CN104376716B - Method for dynamically generating bus timetables on basis of Bayesian network models - Google Patents

Abstract

The invention relates to a method for dynamically generating bus timetables based on a Bayesian network model, comprising: screening micro and macro factors that affect the dynamic generation of bus timetables; constructing a two-layer microcosmic and macroscopic Bayesian network for dynamically generating bus timetables Model, that is: constructing a Bayesian network model for forecasting dynamic changes in the bus environment and dynamically generating a Bayesian network model for bus schedules; predicting the probability of occurrence of various lines' capacity and volume under random interference and the reasons for their imbalance; combined with scheduling strategies, Generate possible timetable schemes around the goal of timely evacuation of passengers; from the perspective of the government, enterprises and passengers, calculate various indicators for evaluating the quality of timetables and evaluate their pros and cons. The invention can realize the function of dynamically adjusting the timetable according to the change of the bus environment, and provides technical support for the daily operation and management of the bus.

Description

A Dynamic Generation Method of Bus Timetable Based on Bayesian Network Model

technical field

The invention relates to the technical field of public transport informationization, in particular to a method for dynamically generating a public transport schedule based on a Bayesian network model.

Background technique

The compilation of bus timetable is one of the core tasks of the daily operation of public transport. According to the spatial and temporal distribution characteristics of residents’ travel, the frequency and type of departures in each time period are reasonably organized and arranged, and the problem of the maximum matching of transport capacity and traffic volume is mainly solved. When random factors in reality cause changes in bus passenger flow or travel time, this will cause an imbalance in bus transport capacity and volume, and thus the bus dispatching scheme will fail. Therefore, it is of great theoretical value and practical significance to dynamically adjust the timetable according to the dynamic changes of the bus environment.

The factors that directly determine the failure of the bus timetable are the capacity and volume of the uplink and downlink, which are affected by external environments such as weather changes, traffic congestion, and large-scale events. When detecting traffic events, evaluate the degree to which they affect passenger flow or travel time. Then analyze the cause of failure of the timetable, and dynamically adjust the timetable accordingly. At present, many domestic and foreign scholars dynamically adjust the timetable according to the dynamic changes of the bus environment. There are two main research ideas:

1. Predict changes in passenger flow or travel time. On the one hand, use methods such as multiple linear regression and structural equations to qualitatively and quantitatively explore the correlation among many factors that affect passenger flow or travel time changes, and conduct related sensitivity analysis; on the other hand, Using the time series method, treat it as a black box, directly reveal the evolution trend of passenger flow or travel time changes, and provide data support for the compilation of timetables.

2. Compile the timetable. On the one hand, on the basis of the above work, study the statistical laws of passenger flow and travel time. When detecting traffic events, build a timetable compilation model and use the optimization theory to generate the timetable; on the other hand, use the neural network Such as artificial simulation technology, simulating the thinking mode of dispatchers, and adjusting the schedule according to changes in the environment.

It can be seen from the above that the existing research methods cannot solve the chain reaction process of the dynamic change of the bus schedule caused by random interference. We should start from the whole and reveal how changes in the external environment cause changes in passenger flow or travel time, and then affect how the dynamic adjustment process of the schedule occurs. , and the complex relationships among them, such as triggering, interfering, transforming and coupling, to predict the generation and probability of bus schedules under complex traffic environment changes.

Bayesian network is a probabilistic graphical model that describes the causal relationship between things, which is very suitable for modeling and analyzing the occurrence of emergencies and the chain reaction process caused by them. Based on this, the present invention analyzes the causes of the imbalance between the transport capacity and traffic volume that affect the dynamic generation of bus timetables, regards external environmental factors as input, analyzes how it affects passenger flow or travel time changes, and then how to cause the imbalance between bus transport capacity and traffic volume. Unbalanced, according to the existing capacity configuration, the output is the result of the maximum matching capacity of the timetable, and the control input can control the change of some states. Based on this, the external environmental condition nodes in the emergency Bayesian network are constructed to input passenger flow or travel time The four-layer topological network structure output by the decision-making node of the timetable for capacity and volume calculation realizes the prediction of the bus timetable and its occurrence probability in the case of complex traffic environment changes, and provides reliable technical support for the dynamic bus dispatching.

Contents of the invention

The invention provides a dynamic generation method of bus timetable based on Bayesian network model, on the basis of analyzing various influencing factors affecting the dynamic change of bus timetable, combined with actual bus dynamic data, to describe the causal relationship between them , when detecting changes in the external environment based on the intelligent bus dispatching platform, infer the reasons for the unbalanced transport capacity and traffic volume under various complex traffic environments and the probability of occurrence, and calculate the departure frequency and its dispatching type accordingly. The timetable is dynamically generated in response to environmental changes, providing technical support for the daily operation and management of public transport.

The solution of the present invention is achieved through the following technical solutions:

The invention provides a method for dynamically generating a bus timetable based on a Bayesian network model, comprising the following steps:

(1) Using a combination of quantitative and qualitative methods to screen out many external environmental factors that affect the dynamic generation of bus timetables, including traffic accidents that cause changes in travel time, which in turn cause insufficient transport capacity or large-scale events that cause passenger flow fluctuations, which in turn cause insufficient transport capacity;

(2) Construct the upper and lower two-layer Bayesian network model for the dynamic generation of bus schedules, in which: the upper model describes the random disturbance of the external environment that causes changes in passenger flow or travel time, and is the input condition of the lower model, which describes the resulting transport capacity and traffic Random interference from the external environment with unbalanced quantity provides data support for schedule generation;

(3) When a traffic event occurs, combined with the real-time operation data of the intelligent bus dispatching platform, using the probability of occurrence of various line capacity and volume predicted by the micro-macro model in step (2) and the cause of the imbalance, combined with the dispatching strategy, the minimum cost Timely evacuation of passengers is the goal, and various timetable schemes are generated;

(4) Calculate the indicators of passenger waiting time, station retention, and operating costs for each timetable scheme in step (3), and evaluate its pros and cons.

As an improvement, a combination of quantitative and qualitative methods is used to screen the process of many factors that affect the dynamic generation of bus schedules, including:

(1) From the microscopic point of view, analyze how N random factors of the external environment X=(X ₁ , X ₂ ,...,X _N ) interfere with passenger flow fluctuation pf _t (X) or travel time change pt _t (X ), such as: road types, road conditions, traffic accidents, large-scale events, traffic control and weather changes, etc., which in turn affect the car demand and the number of available vehicles required for the timetable;

(2) From a macro perspective, reveal the direct determinants of the line uplink generated by the timetable of the current T period and line downlink direct influence factors Uplink car demand in the current T period downlink car demand Number of vehicles available for uplink and the number of vehicles available for downlink And the demand for uplink vehicles in the next T+1 period downlink car demand Number of vehicles available for uplink and the number of vehicles available for downlink

As a further improvement, according to the selected micro and macro factors that affect the generation of bus schedules, a two-layer Bayesian network model for the dynamic generation of bus schedules is constructed, including:

(1) Node abstract definition,

In the Bayesian network model for forecasting dynamic changes in the public transport environment, its conditional nodes are N random factors of the external environment X=(X ₁ , X ₂ ,...,X _N ), including road types, road conditions, traffic accidents , large-scale activities, traffic control and weather changes, etc.; its decision node is the passenger flow fluctuation or travel time change Y={pf _t (X), pt _t (X)} at time t.

In the Bayesian network model for dynamic generation of bus timetables, its conditional nodes are the demand for vehicles in the current T period and the next T+1 period in the up or down direction of the line, as well as the number of available vehicles, that is $Z=\{{\mathrm{pr}}_T^u\left(x\right),{\mathrm{pr}}_T^{\mathrm{D.}}\left(x\right),{\mathrm{PV}}_T^u\left(x\right),{\mathrm{PV}}_T^{\mathrm{D.}}\left(x\right);{\mathrm{pr}}_{T+1}^u\left(x\right),{\mathrm{pr}}_{T+1}^{\mathrm{D.}}\left(x\right),{\mathrm{PV}}_{T+1}^u\left(x\right),{\mathrm{PV}}_{T+1}^{\mathrm{D.}}\left(x\right)\};$ Its decision node is the departure frequency of a line on or off a certain line in the current T period

(2) Structural Learning,

Using the conditional independence test method, all the nodes of the Bayesian network model are dynamically generated for the dynamic change forecast of the bus environment and the bus schedule. If any two nodes are dependent on each other and there are directed edges connected, construct A directed acyclic graph, build their Bayesian network structure graph S.

(3) Parameter learning,

Using the maximum likelihood estimation method, the Bayesian network model is dynamically generated for the dynamic change forecast of the bus environment and the bus schedule, respectively. Given the network topology S and the training sample set D, the prior knowledge is used to determine the respective Bayesian network models. The conditional probability density at each node of the Si network model is:

Describe the probabilistic causal relationship between changes in the external environment and fluctuations in passenger flow or travel time

$\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ;</span>$

Describe the state transition relationship between the random interference of the external environment and the frequency of the uplink and downlink departures of the line

$\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; Z</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Z</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Z</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; .</span>$

As a preference, predict the probability of occurrence of various line capacity and volume under random interference and the reasons for their imbalance, including:

(1) For a certain line, there are K stations and M vehicles in total. Using the actual data of the intelligent bus dispatching platform, combined with the latitude and longitude coordinates of each i vehicle, it is estimated that the number of passengers arriving at station k at time t and the travel time of the vehicle to the first and last station

(2) When a traffic event is detected, determine the values of N external environment random factors X=(X ₁ , X ₂ ,..., X _N ), use the group tree propagation algorithm, and according to the dynamic change model X of the bus environment →Y＝{pf _t (X), pt _t (X)}, predict their fluctuation time pt _t (X) and change passenger flow pf _t (X) and their probability of occurrence;

(3) On the basis of the above, summarize the car demand for uplink and downlink lines in a certain T period ${\mathrm{pr}}_T^u\left(x\right)=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}\left[p_{t-{\mathrm{at}}_t^i-{\mathrm{pt}}_t\left(x\right)}^k{+\mathrm{pf}}_{t-{\mathrm{at}}_t^i-{\mathrm{pt}}_t\left(x\right)}\left(x\right)\right]{,\mathrm{pr}}_T^{\mathrm{D.}}\left(x\right)=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}[p_{t-{\mathrm{at}}_t^i-{\mathrm{pt}}_t\left(x\right)}^k+{\mathrm{pf}}_{t-{\mathrm{at}}_t^i-{\mathrm{pt}}_t\left(x\right)}\left(x\right)],$ number of vehicles available ${\mathrm{PV}}_T^u\left(x\right)=\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}f({\mathrm{at}}_t^i+{\mathrm{pt}}_t\left(x\right)),{\mathrm{PV}}_T^{\mathrm{D.}}\left(x\right)=\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}f({\mathrm{at}}_t^i+{\mathrm{pt}}_t\left(x\right))$ And their probabilities, analyze whether the capacity and volume match, and provide data support for the compilation of bus schedules.

As another priority, combined with scheduling strategies, a possible schedule scheme is generated around the goal of timely evacuation of passengers, including:

(1) Adopt a uniform and balanced departure strategy, consider a single dispatch mode, and determine the departure frequency range on the basis of satisfying the bus capacity constraint c, the degree of congestion γ∈[γ _min , γ _max ], and the government departure interval F and ${\mathrm{trip}}_T^{\mathrm{D.}}\left(x\right)=\max (\frac{{\mathrm{pr}}_T^{\mathrm{D.}}\left(x\right)}{\mathrm{\&gamma;}\&CenterDot;c},f)$ and their probabilities;

(2) According to the imbalance between capacity and volume, use the group tree propagation algorithm and use the timetable to generate a Bayesian network model $Z=\{{\mathrm{pr}}_T^u\left(x\right),{\mathrm{pr}}_T^{\mathrm{D.}}\left(x\right),{\mathrm{PV}}_T^u\left(x\right),{\mathrm{PV}}_T^{\mathrm{D.}}\left(x\right);{\mathrm{pr}}_{T+1}^u\left(x\right),{\mathrm{pr}}_{T+1}^{\mathrm{D.}}\left(x\right),{\mathrm{PV}}_{T+1}^u\left(x\right),{\mathrm{PV}}_{T+1}^{\mathrm{D.}}\left(x\right)\}\&Right\; Arrow;$ $S=\{{\mathrm{trip}}_T^{\mathrm{D.}}\left(x\right),{\mathrm{trip}}_T^u\left(x\right)\},$ Generate possible schedule scenarios.

As a further priority, from the perspective of the government, enterprises and passengers, calculate various indicators for evaluating the quality of timetables and evaluate their pros and cons, including:

(1) On the basis of the timetable scheme, consider the unit mileage operation cost l, and calculate the total passenger waiting time for each scheme $\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}[\frac{T\&Center\; Dot;\underset{T}{\&Integral;}{p}_t^k\mathrm{dt}}{{\mathrm{trip}}_T^u\left(x\right)}+\frac{T\&Center\; Dot;\underset{T}{\&Integral;}{p}_t^k\mathrm{dt}}{{\mathrm{trip}}_T^{\mathrm{D.}}\left(x\right)}],$ Site retention $\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}[{\mathrm{trip}}_T^u\left(x\right)\&CenterDot;\mathrm{\&gamma;}\&Center\; Dot;c-\underset{T}{\&Integral;}{p}_t^k\mathrm{dt}],$ operating cost $\left[\right[{\mathrm{trip}}_T^u\left(x\right)+{\mathrm{trip}}_T^{\mathrm{D.}}\left(x\right)]\&Center\; Dot;l$ and other indicators;

(2) From the perspective of the government, enterprises and passengers, comprehensively evaluate the pros and cons of each of the above timetable schemes, and provide decision support for the public transport management department to choose the best timetable scheme.

The present invention improves by adopting the above-mentioned several measures, uses Bayesian network to describe how changes in the external environment cause changes in passenger flow or travel time, and then affects the process of dynamic adjustment of bus timetables, avoiding the inability of existing methods to solve emergencies The chain reaction process of the unbalanced transport capacity and traffic volume caused by it can mine the coupling relationship between the external environment changes of public transport, the fluctuation of passenger flow or travel time, and the dynamic adjustment of timetable from the original sample data, and the whole process is multi-faceted before, during and after the event Real-time analysis of the reasons and development trends of how traffic events cause changes in bus schedules provides data support for dynamic bus scheduling.

Description of drawings

Fig. 1 is the structural representation that the bus schedule that the present invention relates to generates Bayesian network;

Figure 2 is a flowchart of the implementation of the present invention.

detailed description

Further description will be made below in conjunction with the accompanying drawings provided by the present invention:

As shown in Fig. 1, the present invention provides a kind of dynamic generation method of bus timetable based on Bayesian network model, according to the process of occurrence, development and evolution of traffic events, constructs the two-layer microcosmic and macro Bayeux of dynamic generation of bus timetable This network model analyzes the reasons for the imbalance between transport capacity and traffic volume that affect the dynamic generation of bus schedules, takes external environmental factors as input, and analyzes how they affect passenger flow or travel time changes, and then how to cause the imbalance between bus transport capacity and traffic volume , the output is the result of the maximum matching capacity of the timetable. The control input can control the change of some states. Based on this, the external environmental condition nodes in the emergency Bayesian network are constructed. The four-layer topology network structure of the output of the passenger flow or travel time prediction, the calculation of the timetable, and the output of the decision node of the emergency Bayesian network is realized. The bus schedule and its occurrence probability under complex traffic environment changes provide reliable technical support for bus dynamic scheduling.

As shown in Figure 2, the present invention provides a method for dynamically generating a bus timetable based on a Bayesian network model, including four steps of mechanism analysis, model design, model verification, and model analysis and application. The specific implementation methods are as follows.

Step 1: Mechanism analysis, using a combination of quantitative and qualitative methods to screen many external environmental factors that affect the dynamic generation of bus timetables, and establish a library of factors that affect the failure of bus timetables.

Step 1.1: Microscopically, analyze how N external environmental random factors X=(X ₁ , X ₂ ,...,X _N ) interfere with passenger flow fluctuation pf _t (X) or travel time change pt _t (X ), such as: road types, road conditions, traffic accidents, large-scale events, traffic control and weather changes, etc., which in turn affect the car demand and the number of available vehicles required for the timetable;

Step 1.1.1: Invite experts to discuss and select all n possible influencing factors of passenger flow fluctuations or travel time changes. The former involves seasons, holidays, time periods, large-scale events, traffic control, vehicle failures, weather, etc., and the latter contains road types , traffic flow, traffic congestion, station type, passenger flow and weather, etc.

Step 1.1.2: Combined with the intelligent bus dispatching platform, dynamically obtain the value a _ij of some j external environmental factors at any time i, and their corresponding passenger flow or travel time y _i , and take a total of m data records as sample D, A condition matrix A=(a _ij ) _mn and a decision vector Y=(y _i ) _m are formed.

Step 1.1.3: According to AB=Y, based on the least squares method, calculate B=(b ₁ , b ₂ ,..., b _n )=(A'A) ^-1 (A'Y), for If b _j is greater than the artificial threshold σ, this factor determines passenger flow or travel time, and n influencing factor variables are obtained.

Step 1.2: From a macro perspective, reveal the direct determinants of the uplink line generated by the timetable of the current T period and line downlink direct influence factors Uplink car demand in the current T period downlink car demand Number of vehicles available for uplink and the number of vehicles available for downlink And the demand for uplink vehicles in the next T+1 period downlink car demand Number of vehicles available for uplink and the number of vehicles available for downlink

Step 1.2.1: According to the operating time range of the line, divide it into N time periods, and calculate the car demand in the current T time period and the next T+1 time period in the uplink and downlink directions respectively and the number of vehicles available for them ${\mathrm{PV}}_T^u\left(x\right),{\mathrm{PV}}_T^{\mathrm{D.}}\left(x\right).$

Step 1.2.2: Combined with the intelligent bus dispatching platform, for a certain line, there are K stations and M vehicles in total, combined with the latitude and longitude coordinates of each i vehicle, estimate the number of passengers arriving at station k at time t and the travel time of the vehicle to the first and last station

Step 1.2.3: Summarize the car demand of the line in T period ${\mathrm{pr}}_T^u\left(x\right)=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}\left[p_{t-{\mathrm{at}}_t^i-{\mathrm{pt}}_t\left(x\right)}^k{+\mathrm{pf}}_{t-{\mathrm{at}}_t^i-{\mathrm{pt}}_t\left(x\right)}\left(x\right)\right],$ ${\mathrm{pr}}_T^{\mathrm{D.}}\left(x\right)=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}[p_{t-{\mathrm{at}}_t^i-{\mathrm{pt}}_t\left(x\right)}^k+{\mathrm{pf}}_{t-{\mathrm{at}}_t^i-{\mathrm{pt}}_t\left(x\right)}\left(x\right)],$ number of vehicles available ${\mathrm{PV}}_T^u\left(x\right)=\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}f\left({\mathrm{at}}_t^i{+\mathrm{pt}}_t\left(x\right)\right),$ Analyze whether the capacity and volume match, and provide data support for the compilation of bus schedules.

Step 2: Model design, constructing the upper and lower two-layer microcosmic and macroscopic Bayesian network models dynamically generated by the bus schedule, including the definition of node variables, determining the value range of each condition and decision variable, prior probability distribution, structure learning and parameters Study four parts, in which the upper model describes the random disturbance of the external environment that causes changes in passenger flow or travel time, which is the input condition of the lower model. The lower model describes the random disturbance of the external environment that causes the imbalance of transportation capacity and volume, and provides data support for the generation of timetables; .

Step 2.1: Variable node definition,

In the Bayesian network model for forecasting dynamic changes in the public transport environment, there are n+2 node variables X={X ₁ , X ₂ ,...,X _n }∪Y={pf _t (X), pt _t (X )}, divided into n conditions and 2 decision variable nodes. The former is the input element of the random disturbance of the external environment of the bus, and the latter is the output result of the passenger flow or travel time. It focuses on the interaction between the input elements of the random disturbance of the external environment of the bus and how their changes cause changes in the passenger flow or travel time.

In the Bayesian network model for dynamic generation of bus timetables, its conditional nodes are the demand for vehicles in the current T period and the next T+1 period in the up or down direction of the line, as well as the number of available vehicles, that is $Z=\{{\mathrm{pr}}_T^u\left(x\right),{\mathrm{pr}}_T^{\mathrm{D.}}\left(x\right),{\mathrm{PV}}_T^u\left(x\right),{\mathrm{PV}}_T^{\mathrm{D.}}\left(x\right);{\mathrm{pr}}_{T+1}^u\left(x\right),{\mathrm{pr}}_{T+1}^{\mathrm{D.}}\left(x\right),{\mathrm{PV}}_{T+1}^u\left(x\right),{\mathrm{PV}}_{T+1}^{\mathrm{D.}}\left(x\right)\};\&Right\; Arrow;$ Its decision node is the departure frequency of a line on or off a certain line in the current T period

Step 2.2: Determine the value ranges of the above conditions and decision variable nodes of the microcosmic and macroscopic models and the prior probability distribution between them.

Step 2.2.1: For any node of the micro (macro) model The value range of (S∪Z) Discretize the K feature value state space as $x_i^j=x_{\min }^i+(j-1)*[x_{\max }^i-x_{\min }^i]/K.$

Step 2.2.2: Statistics on any node of the micro (macro) model (S∪Z) value status The probability $p\left(x_i^j\right)=\mathrm{count}\left(x_i^j\right)/\underset{j=1}{\overset{K}{\mathrm{\&Sigma;}}}\mathrm{count}\left(x_i^j\right),$ and two variables (S∪Z) and The prior probability distribution of (S∪Z) between different state values Among them, Count(·) represents the number of events · in the sample set D.

Step 2.3: Construct the topological structure between each node of the micro- and macro-Bayesian network respectively.

Step 2.3.1: Use the K2 algorithm to perform unsupervised machine learning on the training set to obtain the initial network structure of the micro (macro) model.

Step 2.3.2: Using the prior knowledge of experts, based on the conditional independence test method, if any two nodes of the micro (macro) model (S∪Z) and (S∪Z) depend on each other, and there are directed edges to connect Fine-tune the network structure of the micro (macro) model.

Step 2.3.3: Detect whether the adjusted micro (macro) model network structure meets the requirements, if the requirements are met, output the Bayesian network structure diagram S of the bus dynamic environment forecast (timetable generation); otherwise return to step 2.3. 2. Continue the network structure of the micro (macro) model.

Step 2.4: Based on the above network structure, use the maximum likelihood estimation method to estimate the conditional probability distribution table between the nodes in the micro and macro models.

Step 2.4.1: For any node of the micro-macro model (S∪Z), combining the prior distribution and the likelihood function to estimate the parameters

Step 2.4.2: Suppose θ is a random distribution of Dirichlet function, let the likelihood function of θ be $=\underset{k=1}{\overset{r_i}{\mathrm{\&Pi;}}}\underset{i=1}{\overset{\mathrm{no}}{\mathrm{\&Pi;}}}P\left(x_i\right|{\mathrm{\&theta;}}_i)=\underset{i=1}{\overset{\mathrm{no}}{\mathrm{\&Pi;}}}\underset{j=1}{\overset{K}{\mathrm{\&Pi;}}}P\left(x_i^j\right|{\mathrm{\&theta;}}_{\mathrm{ij}}),$ because $L\left({\mathrm{\&theta;}}_{\mathrm{ij}}\right|x)=P\left(x\right|{\mathrm{\&theta;}}_{\mathrm{ij}})=\underset{k=1}{\overset{r_i}{\mathrm{\&Pi;}}}{{\mathrm{\&theta;}}_{\mathrm{ijk}}}^{N_{\mathrm{ijk}}},$ according to $\frac{\&PartialD;\left(L\left(\mathrm{\&theta;}\right|x)\right)}{\&PartialD;\mathrm{\&theta;}}=0,$ Deduced Thus estimating any node of the micro-macro model p(X _i |X _n+1 ) between (S∪Z). Where: _ri is the number of eigenvalues of Xi; N _ijk is the number of _j - _th eigenvalues taken at the parent node when node Xi takes the kth eigenvalue.

Step 2.4.3: According to the derivation process of the above formula, the conditional probability distribution table between the nodes in the micro and macro models can be calculated, that is, the conditional probability between the random disturbance factors in the external environment of the bus and the passenger flow or travel time fluctuations, and then affect the transport capacity and Conditional probability between traffic fluctuations (schedule dynamic adjustment).

Describe the probabilistic causal relationship between changes in the external environment and fluctuations in passenger flow or travel time:

$\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; Y</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ;</span>$

Describe the state transition relationship between the random interference of the external environment and the uplink and downlink departure frequencies of the line:

$\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; Z</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Z</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; S</span>}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; p</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; Z</span>}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; .</span>$

Step 3: Use part of the training set as a test sample to test the accuracy of the micro and macro models. If the model results do not meet the expected goals, return to step 2.

Step 4: Model analysis and application, reasoning and prediction of the dynamic generation process of bus schedules in various complex traffic environments.

Step 4.1: Combined with the intelligent bus dispatching platform, when the external environment of the bus changes, the measured value x=(x ₁ , x ₂ ,..., x _n ) of each influencing factor is dynamically monitored. Calculate its characteristic state (if x _i is in and between), determine the current state of all nodes of the network model

Step 4.2: Using the group tree propagation algorithm, according to the micro model X→Y={pf _t (X), pt _t (X)}, for K stations and M vehicles of a certain line, combine each i vehicle The latitude and longitude coordinates of , estimate the number of passengers arriving at station k at time t and the travel time of the vehicle to the first and last station

Step 4.3: On the basis of the above, summarize the car demand for uplink and downlink lines in a certain T period ${\mathrm{pr}}_T^u\left(x\right)=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}\left[p_{t-{\mathrm{at}}_t^i-{\mathrm{pt}}_t\left(x\right)}^k{+\mathrm{pf}}_{t-{\mathrm{at}}_t^i-{\mathrm{pt}}_t\left(x\right)}\left(x\right)\right]{,\mathrm{pr}}_T^{\mathrm{D.}}\left(x\right)=\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}[p_{t-{\mathrm{at}}_t^i-{\mathrm{pt}}_t\left(x\right)}^k+{\mathrm{pf}}_{t-{\mathrm{at}}_t^i-{\mathrm{pt}}_t\left(x\right)}\left(x\right)],$ number of vehicles available ${\mathrm{PV}}_T^u\left(x\right)=\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}f({\mathrm{at}}_t^i+{\mathrm{pt}}_t\left(x\right)),{\mathrm{PV}}_T^{\mathrm{D.}}\left(x\right)=\underset{i=1}{\overset{m}{\mathrm{\&Sigma;}}}f({\mathrm{at}}_t^i+{\mathrm{pt}}_t\left(x\right))$ And their probability of occurrence, analyze whether the capacity and volume match, and provide data support for analyzing the reasons for the imbalance of the bus schedule.

Step 4.4: Adopt a uniform and balanced departure strategy, consider a single scheduling mode, and preliminarily determine the departure frequency range on the basis of satisfying the bus capacity constraint c, the degree of congestion γ∈[γ _min , γ _max ], and the government departure interval F and And its probability, using the group tree propagation algorithm, according to $Z=\{{\mathrm{pr}}_T^u\left(x\right),{\mathrm{pr}}_T^{\mathrm{D.}}\left(x\right),{\mathrm{PV}}_T^u\left(x\right),{\mathrm{PV}}_T^{\mathrm{D.}}\left(x\right);{\mathrm{pr}}_{T+1}^u\left(x\right),{\mathrm{pr}}_{T+1}^{\mathrm{D.}}\left(x\right),{\mathrm{PV}}_{T+1}^u\left(x\right),{\mathrm{PV}}_{T+1}^{\mathrm{D.}}\left(x\right)\}\&Right\; Arrow;$ $S=\{{\mathrm{trip}}_T^{\mathrm{D.}}\left(x\right),{\mathrm{trip}}_T^u\left(x\right)\},$ Generate possible schedule scenarios.

Step 4.5: Based on the timetable scheme, consider the unit mileage operation cost l, and calculate the total passenger waiting time for each scheme $\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}[\frac{T\&Center\; Dot;\underset{T}{\&Integral;}{p}_t^k\mathrm{dt}}{{\mathrm{trip}}_T^u\left(x\right)}+\frac{T\&Center\; Dot;\underset{T}{\&Integral;}{p}_t^k\mathrm{dt}}{{\mathrm{trip}}_T^{\mathrm{D.}}\left(x\right)}],$ Site retention $\underset{k=1}{\overset{K}{\mathrm{\&Sigma;}}}[{\mathrm{trip}}_T^u\left(x\right)\&Center\; Dot;\mathrm{\&gamma;}\&Center\; Dot;c-\underset{T}{\&Integral;}{p}_t^k\mathrm{dt}],$ operating cost From the perspective of the government, enterprises and passengers, comprehensively evaluate the advantages and disadvantages of each of the above-mentioned timetable schemes, and reversely reason the reasons for possible changes in the timetable failure, so as to provide decision-making for the public transport management department to choose the best timetable scheme support.

What are listed above are only specific embodiments of the present invention. Obviously, the present invention is not limited to the above embodiments, and many variations are also possible, such as: the present invention changes the structure design and parameter learning method of the Bayesian network, and can expand the influencing factors affecting passenger flow or travel time, such as: road type, etc. Use different scheduling strategies and evaluation methods. All deformations that can be directly derived or associated by those skilled in the art from the content disclosed in the present invention should be considered as the protection scope of the present invention.

Claims (6)

1. A method for dynamically generating public transport schedules based on Bayesian network model, characterized in that, comprising the following steps: (1) Using a combination of quantitative and qualitative methods to screen out many external environmental factors that affect the dynamic generation of bus timetables, including traffic accidents that cause changes in travel time, which in turn cause insufficient transport capacity or large-scale events that cause passenger flow fluctuations, which in turn cause insufficient transport capacity; (2) Construct the upper and lower two-layer Bayesian network model for the dynamic generation of bus schedules, in which: the upper model describes the random disturbance of the external environment that causes changes in passenger flow or travel time, and is the input condition of the lower model, which describes the resulting transport capacity and traffic Random interference from the external environment with unbalanced quantity provides data support for schedule generation; (3) When a traffic event occurs, combined with the real-time operation data of the intelligent bus dispatching platform, using the Bayesian network model in step (2) to predict the occurrence probabilities of various lines and the reasons for the imbalance, combined with the dispatching strategy, around With the goal of timely evacuating passengers at the minimum cost, various timetable schemes are generated; (4) Calculate the indicators of passenger waiting time, station retention, and operating costs for each timetable scheme in step (3), and evaluate its pros and cons. 2. a kind of bus schedule dynamic generation method based on Bayesian network model according to claim 1, is characterized in that: the method that quantitatively and qualitatively combine in described step (1), comprises from microcosmic, analysis How do N external environmental random factors X=(X ₁ , X ₂ ,...,X _N ) interfere with passenger flow fluctuation pf _t (X) or travel time change pt _t (X) at time t; from a macro perspective, reveal the current The direct determinants of the line uplink generated by the T-slot timetable and line downlink direct influence factors Uplink car demand in the current T period downlink car demand Number of vehicles available for uplink and the number of vehicles available for downlink And the demand for uplink vehicles in the next T+1 period downlink car demand Number of vehicles available for uplink and the number of vehicles available for downlink 3. the dynamic generation method of bus timetable based on Bayesian network model according to claim 2, is characterized in that: described step (2) influences the microcosmic and macrocosm that step (1) screens out generation of bus timetable Factors, construct the Bayesian network model of dynamic change of bus environment X→Y={pf _t (X), pt _t (X)} and the Bayesian network model of bus timetable generation Describe the probabilistic causal relationship between changes in the external environment and fluctuations in passenger flow or travel time Then describe the state transition relationship between the random interference of the external environment and the vehicle demand of the line and the number of available vehicles Where P represents probability, Y/X represents the probability of Y in the case of X, X/Y represents the probability of X in the case of Y, Z/S represents the probability of Z in the case of S, and S/Z represents the probability of S in the case of Z. 4. the method for dynamically generating public transport timetable based on Bayesian network model according to claim 2, is characterized in that: step (3) utilizes the actual data of intelligent public transport dispatching platform, has K sites and M cars altogether, when When a traffic event is detected, according to the dynamic change model of the bus environment, combined with the number of passenger arrivals of each i vehicle arriving at k station at time t And the travel time of the i-th vehicle to the first and last station Predict their fluctuation time pt _t (X) and changing passenger flow pf _t (X), and then generate a model based on the bus schedule to summarize the car demand for up and down lines in a certain T period number of vehicles available And their probabilities provide data support for the compilation of bus schedules. 5. the bus schedule dynamic generation method based on Bayesian network model according to claim 2, is characterized in that: step (4) is on the basis of the data of step (3), around in time evacuation passenger is target, adopts uniform Balanced departure strategy, considering a single scheduling mode, on the basis of satisfying the capacity constraint c of the bus, the degree of congestion γ∈[γ _min , γ _max ], and the government departure interval F, determine the frequency range of uplink and downlink departures and and their probabilities, generating possible schedule scenarios. 6. the dynamic generation method of bus schedule based on Bayesian network model according to claim 2, is characterized in that: step (5) considers unit mileage operation cost 1 on the schedule scheme basis of step (4), Calculate the total passenger waiting time for each option Site retention operating cost indicators, and analyze whether each timetable scheme is suitable for traffic environment changes, and provide decision support for the public transport management department to choose the best timetable scheme, among which Represents the number of passengers arriving at station k at time t, c represents the capacity constraint of the bus, and γ represents the degree of congestion.