Multi-agent simulation analysis method for operation optimization of urban parking system

Abstract

The invention provides a multi-agent simulation analysis method for operation optimization of an urban parking system, and the method comprises the following steps: S1, obtaining influence factors which remarkably influence the parking selection behavior of a driver through questionnaire design and experiment based on will survey; s2, through a Binomial Logit discrete selection model, carrying out quantitative analysis on the influence factors of the parking selection behavior of the driver; s3, building a multi-agent parking lot microscopic simulation platform in combination with a Netlogo development environment; and S4, constructing a double-layer model for parking lot operation management optimization, and solving by adopting a simulated annealing algorithm. On the basis of analyzing the parking selection behavior of the driver, the operation efficiency of the parking lot is effectively improved through multi-agent simulation, and a technical tool is provided for operation organization and management strategy evaluation of the parking lot.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent parking system management and control, and in particular relates to a multi-agent simulation analysis method for optimizing the operation of urban parking systems. Background Art

[0002] With the acceleration of urbanization and the continuous growth of car ownership, the operational management and traffic organization design of large parking lots face increasingly complex challenges. However, the current theoretical framework for parking lot internal operations management and traffic organization is still incomplete. Existing research mainly focuses on the relationship between parking supply and demand and overall planning, lacking in-depth research on the parking behavior patterns of drivers within parking lots. This research gap limits the optimization potential of parking lot design, management and evaluation, leading to persistent problems such as inefficient parking resource utilization, poor user experience, and traffic congestion.

[0003] Traditional approaches to parking lot design and optimization typically rely on macro-statistics and empirical rules, but these methods overlook the diverse and complex nature of individual drivers' behaviors. For example, static models cannot accurately reflect drivers' dynamic decision-making processes in real-world scenarios, while analyses based on a single factor struggle to fully capture the diversity of parking choice behaviors. Furthermore, there is a lack of efficient technical tools to simulate and evaluate the specific impact of different management strategies on parking lot operations, resulting in deficiencies in both scientific and practical aspects of parking lot management and planning.

[0004] In this context, studying the patterns of parking choice behavior in parking lots is crucial. Analyzing the key factors influencing driver decisions can provide a more precise basis for parking lot planning and design. Furthermore, quantitative analysis using discrete choice models can clarify the degree to which different factors influence driver parking choices, which plays a crucial role in optimizing parking resource allocation and improving overall parking lot efficiency. This research, combined with multi-agent simulation technology, can recreate parking lot operations at a microscopic level, providing technical support and a theoretical basis for developing scientific parking lot management strategies and optimizing traffic flow. Summary of the Invention

[0005] The purpose of the present invention is to overcome the defects of the above-mentioned traditional methods and provide a multi-agent simulation analysis method for optimizing the operation of urban parking systems.

[0006] The purpose of the present invention can be achieved by the following technical solutions:

[0007] A multi-agent simulation analysis method for optimizing urban parking system operations, the method comprising the following steps:

[0008] S1. Through questionnaire design and experiments based on willingness survey, we can identify the factors that significantly influence drivers' parking behavior.

[0009] S2. Quantitatively analyze the factors influencing drivers’ parking choice behavior using the Binomial Logit discrete choice model;

[0010] S3. Build a multi-agent parking lot micro-simulation platform using the Netlogo development environment;

[0011] S4. Construct a two-layer model for parking lot operation management optimization, use simulated annealing algorithm to solve it, and optimize the parking lot operation efficiency.

[0012] Furthermore, the parking selection behavior is mainly divided into three stages, namely the target area selection stage, the stage of searching for empty spaces in the area, and the stage of evaluating parking spaces and parking; the influencing factors are divided into personal attribute factors and parking attribute influencing factors. Personal attribute factors include gender, age, driving experience, educational background, and income. Parking attribute influencing factors include walking distance after parking, parking feasibility, driving distance, parking safety, parking convenience, and parking cost.

[0013] Furthermore, the S1 is specifically:

[0014] S101. Determine the research question and purpose of the investigation;

[0015] S102. Determine the influencing factors, eliminate factors with insignificant influence, and determine the level number and level value of relevant parameters based on the actual situation, survey purpose, and survey principles;

[0016] S103, scenario design based on utility balance;

[0017] S104. Complete the willingness survey questionnaire and conduct on-site questionnaire survey;

[0018] S105. Statistically analyze the results of the parking choice behavior intention survey to obtain factors that significantly affect the drivers' parking choice behavior.

[0019] Furthermore, the statistical analysis of the results can be divided into statistical analysis of driver personal information, statistical analysis of driver parking behavior, and statistical analysis of parking space selection tendency.

[0020] Furthermore, the S2 is specifically:

[0021] S201, using SPSS software to calibrate parameters of a discrete BL model of the driver's parking choice behavior in a parking lot;

[0022] S202, bringing the driver's personal attributes and parking space attribute variables into the model, and performing model fit analysis using SPSS software;

[0023] S203, using χ 2 Independence test was used to analyze the correlation between influencing factors and parking choice behavior.

[0024] Furthermore, if the results of the model fit analysis show that the model fit index value does not meet the required value, it is necessary to adjust the variables of the model accordingly, that is, remove the independent variables that are not significant in the model test, re-establish the model analysis, and repeatedly adjust and calibrate the model until the various fit indices of the model meet the requirements.

[0025] Furthermore, the simulation logic of the simulation platform includes a random selection method and a method based on a discrete selection model to maximize personal utility, and the relevant parameters of the simulation model are calibrated and the simulation effect is verified through measured data.

[0026] Furthermore, the simulation platform constructed by Netlogo can be divided into two types. One describes the random selection of parking strips and parking spaces by drivers in the absence of parking guidance information and obeys a certain probability distribution. The other describes the selection of parking strips and parking spaces by drivers in the presence of parking guidance information according to the discrete choice model calibrated above.

[0027] Furthermore, the S4 is specifically:

[0028] S401. Set up the upper-level model so that its goal is to optimize social benefits;

[0029] S402, introducing the evaluation model based on microscopic traffic simulation into the lower model;

[0030] S403. Solve the double-layer model using a simulated annealing algorithm to optimize parking lot operation efficiency.

[0031] Compared with the prior art, the present invention has the following beneficial effects:

[0032] By studying the parking behavior patterns of drivers in parking lots, this invention effectively reflects the dynamic decision-making process of individual drivers, while quantifying the specific impact of different factors on parking selection behavior. It also constructs a multi-agent simulation platform through a development environment to dynamically simulate the parking lot operation process at a micro level. Combined with a two-layer optimization model and a simulated annealing algorithm, it effectively improves the operating efficiency of parking lots and the scientific nature of resource allocation, providing innovative solutions and technical support for the planning, design, operation management and strategy evaluation of parking lots. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1A schematic diagram of the form and content of the parking choice behavior intention questionnaire of the present invention;

[0034] Figure 2 It is a flowchart of the parameter estimation and testing flow chart of the binomial logit model of the present invention;

[0035] Figure 3 This is the flow chart of the random simulation model;

[0036] Figure 4 This is a flow chart of the parking lot micro-simulation platform based on parking behavior rules of the present invention;

[0037] Figure 5 A schematic diagram of the parking space layout and flow lines on the B1 floor of the underground parking lot of the present invention;

[0038] Figure 6 Schematic diagram of the comparison between the actual measurement and simulation of the average time occupancy rate of parking spaces of the present invention;

[0039] Figure 7 Schematic diagram showing the comparison between the actual measurement and simulation of the average daily turnover rate of parking spaces according to the present invention;

[0040] Figure 8 A schematic diagram of the lower layer simulation model logic of the present invention;

[0041] Figure 9 A flow chart of the model optimization combining the dual-layer optimization model and the simulated annealing algorithm of the present invention;

[0042] Figure 10 A schematic diagram of a parking lot network structure according to the present invention;

[0043] Figure 11 This is a flow chart of the multi-agent simulation analysis method for urban parking system operation optimization of the present invention. DETAILED DESCRIPTION

[0044] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments. This embodiment is implemented based on the technical solution of the present invention, and provides a detailed implementation method and specific operation process, but the protection scope of the present invention is not limited to the following embodiments.

[0045] Example

[0046] like Figure 11 As shown, a multi-agent simulation analysis method for optimizing the operation of an urban parking system is provided, wherein the method comprises the following steps:

[0047] S1. Through questionnaire design and experiments based on willingness survey, the factors that significantly affect drivers' parking choice behavior are obtained.

[0048] Parking selection behavior is mainly divided into three stages: target area selection stage, searching for empty spaces within the area stage, and evaluating parking spaces and parking stage;

[0049] The influencing factors are divided into personal attribute factors and parking space attribute factors;

[0050] Personal attribute factors include gender, age, driving experience, educational background, and income;

[0051] Factors influencing parking space attributes include driving distance to the parking lot, walking time from the parking space to the destination, parking feasibility, parking safety, parking convenience, and parking cost.

[0052] S101. The purpose of the survey is to understand the decision-making process of drivers' parking choices, identify the characteristics that significantly influence their parking choices, and provide a reasonable basis for parking lot operation and management.

[0053] S102. Determine the influencing factors, eliminate factors with insignificant impact, and determine the level number and level value of relevant parameters based on the actual situation, survey purpose, and survey principles, as shown in Table 1:

[0054] Table 1 Related parameter level settings

[0055]

[0056] That is, among the factors influencing personal attributes, remove the variable education background and retain gender, age, driving experience, and income.

[0057] Among the factors affecting parking space attributes, the driving distance to the parking lot, the walking time from the parking space to the destination, and the parking feasibility, i.e., the probability of the existence of available parking spaces in the target parking block, are retained.

[0058] S103. Based on utility equilibrium, scenario design was conducted using the design procedure in Ngene with the objective function of minimizing D-error. Eight survey scenarios were obtained. Considering the actual survey difficulty and respondents' acceptance, the eight scenarios were split into two parts. Each questionnaire only contained four survey scenarios to reduce the difficulty of the questionnaire.

[0059] S104. Complete willingness questionnaire was formed and field survey was carried out. A total of 536 questionnaires were collected, and 367 questionnaires were effectively collected after screening. Figure 1 Select the form and content of the behavioral intention questionnaire for parking;

[0060] S105. Conduct statistical analysis on the results of the parking choice behavior intention survey, which can be divided into statistical analysis of driver personal information, statistical analysis of driver parking behavior, and statistical analysis of parking space selection tendency, so as to obtain influencing factors that significantly affect the driver's parking choice behavior.

[0061] (1) Statistical analysis of driver’s personal information

[0062] A univariate statistical analysis was conducted on the basic personal information of the drivers in the questionnaire, including gender, age, driving experience, education level, monthly income, etc. The statistical indicators were descriptive indicators such as frequency, effective percentage, and cumulative percentage, as shown in Table 2:

[0063] Table 2 Summary of statistical results of drivers' personal information

[0064]

[0065] Basic information on the respondents revealed a relatively even gender distribution (with a higher number of male drivers) with a male to female ratio of nearly 3:1. The majority were young or middle-aged, with 78.7% under the age of 39. 54.2% had three or more years of driving experience, while 16.6% had 10 or more years. 85% of the respondents had a bachelor's degree or higher. Monthly incomes were generally high, with 69.2% earning over 5,000 yuan per month and 28.3% over 10,000 yuan. Overall, this group represents a relatively high income, with a preponderance of male drivers and extensive driving experience.

[0066] (2) Statistical analysis of driver parking behavior

[0067] The parking behaviors that parkers have already taken place or used to take in the past are an important part of the basic information survey, the purpose of which is to understand the drivers' choice habits and preferences during parking. The statistical results are shown in Table 3:

[0068] Table 3 Summary of statistical results of drivers' parking behavior

[0069]

[0070]

[0071] Through the above parking behavior statistics, we found that in the survey results, most drivers would choose parking lots (garages) to park their vehicles. Among them, the proportions of those who preferred and generally chose parking lots (garages) were 24% and 57.5% respectively, with a cumulative percentage of 81.5%. The proportion of those who generally did not choose parking lots was 18%. The results show that parking lots (garages) are the main parking choice for drivers, and it is very necessary to analyze parking behavior in parking lots in order to optimize their management model.

[0072] When parking lots are open, drivers generally don't park immediately. 33.8% of drivers search for a satisfactory spot before parking, while 40% drive directly to the target area to find a satisfactory spot. During peak hours, when parking lots are crowded and occupancy is high, most drivers park immediately. 61.9% of drivers park immediately, while 33.9% search for a satisfactory spot before parking. Only 4.1% drive directly to a desired area and then search for a suitable spot.

[0073] S2. Quantitatively analyze the factors influencing drivers' parking choices using a Binomial Logit discrete choice model. Based on the quantitative analysis results, select several of these factors as the final independent variables of the Binomial Logit discrete choice model and obtain the corresponding independent variable parameters.

[0074] S201, using SPSS software to calibrate the parameters of the discrete BL model (BinomialLogit discrete choice model) of the driver's parking choice behavior in the parking lot. The basic steps for calibrating the parameters of the Binomial Logit model are as follows: Figure 2 As shown;

[0075] S202. The driver's personal attributes and the three parking space attribute variables in Table 1 are introduced into the calibrated discrete BL model, and the model fit analysis is performed using SPSS software.

[0076] Furthermore, if the results of the model fit analysis show that the model fit index value does not meet the required value, it is necessary to adjust the variables of the model accordingly, that is, remove the independent variables that are not significant in the model test, re-establish the model analysis, and repeatedly adjust and calibrate the model until the various fit indices of the model meet the requirements.

[0077] S203, using χ 2 Independence test is used to analyze the correlation between influencing factors and parking choice behavior, and the final independent variables and corresponding independent variable parameters of the discrete BL model are obtained.

[0078] All the independent variable data and dependent variable data (whether to choose the berth) collected by the questionnaire were input into SPSS software, and the SPSS software was used to perform χ 2 Independence test. When testing the independence of variables, the significance level is generally set at 0.05. If the P value (the probability value corresponding to the value of the test statistic, P-value) calculated by SPSS software is less than or equal to 0.05, it is considered that there is a significant relationship between the independent variable and the dependent variable.

[0079] Table 4 Chi-square test between independent variables and dependent variables

[0080]

[0081] The “on-site driving distance” refers to the “driving distance into the parking lot” in Table 1.

[0082] “Walking time” refers to the “Walking time from berth to destination” in Table 1.

[0083] “Probability of existence of valid vacant parking spaces” refers to the “Probability of existence of valid vacant parking spaces in the target parking area” in Table 1.

[0084] The above table summarizes the χ2 relationship between each variable and the dependent variable. 2 Independence test results.

[0085] Among them, the chi-square test value Asymp.sig (2-sided) of the driving distance, walking time and the probability of the existence of valid vacant parking spaces is less than or equal to 0.05, indicating that these variables have a certain impact on travel choices and should be included when establishing the choice behavior model.

[0086] However, the personal attribute characteristic variables were not significant and were not included in the independent variables in this study. That is, the final parameters of the BL model only used the driving distance within the venue, walking time, and the probability of the existence of an effective vacant parking space.

[0087] Here, "choice behavior model" refers to the step Binomial Logit discrete choice model.

[0088] The final parameter estimation results of the model are shown in Tables 5 and 6.

[0089] Table 5 Results of BL model for drivers’ on-site parking preference

[0090]

[0091] Table 6 Overall statistical indicators of the model

[0092]

[0093] As can be seen from the table above, the model's various fitting indices and the P-value tests of each parameter show that the model's accuracy has met the response requirements. 2 is a pseudo-determination coefficient based on the likelihood function, Nagelkerke R 2 It is Cox&Snell R 2 The adjusted versions of the models can reflect the model's ability to explain the variation of the dependent variable. Cox & Snell R 2 is 0.278 and Nagelkerke R2 The value of 0.337 indicates that the model has a certain explanatory power for the dependent variable. McFadden R 2 It is another pseudo coefficient of determination that assesses the degree of model improvement based on the likelihood ratio test by comparing the log-likelihood of the fitted model with the log-likelihood of the null model containing only the constant term. Typically, McFadden R 2 Values ​​between 0.2 and 0.4 are considered to indicate a good fit of the model. 2 The value is 0.241, indicating that the model has a good fit.

[0094] From the estimated values ​​of various parameters we can get:

[0095] The regression coefficient (estimated value) of the independent variable, on-site driving distance, is negative, indicating that the longer the on-site driving distance, the less likely the parking space will be chosen.

[0096] The regression coefficient of the independent variable walking time is negative, indicating that the longer the walking time, the less likely the parking space will be chosen;

[0097] The regression coefficient of the independent variable probability of available parking spaces is positive, indicating that drivers are more inclined to park in open areas with available parking spaces.

[0098] From the comparison of the regression coefficient values ​​of the independent variables, it can be seen that among the factors affecting drivers' preference for parking spaces in parking lots, the probability of available parking spaces has the greatest impact (0.0764), followed by the walking time to the destination (-0.0524), and the weakest influence is the driving distance within the parking lot (-0.0108).

[0099] This shows that when drivers choose a parking space in a parking lot, they first consider the probability of vacant parking spaces in the area, which can be obtained based on the parking guidance information in the parking lot.

[0100] From the above, the inputs of Binomial Logit discrete choice model include: driving distance, walking time and the probability of having an available parking space;

[0101] Its output is: the probability of choosing a parking space.

[0102] A parking strip can be thought of as a large parking space.

[0103] The BL model calibrates a utility function, which calculates the utility of all optional berths and the probability of a berth being selected based on the utility of each berth.

[0104] S3. Build a multi-agent parking lot micro-simulation platform using the Netlogo development environment;

[0105] The simulation logic of the simulation platform includes a random selection method and a method based on a discrete selection model to maximize personal utility, and the relevant parameters of the simulation model are calibrated and the simulation effect is verified through measured data.

[0106] The simulation platform constructed by Netlogo can be divided into two types. One is to describe the situation where the driver makes random choices in parking strips and parking spaces in the absence of parking guidance information, such as Figure 3 As shown, the other is to describe the choice made by the driver according to the discrete choice model calibrated above when there is parking guidance information, such as Figure 4 shown.

[0107] The input of the “Parking Lot Micro-Simulation Platform” is:

[0108] The actual total vehicle demand distribution, arrival time distribution and parking duration distribution of the parking lot.

[0109] The output of the “Parking Lot Micro-Simulation Platform” is:

[0110] Parking space occupancy and turnover are characterized by parking space parameters, including average time occupancy rate and average daily turnover rate;

[0111] The speed of the parker after entering the parking lot;

[0112] Expected travel time (the shortest travel time from the target parking space to the parking user's entrance);

[0113] Search time (the parker's actual travel time minus the expected travel time);

[0114] The walking distance from the parking location to the destination.

[0115] like Figure 4 As shown, "making a selection" refers to parking strip selection and parking space selection.

[0116] Parking strip selection includes the following steps:

[0117] Based on the parking occupancy information (the probability of the existence of valid vacant parking spaces) provided by the parking guidance information and the location attributes of the parking strip, the discrete choice model gives the parking strip selection probability.

[0118] Parking space selection includes the following steps:

[0119] The parking space selection probability is calculated based on the attribute information of the parking space and the driving distance to the parking space and the walking distance to the destination after parking.

[0120] The attribute information of the berth refers to the driving distance within the yard, walking time, and the probability of the existence of an available berth.

[0121] According to the actual situation of the underground parking lot of the gymnasium of Tongji University Siping Road Campus, there are 946 parking spaces in total, divided into two floors, of which 464 are on the first underground floor and 482 are on the second underground floor. The garage has three entrances and exits, and the interior adopts a streamlined design combining one-way large circulation and small circulation. Currently, only the first underground floor is officially put into operation. Therefore, Netlogo software is used to develop a parking simulation platform for the underground floor of the new gymnasium of Tongji University Siping Road Campus. Figure 5 shown.

[0122] The parking simulation platform includes four types of parking users: weekday commuters (registered vehicles), weekday visitors (unregistered vehicles), non-weekday commuters (registered vehicles), and non-weekday visitors (unregistered vehicles). The platform can select the parking user type based on whether the simulation is for a weekday or a non-weekday. During vehicle operation within the parking lot, based on the following vehicle model, if a parker is parking in or exiting a target parking space, following vehicles must stop and wait until they have fully entered or exited before passing through the target space.

[0123] We use the actual total vehicle demand distribution, arrival time distribution and parking duration distribution in the parking lot as the demand input of our parking lot simulation platform.

[0124] The simulation platform can output in real time the parking space occupancy and turnover status, the driving speed of the parker after entering the parking lot, the expected driving time (the driving time of the shortest path from the target parking space to the parker's entrance), the search time (the parker's actual driving time minus the expected driving time) and the walking distance after parking.

[0125] The underground garage of the newly built gymnasium on the Siping Road campus of Tongji University is equipped with a parking guidance system (existing technology). We used a simulation model based on parking behavior rules with the parking guidance system to simulate the parking space parameters (average time occupancy rate and average daily turnover rate) and compared the parking space parameters (average time occupancy rate and average daily turnover rate) output by the simulation with the actual detection data of the parking space video detector in the parking lot to verify the simulation effect of the parking lot micro-simulation platform. Figure 6 and Figure 7 shown.

[0126] The “parking behavior pattern” is obtained based on the discrete BL model finally established in step S2.

[0127] The “simulation model based on parking behavior rules with a parking guidance system” refers to the parking lot micro-simulation platform in step S3.

[0128] A t-test comparing the measured and simulated data on the average time occupancy of parking spaces showed that, at the 5% significance level, the two data sets can be considered to come from the same distribution, with sig = 0.8870. A t-test comparing the measured and simulated data on the average daily turnover rate of parking spaces also showed that, at the 5% significance level, the two data sets can be considered to come from the same distribution, with sig = 0.7067. Therefore, it can be concluded that the parking lot micro-simulation platform can effectively describe actual parking behavior and provide a scientific evaluation tool for optimizing parking lot operations and management.

[0129] S4. Construct a two-layer model for parking lot operation management optimization, use simulated annealing algorithm to solve it, and optimize the parking lot operation efficiency.

[0130] S401. Set the upper model so that its goal is to optimize social benefits, such as;

[0131]

[0132] st

[0133] g i (y,x(y))≤0,i=1,...,p1 (4-2)

[0134] h j (y,x(y))≤0,j=1,...,p2 (4-3)

[0135] where g i , h j They represent various external constraints respectively, and x(y) is the system parking allocation and operation result under any given project measure decision variable y, but it is taken from the parking flow simulation process of the study parking lot under the following project measure implementation scenario.

[0136] S402: Introduce the evaluation model based on microscopic traffic simulation (the parking lot microscopic simulation platform in step S3) into the lower model, and introduce the simulation model into the lower model of such a two-layer model.

[0137] In the agent-based simulation system, users can select user travel behavior criteria according to the needs of the model, such as the user utility maximization allocation method based on the results of the parking lot selection behavior survey.

[0138] The schematic diagram of the lower-level simulation model logic is as follows Figure 8 shown.

[0139] S403, solve the double-layer model through the simulated annealing algorithm to optimize the parking lot operation efficiency, and combine the double-layer optimization model with the simulated annealing algorithm to optimize the model process as follows: Figure 9 shown.

[0140] The newly constructed underground parking garage at the stadium on Tongji University's Siping Road campus has 946 parking spaces. During weekdays, it primarily serves two types of users: commuters and visitors. This often results in prime parking spaces being occupied by long-term commuters, leaving short-term visitors and other vehicles without suitable spaces.

[0141] External constraints include parking lot capacity constraints (the number of parking lots cannot be greater than the total number of parking spaces) and other conditions.

[0142] Therefore, by setting up parking time restriction areas, such as n partitions in a parking lot, among which m partitions are subject to parking time restrictions, the parking time shall not exceed t hours, with the goal of minimizing the total driving time and walking time of the system, and without demand overflow.

[0143] The optimization model is shown below:

[0144] Decision variables y,z:

[0145] There are i = 1, 2, 3, ..., n zones in the parking lot. From these zones, m = 1, ..., n zones are selected, and y = {y1, ..., ym} is used as the parking time limit zone.

[0146] z is the parking time limit parameter, z = {0.5, 1, 1.5, 2, 2.5, 3, 3.5, 4}, unit: hour.

[0147] Objective function F(x(y,z),(y,z)): The objective function of minimizing system costs is adopted, that is, the system's driving time and walking time costs are minimized. It is expressed in functional form as follows:

[0148]

[0149] in:

[0150] A—Partition collection within the parking lot;

[0151] v a —Parking flow in parking zone a is shown in Table 7;

[0152] b—parking lot users;

[0153] travel b — User b’s parking travel time; the travel time to each parking space is calculated based on the user’s entry point into the parking lot, the location of each parking space, and the speed of the user in the parking lot.

[0154] walk b — User b’s walking time after parking is calculated based on the location of each parking space, the location of the elevator, and the walking speed.

[0155] (1) Parking demand input

[0156] Overall distribution of arrival times: estimated probability ω k , expected value μ k and variance σ k As shown in Table 9 below:

[0157] Table 7 Fitting parameters of mixed Gaussian distribution of overall vehicle arrival time

[0158]

[0159] In Table 7, k is the number of driver behavior categories. For each category, k = 1,…,K, there is a set of parameters and representing its mean and variance.

[0160] Overall distribution of parking duration: estimated probability ω k , expected value μ k and variance σ k As shown in Table 10 below:

[0161] Table 8 Fitting parameters of mixed Gaussian distribution of overall parking time of vehicles

[0162]

[0163] In Table 8, k is the number of driver behavior categories. For each category, k = 1,…,K, there is a set of parameters and representing its mean and variance;

[0164] Daily vehicle arrivals refer to the number of vehicles arriving at the parking lot (or parking area) from 0:00 to 24:00 on the same day. This simulation uses the average value of 2077 vehicles per day on Monday.

[0165] (2) Parking lot network structure

[0166] The parking lot has 18 sections, such as Figure 10 As shown, each partition corresponds to 50 parking spaces, and each parking space corresponds to two types of simulation parameters in the simulation model, the shortest driving distance to each entrance and the shortest walking distance to each elevator entrance.

[0167] Figure 10 ABCD in the number represents elevators, and the numbers 1, 2, ..., 9 on both sides represent a parking partition, for a total of 18 parking partitions.

[0168] The simulated annealing optimization algorithm is used to solve the problem. In the optimization operation, the initial temperature T(0) = 1,000,000 and the cooling coefficient tfactor = 0.9 are selected.

[0169] The optimization results are shown in Table 11 below.

[0170] Table 9 Optimization results

[0171]

[0172] The optimized management strategy was set in a simulation model (parking lot micro-simulation platform), and the simulation results of the current management strategy were compared with those of the short-term parking area management strategy. It was found that after the short-term parking area was set, the average driving search time of vehicles in the parking lot was reduced by 14.29%, and the average walking distance was reduced by 22.29%. The optimization effect was obvious.

Claims

1. A multi-agent simulation analysis method for urban parking system operation optimization, characterized by: The method comprises the following steps: S1. Through questionnaire design and experiments based on willingness survey, we can identify the factors that significantly influence drivers' parking behavior. S2. Quantitatively analyze the factors influencing drivers’ parking choice behavior using the Binomial Logit discrete choice model; S3. Build a multi-agent parking lot micro-simulation platform using the Netlogo development environment; S4. Construct a two-layer model for parking lot operation management optimization, use simulated annealing algorithm to solve it, and optimize the parking lot operation efficiency.

2. The multi-agent simulation analysis method for urban parking system operation optimization according to claim 1 is characterized in that: The parking selection behavior includes three stages: target area selection stage, area search stage, and parking space evaluation stage. The influencing factors are divided into personal attribute factors and parking attribute influencing factors; Personal attribute factors include gender, age, driving experience, educational background, and income; Factors influencing parking attributes include walking distance after parking, parking feasibility, driving distance, parking safety, parking convenience, and parking cost.

3. The multi-agent simulation analysis method for urban parking system operation optimization according to claim 1 is characterized in that: The S1 specifically includes the following steps: S101. Determine the research question and purpose of the investigation; S102. Determine the influencing factors, eliminate factors with insignificant influence, and determine the level number and level value of relevant parameters based on the actual situation, survey purpose, and survey principles; S103, scenario design based on utility balance; S104. Complete the willingness survey questionnaire and conduct on-site questionnaire survey; S105. Statistically analyze the results of the parking choice behavior intention survey to obtain factors that significantly affect the drivers' parking choice behavior.

4. The multi-agent simulation analysis method for optimizing urban parking system operations according to claim 3, characterized in that: The statistical analysis of the results can be divided into statistical analysis of driver personal information, statistical analysis of driver parking behavior and statistical analysis of parking space selection tendency.

5. The multi-agent simulation analysis method for urban parking system operation optimization according to claim 1 is characterized in that: The S2 specifically includes the following steps: S201, using SPSS software to calibrate parameters of a discrete BL model of the driver's parking choice behavior in a parking lot; S202, bringing the driver's personal attributes and parking space attribute variables into the model, and performing model fit analysis using SPSS software; S203, using χ 2 Independence test was used to analyze the correlation between influencing factors and parking choice behavior.

6. The multi-agent simulation analysis method for optimizing urban parking system operations according to claim 5, characterized in that: If the results of the model fit analysis show that the model fit index value does not meet the required value, it is necessary to adjust the variables of the model accordingly, that is, remove the independent variables that are not significant in the model test, re-establish the model analysis, and repeatedly adjust and calibrate the model until the various model fit indices meet the requirements.

7. The multi-agent simulation analysis method for urban parking system operation optimization according to claim 1 is characterized in that: The simulation logic of the simulation platform includes a random selection method and a method based on a discrete selection model to maximize personal utility, and the relevant parameters of the simulation model are calibrated and the simulation effect is verified through measured data.

8. The multi-agent simulation analysis method for urban parking system operation optimization according to claim 1 is characterized in that: The simulation platform constructed by Netlogo can be divided into two types; One is to describe the random selection of parking strips and parking spaces by drivers in the absence of parking guidance information and in accordance with a certain probability distribution; The other is to describe the choices made by drivers based on the discrete choice model calibrated above when parking guidance information is available.

9. The multi-agent simulation analysis method for urban parking system operation optimization according to claim 1, characterized in that: The S4 specifically includes the following steps: S401. Set up the upper-level model so that its goal is to optimize social benefits; S402, introducing the parking lot micro-simulation platform into the lower model; S403. Solve the double-layer model using a simulated annealing algorithm to optimize parking lot operation efficiency.

Images