Parameter calibration method for urban road network microscopic traffic simulation system based on macroscopic simulation fusion

Abstract

The application discloses a kind of fusion macroscopic simulation city road network microcosm traffic simulation system parameter calibration method, comprising: building microcosm traffic simulation system and macroscopic traffic simulation system;Initial simulation parameter x is used to execute microcosm traffic simulation system and macroscopic traffic simulation system respectively;Microcosm simulation result and macroscopic simulation result are used to build metamodel m (·), and parameter feasible solution x is estimated using metamodel m (·) h ;Parameter feasible solution x h Microcosm traffic simulation system and macroscopic traffic simulation system are executed respectively, and microcosm simulation error l mic (x h ) is calculated;The microcosm simulation error of two simulation parameters x and x h Corresponding, the simulation parameter x of microcosm traffic simulation system is updated to the simulation parameter corresponding to smaller microcosm simulation error.The application quickly estimates the error gradient of large-scale microcosm simulation result and actual state using metamodel, realizes the rapid search of simulation parameter, constructs metamodel to fit the result of microcosm traffic simulation, and has higher consistency with microcosm simulation result.

Description

Technical Field

[0001] This invention relates to the field of road traffic simulation and simulation optimization, and in particular to a method for calibrating parameters of a micro-traffic simulation system for urban road networks that integrates macro-simulation. Background Technology

[0002] For a range of problems in the field of traffic engineering, ensuring the smooth operation of transportation systems often presents challenges due to the difficulty of organizing on-site experiments and analyses. Traffic simulation, a method that uses computer simulation technology to reproduce the operational state of a traffic system, allows researchers and engineers to use it as a platform for designing and analyzing various traffic problems. Before studying traffic problems using traffic simulation technology, it is necessary to calibrate the key parameters of the simulation in advance to ensure a high degree of consistency between the simulation results and actual traffic conditions. Microscopic traffic simulation, using individual travelers or vehicles as the basic unit, offers even higher accuracy. Therefore, quickly and accurately calibrating the parameters of the microscopic simulation system to ensure consistency between the simulation results and actual conditions has significant research and engineering value.

[0003] Traditional calibration processes often employ random search algorithms, involving several rounds of micro-simulation by adjusting the specific values ​​of the parameters to be calibrated, calculating the error between the simulation results and the actual state, and searching for parameter values ​​that reduce the error. However, micro-simulation generally suffers from high computational complexity, high space consumption, and the long time required to complete a single large-scale urban road network micro-simulation. Therefore, directly performing multiple rounds of simulation to randomly search for optimal simulation parameter values ​​incurs significant time and computational costs and places high demands on hardware performance. Furthermore, as the number of parameters to be calibrated increases, the number of searches required during calibration also increases, further increasing the aforementioned time and computational burden. Summary of the Invention

[0004] Purpose of the invention: To address the above problems, the purpose of this invention is to provide a method for calibrating parameters of a micro-traffic simulation system for urban road networks that integrates macro-simulation. This method utilizes prior information provided by macro-simulation results, which have lower accuracy but lower computational cost, to assist in the calibration of micro-simulation parameters. On the one hand, it achieves efficient searching, and on the other hand, it minimizes the number of searches by maximizing speed limits, thereby improving the efficiency and quality of micro-simulation parameter calibration.

[0005] Technical solution: The present invention provides a parameter calibration method for a micro-traffic simulation system of urban road networks that integrates macro-simulation, comprising the following steps:

[0006] Step 1: Construct the actual state dataset y of the traffic network as the basis for calibration. real ;

[0007] Step 2: Build a micro-level traffic simulation system based on travelers or vehicles, and a macro-level traffic simulation system based on path flow.

[0008] Step 3: Using the initial simulation parameters x, execute the microscopic traffic simulation system and the macroscopic traffic simulation system respectively to obtain the microscopic simulation results and the macroscopic simulation results, and calculate the microscopic simulation error l. mic (x);

[0009] Step 4: Construct a meta-model m(·) using the microscopic and macroscopic simulation results, and estimate the feasible solution x using the meta-model m(·). h ; Using the feasible solution x with parameters h Execute both the microscopic and macroscopic traffic simulation systems separately, update the microscopic and macroscopic simulation results, and calculate the microscopic simulation error l. mic (x h ), where h represents the number of iterations;

[0010] Step 5, compare the two simulation parameters x and x h The corresponding microscopic simulation error is used to update the simulation parameters x of the microscopic traffic simulation system to the simulation parameters corresponding to the smaller microscopic simulation error; it is then determined whether optimization is complete, and if so, the updated simulation parameters x are output.

[0011] Furthermore, step 5 also includes: if optimization is not completed, proceed to step 4 to start a new round of iteration.

[0012] Furthermore, step 3 also includes:

[0013] After executing the microscopic traffic simulation system with initial simulation parameters x, the microscopic simulation result network simulation state y is obtained. mic (x), using network simulation state y mic (x) and the actual state dataset y real Calculate the microscopic simulation error l mic (x);

[0014] After executing the macroscopic traffic simulation system with initial simulation parameters x, M macroscopic simulations are performed in parallel with different randomization seeds to obtain M sets of corresponding macroscopic simulation results.

[0015] Furthermore, the actual state dataset y mentioned in step 1 real This includes the hourly outflow t of each segment i of key arterial roads in the road network. γ of motorized bus occupancy rate within the road network real (x); Traffic data is written in vector form as Motorized bus occupancy rate is defined as the proportion of passengers who choose conventional buses or subways within the road network to the total number of motorized passengers.

[0016] Furthermore, the microscopic simulation error l mentioned in step 3mic (x) is from and The weighted average of the two error components is calculated as follows:

[0017]

[0018] In the formula, θ represents the weighting coefficient. The error between the simulated flow rate and the actual flow rate is expressed as:

[0019]

[0020] The error between the micro-level bus occupancy rate and the actual occupancy rate is expressed as:

[0021]

[0022] In the formula, γ represents the hourly outflow rate t of each section i of the key arterial road. mic (x) represents the occupancy rate of motorized buses within the road network.

[0023] Furthermore, the meta-model is defined as follows: Functions, including meta-models With meta-model The two parts are used to fit the hourly outflow t of each segment i of the key arterial road in the microscopic traffic simulation results. γ of motorized bus occupancy rate within the road network real (x), the expressions are as follows:

[0024]

[0025]

[0026] In the formula, β q β γ Metamodel With meta-model The parameters are all (M+1)×1 dimensional vectors; For constant terms; M sets of results for macroscopic simulation Perform linear combination coefficients; This represents the hourly outflow t of each segment i of the key arterial road obtained from the j-th macroscopic simulation. This represents the occupancy rate of motorized buses within the road network obtained from the j-th macroscopic simulation.

[0027] Furthermore, metamodel With meta-model The parameter β in qand β γ Based on least squares linear regression, the expressions are as follows:

[0028]

[0029]

[0030] Furthermore, the meta-model m(·) is used to quickly estimate the microscopic simulation error l. mic The gradient g of (x) at the simulation parameter x m (x), the expression is:

[0031]

[0032] In the formula, δ represents a randomly generated small perturbation; Let m(·) represent the microscopic simulation errors estimated rapidly based on the meta-model m(·) with simulation parameters x+δ and x-δ, respectively, and their expressions are:

[0033]

[0034]

[0035] In the formula, 0≤θ≤1 are weighting coefficients, representing the weight of the errors of the two indicators, hourly traffic flow and bus occupancy rate, in the total simulation error.

[0036] Furthermore, in step 5, the criterion for determining whether optimization is complete is whether the current iteration number h exceeds the preset iteration number H.

[0037] Beneficial effects: Compared with the prior art, the significant advantages of this invention are:

[0038] 1. This invention utilizes a meta-model to quickly estimate the error gradient between large-scale micro-simulation results and actual conditions, enabling rapid search for simulation parameters. Compared to directly adjusting micro-simulation parameters to calculate errors and find the optimal parameter combination, the meta-model constructed in this invention fits the results of micro-traffic simulation, resulting in faster calculation speed and higher consistency between the results and micro-simulation results. Therefore, based on the meta-model, the error gradient between large-scale micro-simulation results and actual conditions can be quickly estimated, and potential better parameters can be quickly searched.

[0039] 2. This invention achieves efficient calibration of simulation parameters by fully utilizing the prior information provided by macroscopic traffic simulation results; in order to enrich the prior information and execute multiple independent macroscopic traffic simulations in parallel, a good calibration result can be obtained with only a small number of search rounds.

[0040] 3. This invention reduces the time and computational cost in the process of calibrating micro-simulation parameters for large-scale urban traffic, quickly completes parameter calibration, improves search efficiency, and finds simulation parameter values ​​that make the simulation results as close as possible to the actual results through micro-simulation with as few rounds as possible. Attached Figure Description

[0041] Figure 1 This is a flowchart of the parameter calibration process for the microscopic traffic simulation system of the present invention;

[0042] Figure 2 This is a schematic diagram of a large-scale road network in the embodiment;

[0043] Figure 3 This is a schematic diagram of the road segment for collecting hourly traffic flow in the embodiment. Detailed Implementation

[0044] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments.

[0045] This embodiment describes a parameter calibration method for a micro-traffic simulation system of urban road networks that integrates macro-simulation. The flowchart is as follows: Figure 1 As shown, the urban road network is illustrated using the large-scale road network of Qingdao's central urban area as an example, including the following steps:

[0046] Step 1: Construct the actual state dataset y of the traffic network as the basis for calibration. real ;

[0047] Step 2: Build a micro-level traffic simulation system based on travelers or vehicles, and a macro-level traffic simulation system based on path flow.

[0048] Step 3: Using the initial simulation parameters x, execute the microscopic traffic simulation system and the macroscopic traffic simulation system respectively to obtain the microscopic simulation results and the macroscopic simulation results, and calculate the microscopic simulation error l. mic (x);

[0049] Step 4: Construct a meta-model m(·) using the microscopic and macroscopic simulation results, and estimate the feasible solution x using the meta-model m(·). h ; Using the feasible solution x with parameters h Execute both the microscopic and macroscopic traffic simulation systems separately, update the microscopic and macroscopic simulation results, and calculate the microscopic simulation error l. mic (x h ), where h represents the number of iterations;

[0050] Step 5, compare the two simulation parameters x and x hThe corresponding microscopic simulation error is used to update the simulation parameters x of the microscopic traffic simulation system to the simulation parameters corresponding to the smaller microscopic simulation error; it is then determined whether the optimization is complete. If the optimization is complete, the updated simulation parameters x are output; if the optimization is not complete, the process proceeds to step 4 to start a new round of iteration.

[0051] The actual state dataset y in step 1 above real This includes the hourly outflow t of each segment i of key arterial roads in the road network. γ of motorized bus occupancy rate within the road network real (x); Traffic data is written in vector form as The data in the above actual status dataset can be collected through video surveillance or induction loops deployed at major road intersections. The main road sections where hourly traffic flow is monitored include... Figure 3 As shown. Motorized public transport occupancy rate is defined as the proportion of passengers choosing regular public transport or subway within the road network to the total number of motorized trips (cars + regular buses + subways). The relevant data is obtained from resident travel survey results. In this embodiment, the total number of motorized trips includes passengers traveling by car, bus, and subway, which can be obtained from resident travel survey results.

[0052] The large-scale microscopic traffic simulation system built in step 2 above includes a ground road network design section, a public transport network design section, and a traffic demand loading section. It supports simulation of car and public transport travel. In this embodiment, the microscopic simulation software MATSim is used as the microscopic traffic simulation system. Figure 2 The large-scale road network in the central urban area of ​​Qingdao is shown in the figure for micro-simulation. The software MATSim models the road network traffic state on a per-traveler basis and allows each traveler to autonomously choose their departure time, mode of transport, and route based on the network status. The simulation has high computational complexity and large space occupancy, but it better reflects the actual state of the road network. At the same time, a macro-level traffic simulation system is built for the same road network and traffic demand, which also supports car and public transport travel. In this embodiment, the traditional Logit-DTA model is used as the macro-level traffic simulation system. The road network traffic state is modeled on a path flow basis, and traffic assignment is based on the traditional Logit algorithm. The computational complexity is low and the space occupancy is small, but the overall error is relatively large.

[0053] In step 3 above, the simulation parameters in the microscopic traffic simulation system to be calibrated are denoted as x, where x is a one-dimensional numerical vector. The length of the vector is the number of parameters to be calibrated, and the value of each element is the value of the corresponding parameter. In this embodiment, the microscopic simulation software MATSim uses parameters to be calibrated, including traveler behavior selection parameters such as the marginal utility of car travel time and the marginal utility of public transport travel time. These parameters are also the simulation parameters of the selected macroscopic simulation model Logit-DTA.

[0054] Given initial simulation parameters x, after executing the microscopic traffic simulation system, the microscopic simulation result network simulation state y is obtained. mic (x), using network simulation state y mic (x) and the actual state dataset y real Calculate the microscopic simulation error l mic (x). Microscopic simulation results: Network simulation state y mic (x) includes the hourly outflow t of each section i, i = 1, ..., N of the key main roads. γ of motorized bus occupancy rate within the road network mic (x).

[0055] The microscopic simulation error l in step 3 above mic (x) is from and The result is obtained by weighted calculation of the two parts, and the calculation expression is:

[0056]

[0057] In the formula, θ represents the weighting coefficient, and in this embodiment, θ can be set to 0.6; The error between the simulated flow rate and the actual flow rate is expressed as:

[0058]

[0059] The error between the micro-level bus occupancy rate and the actual occupancy rate is expressed as:

[0060]

[0061] In the formula, γ represents the hourly outflow rate t of each section i of the key arterial road. mic (x) represents the occupancy rate of motorized buses within the road network.

[0062] Step 3 above also includes: after executing the macroscopic traffic simulation system with the initial simulation parameters x, performing M macroscopic simulations in parallel with different randomization seeds to obtain the corresponding M sets of macroscopic simulation results. In this embodiment, M = 5. Macroscopic simulation results. Specifically, this includes the hourly outflow volume of each section i on key arterial roads. The occupancy rate of motorized buses within the road network

[0063] Based on the macroscopic simulation results obtained above Microscopic simulation results y under corresponding parameters mic(x) Construct a meta-model m(·) and estimate a new feasible solution x′ with reduced error based on the meta-model m(·).

[0064] Furthermore, the meta-model is defined as follows: Functions, including meta-models With meta-model The two parts are used to fit the hourly outflow t of each segment i of the key arterial road in the microscopic traffic simulation results. γ of motorized bus occupancy rate within the road network real (x), the expressions are as follows:

[0065]

[0066]

[0067] In the formula, β q β γ Metamodel With meta-model The parameters are all (M+1)×1 dimensional vectors; For constant terms; M sets of results for macroscopic simulation Perform linear combination coefficients; This represents the hourly outflow t of each segment i of the key arterial road obtained from the j-th macroscopic simulation. This represents the occupancy rate of motorized buses within the road network obtained from the j-th macroscopic simulation.

[0068] Metamodel With meta-model The parameter β in q and β γ Based on least squares linear regression, the expressions are as follows:

[0069]

[0070]

[0071] Rapid estimation of microscopic simulation error l using meta-model m(·) mic The gradient g of (x) at the simulation parameter x m (x), the expression is:

[0072]

[0073] In the formula, δ represents a randomly generated small perturbation; Let m(·) represent the microscopic simulation errors estimated rapidly based on the meta-model m(·) with simulation parameters x+δ and x-δ, respectively, and their expressions are:

[0074]

[0075] In the formula, 0≤θ≤1 are weighting coefficients, representing the weight of the errors of the two indicators, hourly traffic flow and bus occupancy rate, in the total simulation error.

[0076] In step 5, the criterion for determining whether optimization is complete is whether the current iteration number h exceeds the preset iteration number H. If the preset iteration number is reached, the optimization ends and the parameters x of the microscopic traffic simulation system are output. Otherwise, the process continues to step 4 for a new round of optimization.

[0077] This method rapidly estimates the error gradient between large-scale microscopic simulation results and actual conditions based on a meta-model, enabling rapid search of simulation parameters in microscopic traffic simulation systems. At the same time, it fully utilizes the prior information provided by macroscopic traffic simulation results to achieve efficient calibration of simulation parameters.

Claims

1. A parameter calibration method for a micro-traffic simulation system of urban road networks integrating macro-simulation, characterized in that, Includes the following steps: Step 1: Construct the actual state dataset of the traffic network as the basis for calibration. ; Step 2: Build a micro-level traffic simulation system based on travelers or vehicles, and a macro-level traffic simulation system based on path flow. Step 3, using the initial simulation parameters Execute the microscopic traffic simulation system and the macroscopic traffic simulation system separately to obtain the microscopic simulation results and the macroscopic simulation results, and calculate the microscopic simulation error. ; Step 4: Build a meta-model using microscopic and macroscopic simulation results. Using meta-model Estimated parameter feasible solution ; Using the parameter feasible solution Run both the microscopic and macroscopic traffic simulation systems separately, update the microscopic and macroscopic simulation results, and calculate the microscopic simulation error. h represents the number of iterations; Step 5: Compare the two simulation parameters The corresponding microscopic simulation error will affect the simulation parameters of the microscopic traffic simulation system. Update the simulation parameters to correspond to the smaller microscopic simulation error; determine if optimization is complete; if so, output the updated simulation parameters. ; The actual state dataset mentioned in step 1 Including key arterial roads in the road network Hourly outflow The occupancy rate of motorized buses within the road network Traffic data is written in vector form as Motorized bus occupancy rate is defined as the proportion of passengers who choose conventional buses or subways within the road network to the total number of motorized passengers. The microscopic simulation error mentioned in step 3 Depend on and The weighted average of the two error components is calculated as follows: ； In the formula, Indicates the weighting coefficient. The error between the simulated flow rate and the actual flow rate is expressed as: ； The error between the micro-level bus occupancy rate and the actual occupancy rate is expressed as: ； In the formula, Indicates key main road sections hours Outflow traffic This indicates the occupancy rate of motorized buses within the road network.

2. The parameter calibration method according to claim 1, characterized in that, Step 5 also includes: if optimization is not completed, proceed to step 4 to start a new round of iteration.

3. The parameter calibration method according to claim 1, characterized in that, Step 3 also includes: In the initial simulation parameters After executing the microscopic traffic simulation system, the microscopic simulation results and network simulation state are obtained. Using network simulation state and actual state dataset Calculate microscopic simulation error ; In the initial simulation parameters After executing the macroscopic traffic simulation system, it is run in parallel with different randomization seeds. The macroscopic simulation yields M sets of corresponding macroscopic simulation results. .

4. The parameter calibration method according to claim 1, characterized in that, The metamodel is defined as follows: Functions, including meta-models With meta-model The two error components are used to fit the key arterial road segments in the microscopic traffic simulation results. Hourly outflow The occupancy rate of motorized buses within the road network The expressions are as follows: ； ； In the formula, Metamodel With meta-model The parameters are all dimensional vector; For constant terms; For macroscopic simulation Group Results Perform linear combination coefficients; Indicates the first Key main road sections obtained from the macroscopic simulation hours Outflow traffic Indicates the first The occupancy rate of motorized buses within the road network was obtained from a macroscopic simulation.

5. The parameter calibration method according to claim 4, characterized in that, Metamodel With meta-model Parameters in Based on least squares linear regression, the expressions are as follows: 。 6. The parameter calibration method according to claim 5, characterized in that, Using meta-model Rapid estimation of microscopic simulation errors Based on simulation parameters gradient at The expression is: ； In the formula, This represents a randomly generated small perturbation; They represent the meta-models respectively. Rapid estimation based on simulation parameters and The microscopic simulation error at that time is expressed as: ； In the formula The weighting coefficients represent the weights of the errors in the two indicators, hourly traffic flow and bus occupancy rate, in the total simulation error.

7. The parameter calibration method according to claim 1, characterized in that, In step 5, the criterion for determining whether optimization is complete is the current iteration number. Has the preset number of iterations been exceeded? .

Images