields: vehicle ID, time, longitude, latitude, driving out Platform identification, arrival platform identification. At the same time, extract the data of all bus lines passing through the intermediate platform of No. 68, and the data format is as above.

2), extract the departure time and the number of people on board. The arrival time is calculated as the time when the arrival platform logo appears for the first time; the departure time is the time when the departure station logo appears for the first time; the number of people getting on the bus is the number of card swiping records within the time period of two consecutive platform arrival logos. The extraction methods of other lines passing through the intermediate platform are the same as above.

3), travel time fitting of road sections between platforms. 11 bus stops divide 68 routes into 10 sections, and the fitting results with the offset lognormal distribution are shown in Table 1 below:

Table 1 Travel time of road sections

Through the Kolmogorov-Smirnov (KS) test, all fittings were significant at the 0.05 level, indicating the feasibility of the method of the present invention.

4), queuing time fitting at the station. Take the 10th platform "Wenhui Intersection" as an example, lines 135, 187, 535, 2, 105, 84 and 90 all pass through this platform; when line 68 reaches this platform, the probability that other lines are waiting for passengers to get on and off at this platform, and the average of each line. The number of people on board can be counted as shown in Table 2 below:

Table 2 The queuing time for entering the station at "Wenhui Intersection" on the platform

Assuming that it takes 3.23 seconds for each person to get on and off the bus, the queuing time of Route 68 at this platform can be calculated as: 3.23*(4.3%*3.49+9.2%*3.58+2.0%*2.96+5.4%*2.67+4.2%*1.6+ 3.6%*1.7+4.4%*2.77)=6.24s.

5), the per capita boarding time fitting. The present invention uses the least squares regression OLS to fit the relationship between the pick-up and drop-off time and the number of passengers. The obtained relationship is y=3.23x+7.21, R ² =0.346, and the regression coefficient is significant at the 0.05 level.

6) Time-consuming fitting when entering the main road. Using the normal distribution to fit the time taken to enter the main road, the following relationship is obtained

β～N(9.74,1.55 ² ) (12)

7) Calculation of platform stop time. The stop time at the platform is the sum of the queuing time, the time for passengers getting on and off the bus and the time for entering the main road. The calculation results are shown in Table 3 below:

Table 3 The platform stop time of Route 68 vehicles

8), route travel time estimation. The travel time of the route is the sum of the travel time of the road segment and the stop time of the platform.

Therefore, the expectation of the route travel time of the No. 68 car is 1333.61+216.72=1550.33s, and the variance of the route travel time of the No. 68 car is 264.39+79.99=344.38s. Through bootstrap sampling, the distribution of route travel time can be obtained, Figure 4 shows the fitted route travel time and the true route travel time. It can be seen from the figure that the difference between the travel time of the route fitted by the method of the present invention and the travel time of the real route is relatively small. Therefore, the method of the present invention can be used to more accurately analyze the travel time of the bus under different traffic states and passenger demand states. reliability to improve public transport service levels and operational efficiency.

Claims (6)

1. a bus vehicle travel time modeling method based on a probability model, is characterized in that, comprises the steps: Step 1: Collect the operation data of all public transport vehicles on the line to be studied, including: departure time from the starting station, time to arrive at each intermediate platform, time to leave each intermediate platform, time to arrive at the terminal station, and number of passengers on each platform; Step 2, using the offset lognormal distribution to fit the travel time between the platforms; The travel time t _i of the bus vehicle on the road segment i is calculated as follows: t _i =λ _i +exp(μ _i +σ _{i zi} ); Among them, λ _i represents the clear travel time of the road section, which is the distribution offset; exp(μ _i +σ _{i zi} ) represents the part beyond the clear travel time, which obeys the lognormal distribution lognormal distribution, where μ _i , σ _i are the mean and variance of the excess travel time, respectively, and _zi is the standard normal distribution; The expected E(t _i ) and variance Var(t _i ) of the travel time of road segment i are obtained as:

Step 3: Model the process of entering the bus vehicle into the platform. The process is: queuing to enter the platform, waiting for passengers to get on and off the bus, and entering the main road from the platform; modeling the queuing entering the platform as a first-in-first-out queue to calculate the bus queue Time, calculate the time of passengers getting on and off the bus according to the number of passengers getting on and off, and use the normal distribution to fit the time for the bus to enter the main road from the platform; obtain the stop time of the bus platform, and the stop time of the bus platform is the queuing time

The sum of the passenger's getting on and off time t ^b and the time β entering the main road; Calculate the queuing time of bus vehicles as follows: Assuming the bus station s of the line to be studied, the set of bus lines entering the platform except the line to be studied is L _s , the probability that the bus from line l is also in the queue when the bus of the line to be studied enters the platform s _pls is:

in,

is the encounter probability between the line to be studied and the bus of line l obtained according to the operating experience, M _s is the number of encounters between the line to be studied and the bus of line l at platform s, and N _s is the bus vehicle of the line to be studied passing through platform s number of times,

for the parameter to be 1,

The Bernoulli distribution of ; then the bus queue time

Expressed as:

Among them, t _ls is the sum of the time of getting on and off passengers at platform s and the time of leaving the platform of the bus of line l; Step 4, obtain the route travel time of the bus vehicle, and analyze the distribution, expectation, variance and reliability of the travel time; Among them, the route travel time T of the bus vehicle is the sum of the travel time of all road segments and the stop time of all platforms, which is expressed as:

IR is all the sections of the line to be studied, _SR is all the bus stops of the line to be studied, _{t i} is the travel time of the bus on the section i, and _ts is the stop time of the bus at the platform s. 2 . The method for modeling the travel time of public transport vehicles based on a probability model according to claim 1 , wherein in the step 1, operation data of public transport vehicles passing through different routes of all intermediate platforms is also collected. 3 . 3. a kind of bus vehicle travel time modeling method based on probability model according to claim 1, is characterized in that, in described step 3, the method for calculating passenger's getting on and off time is as follows: Set the time it takes for each person to get on and off the bus as α, then the time for getting on and off the bus depends on the maximum number of passengers getting on and off, and the time t ^b for passengers getting on and off is:

in,

are the number of people getting off and getting on at platform s, respectively. 4. a kind of bus vehicle travel time modeling method based on probability model according to claim 1, is characterized in that, in described step 3, the time β of bus vehicle entering main road from platform, uses normal distribution fit, expressed as follows:

Among them, μ _m and σ _m are the average time and time variance of the bus entering the main road, respectively. 5. a kind of bus vehicle travel time modeling method based on probability model according to claim 1, is characterized in that, in described step 3, bus stop time t _s at platform is as follows:

The expectations and variances of platform stop times are as follows:

6. a kind of bus vehicle travel time modeling method based on probability model according to claim 1, is characterized in that, in described step 4, The route travel time expectation E(T) is:

The route travel time variance Var(T) is:

Among them, ρ _ij represents the correlation coefficient between road segment i and road segment j. Assuming that the speed vectors of road segment i and road segment j are X _i and X _j respectively, then ρ _ij can be expressed as:

Images