A method and system for estimating queue length at signalized intersections based on floating vehicle data

Abstract

This invention proposes a method and system for estimating queue lengths at signalized intersections based on floating vehicle data. The method first performs map matching on low-frequency floating vehicle data. Next, a relationship model is constructed between the periodic maximum queue length and the spatial distribution of floating vehicles. Based on historical floating vehicle data, the mean of the periodic maximum queue is estimated. Then, based on the historical floating vehicle data and intersection signal timing data, the periodic average maximum queue length of floating vehicle samples is extracted and the variance is obtained. Finally, the obtained mean and variance are used to define a prior distribution of the periodic maximum queue length. Based on the floating vehicle data observed during the current period, a Bayesian approach is used to estimate the posterior distribution of the periodic queue length during the current period. This invention utilizes low-frequency, low-penetration floating vehicle data, which is widely available in real urban road networks, to achieve widespread and applicable acquisition of queue status at signalized intersections in urban road networks, providing fundamental support for urban road network traffic management and control.

Description

Technical Field

[0001] The invention relates to a method and system for estimating queue length at a signal intersection based on floating vehicle data, and belongs to the technical field of transportation. Background Art

[0002] Queue length at signalized intersections is not only a core metric for evaluating urban road traffic performance but also a crucial input for traffic control applications such as optimizing traffic signal timing parameters. Accurately estimating queue length is crucial for optimizing traffic efficiency at signalized intersections and alleviating urban congestion.

[0003] The existing methods for estimating the queue length at signalized intersections on urban roads mainly include: (1) Deterministic methods based on traffic wave reconstruction: by identifying the key points where the traffic status of queued vehicles changes, the gathering wave and dissipation wave generated during the queuing process are reconstructed, and the maximum queue length is determined according to the position of the intersection of the two waves in space; (2) Stochastic methods based on probability statistics: the distribution of vehicles arriving at the intersection is pre-set, and the parameters in the hypothetical distribution are solved based on the observed data, and the queue length at the intersection is described according to the mean and variance of the probability distribution finally obtained; (3) Data-driven methods based on machine learning: by mining the traffic pattern information in a large amount of historical data of the intersection, machine learning technology is used to simulate the arrival pattern of vehicles at the intersection, traffic operation status and other information, and the queue length of the corresponding road is inferred.

[0004] Although the above methods have achieved good application results in specific data and traffic scenarios, they are still unable to meet the needs of widely obtaining the queue status of urban roads due to the constraints of the current traffic data available in urban road networks. Among them, method (1) needs to identify the key points where the traffic flow status changes. When the vehicle queue exceeds the detection area or the detector detection frequency is low, the traffic flow status will be difficult to identify; method (2) Most studies only consider the queue vehicle data and do not fully consider non-parked vehicles. The position of non-parked vehicles also implies the upper bound of the queue to a certain extent; method (3) requires a large amount of high-precision historical data to calibrate the model parameters, and has high requirements for data quality. When the data quality is poor, the estimation accuracy of the model will be difficult to guarantee. Therefore, the above three methods cannot fully meet the needs of queue length estimation under the conditions of low-frequency and low-penetration floating vehicle data.

[0005] Low-frequency, low-penetration floating vehicle data is a widely available data source in urban road networks, with a wide spatial distribution (e.g., floating vehicle data collected by taxis). This low frequency and low penetration rate is primarily characterized by a floating vehicle sampling frequency of 20 to 60 seconds and a penetration rate of less than 5%. How to leverage this low-frequency, low-penetration data to broadly and cost-effectively obtain information on urban road queuing status remains an unresolved challenge. Summary of the Invention

[0006] Purpose of the invention: In response to the above-mentioned deficiencies in the prior art, the present invention aims to provide a method and system for estimating the queue length at a signalized intersection based on low-frequency and low-penetration floating vehicle data. This method fully exploits the queue status information implicit in the historical floating vehicle data for both parked and non-parked vehicles, and combines it with the floating vehicle data observed in the current period to estimate the maximum queue mean and variance of the period, thus providing basic support for obtaining a large-scale picture of urban road traffic status.

[0007] Technical solution: To achieve the above-mentioned purpose, the technical solution adopted by the present invention is as follows:

[0008] A method for estimating queue length at a signalized intersection based on floating vehicle data comprises the following steps:

[0009] Perform map matching on the floating vehicle data to obtain the distance and direction from the floating vehicle trajectory matching point to the downstream signal intersection;

[0010] Based on traffic wave theory and probability statistics theory, a probability model of the maximum queue length and the spatial distribution of floating vehicles is constructed. The model parameters are calibrated to obtain the mean maximum queue length over the historical period.

[0011] Based on historical floating vehicle data and intersection signal timing data, the maximum queue length of the floating vehicle sample is extracted, and the variance of the maximum queue length in the historical period is obtained;

[0012] The mean and variance obtained from historical data are used to define the prior distribution of the maximum queue length of a period. A posterior estimation model of the maximum queue length of a period is constructed based on Bayesian theory, and the Markov Chain Monte Carlo (MCMC) method is used to solve the parameters of the posterior estimation model.

[0013] Preferably, the map matching of the floating vehicle data includes:

[0014] The floating vehicle data is screened to remove data outside the study area, duplicate data, and abnormal drift trajectory points. The screened floating vehicle data is then divided into multiple travel trajectories using the time threshold method.

[0015] The road network is represented as a topological graph with intersections as nodes and road sections as edges;

[0016] Based on a single floating vehicle travel trajectory, a map matching algorithm based on the hidden Markov model is used to obtain the trajectory points that match the road section.

[0017] Preferably, the probability model of the maximum queue length of the period and the spatial position distribution of floating vehicles is expressed as:

[0018]

[0019] Among them, f X (x) is the probability density of floating vehicle sampling at position x, L is the total length of the research section, l r is the remaining queue length under saturation, l max is the distance from the remaining queue length to the maximum queue length of the period, is the normalized arrival density, To normalize the increment of arrival density in the queue, the maximum queue length l = l max +l r .

[0020] As a preferred method, based on the historical floating vehicle spatial position distribution dataset on the study section, the maximum likelihood criterion was used and the simulated annealing algorithm was used to obtain the mean of the maximum queue length in the historical period. The objective function of the maximum likelihood estimation is expressed as:

[0021]

[0022] Among them, x0 is the historical floating vehicle spatial position distribution dataset x on the research section O The samples in .

[0023] Preferably, the method of extracting the maximum queue length of a floating vehicle sample based on historical floating vehicle data and intersection signal timing data, and obtaining the variance of the maximum queue length of the period within the historical period, includes:

[0024] The matched trajectory points are divided into intersections and entrances, and the floating vehicle data of each intersection and entrance is time-sliced. According to the timing information of the signalized intersection, the floating vehicle data within each slice time is divided into periods.

[0025] The floating vehicle data in each cycle is used to determine the vehicle state. Vehicles with a speed less than or equal to the preset threshold are considered to be in a parked state, and vice versa.

[0026] The trajectory point farthest from the downstream intersection and in a parked state in each cycle is extracted as the observation sample of the maximum queue length of the cycle;

[0027] Aggregate the observation samples of the maximum queue length in the same period of history, fit them according to the normal distribution, and use the obtained variance as the variance of the maximum queue length in the historical period.

[0028] Preferably, the a posteriori estimation model of the periodic maximum queue length is expressed as:

[0029]

[0030] Among them, l is the maximum queue length of the cycle, l ris the remaining queue length of the cycle, is the period-normalized arrival density, is the likelihood function, f L (l) l, l respectively r 、 The prior distribution of , x represents the sample data of the floating vehicle spatial position observation in the current analysis period.

[0031] Preferably, the prior distributions of the periodic maximum queue length and the remaining queue length are set to normal distribution, and the prior distribution of the normalized arrival density is set to truncated normal distribution.

[0032] Preferably, the Markov Chain Monte Carlo (MCMC) method is used to estimate the posterior distribution of the maximum queue length of the period, including: generating multiple posterior distribution samples using the observations in the current period and the prior distribution derived from historical data, and calculating the acceptance probability of each posterior distribution sample, stopping when the number generated meets the iteration condition, and using the final posterior distribution sample to estimate the posterior distribution, thereby realizing the estimation of the maximum queue distribution of the period.

[0033] Based on the same inventive concept, the present invention provides a signalized intersection queue length estimation system based on floating vehicle data, comprising:

[0034] The pre-processing module is used to perform map matching on the floating vehicle data and obtain the distance and direction from the floating vehicle trajectory matching point to the downstream signal intersection;

[0035] The prior information calculation module is used to construct a probability model of the maximum queue length and the spatial distribution of floating vehicles based on traffic wave theory and probability statistics theory, calibrate the model parameters, and obtain the mean of the maximum queue length over a historical period. Based on historical floating vehicle data and intersection signal timing data, it extracts the maximum queue length for floating vehicle samples and obtains the variance of the maximum queue length over a historical period.

[0036] And the posterior estimation module is used to define the prior distribution of the maximum queue length of the period based on the mean and variance obtained from historical data, construct a posterior estimation model of the maximum queue length of the period based on Bayesian theory, and use the MCMC method to solve the parameters of the posterior estimation model.

[0037] Based on the same inventive concept, the present invention provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the method for estimating the queue length of a signalized intersection based on floating vehicle data.

[0038] Beneficial Effects: From the perspective of historical patterns, this invention fully exploits the prior information in historical floating vehicle data, taking into account both parked and non-parked vehicles, and using the probability model method and the relationship between the sample and the population to obtain the mean and variance of the historical period maximum queue. From the perspective of real-time estimation, based on Bayesian theory, using the floating vehicle observation data in the current analysis period and the prior information obtained from historical floating vehicle data, a posterior distribution estimation model for the period maximum queue is constructed, enabling dynamic estimation of the vehicle queue status at the intersection. Compared with the existing technology, this invention has the following advantages:

[0039] 1. Compared with deterministic methods based on traffic wave reconstruction and data-driven methods based on machine learning, the proposed method can achieve periodic maximum queue mean and variance estimation under limited data conditions, providing technical support for obtaining urban road traffic status on a large scale.

[0040] 2. Compared with the random method based on probability statistics, the proposed method makes full use of the parking and non-parking vehicle information in the low-frequency and low-penetration floating vehicle data, thereby improving the accuracy and reliability of the intersection cycle maximum queue estimation. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is an overall flow chart of an embodiment of the present invention.

[0042] Figure 2 This is a flowchart of estimating the posterior distribution of the average maximum queue length of floating vehicles during the analysis period based on Bayesian theory in an embodiment of the present invention.

[0043] Figure 3 The spatiotemporal trajectory diagram of the vehicle in the undersaturated state and the saturated state.

[0044] Figure 4 This is a trend diagram of the queue length change at the north entrance of intersection 1 as an example in an embodiment of the present invention.

[0045] Figure 5 This is a trend diagram of the queue length change at the south entrance of intersection 2 as an example in an embodiment of the present invention.

[0046] Figure 6 This is a trend diagram of the queue length change at the west entrance of intersection 3 as an example in an embodiment of the present invention.

[0047] Figure 7 This is a trend diagram of the queue length change at the west entrance of intersection 4, as illustrated in an embodiment of the present invention.

[0048] Figure 8 This is a comparison diagram of the estimated distribution and the actual distribution of the period-averaged maximum queue length used as an example in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The technical solution of the invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0050] The relevant terms involved in this invention are explained as follows: Floating vehicle data: refers to data such as location, time, vehicle speed, and driving direction collected by vehicles equipped with satellite positioning modules. Undersaturated state: refers to a traffic state in which the queue of vehicles at an intersection can be completely dissipated during a single green light. Saturated state: refers to a traffic state in which the queue of vehicles at an intersection cannot be completely dissipated during a single green light. Before passing through the intersection, some vehicles must queue twice or more times. Residual queue: refers to the queue of vehicles that has not been dispersed at the end of the current green light.

[0051] like Figure 1 As shown, an embodiment of the present invention discloses a method for estimating queue length at a signalized intersection based on low-frequency and low-penetration floating vehicle data, comprising the following steps:

[0052] S1: Perform map matching on the floating vehicle data to obtain the distance and direction from the floating vehicle trajectory matching point to the downstream signal intersection.

[0053] S2: Based on traffic wave theory and probability statistics theory, a probability model of the maximum queue length in a period and the spatial distribution of floating vehicles is constructed, the model parameters are calibrated, and the mean maximum queue length in a period within the historical period is obtained.

[0054] S3: Based on the historical floating vehicle data and the intersection signal timing data, the maximum queue length of the floating vehicle sample is extracted, and the variance of the maximum queue length of the period in the historical period is obtained.

[0055] S4: Using the mean and variance obtained from historical data, define the prior distribution of the maximum queue length of a period, construct a posterior estimation model of the maximum queue length of a period based on Bayesian theory, and use the MCMC method to solve the model parameters.

[0056] For example, in step S1, the path-level map matching algorithm based on the hidden Markov model is used to implement map matching of low-frequency floating vehicle data. The specific process is as follows:

[0057] S11: Eliminate non-compliant floating vehicle data, including data outside the study area, duplicate data (multiple data points generated by the same vehicle at the same time), and abnormal drift trajectory points. Abnormal drift trajectory points can be eliminated using speed thresholds, distance thresholds, and angle thresholds. The filtered floating vehicle data is then segmented into multiple travel trajectories using a time threshold, typically 25-30 minutes.

[0058] S12: The road network is represented as a topological graph consisting of intersections (nodes) and road sections. The topological requirement is that each road section should connect two adjacent nodes to form a complete network structure.

[0059] S13: Based on a single floating vehicle travel trajectory, a map matching algorithm based on a hidden Markov model is used to obtain trajectory points that match road segments. First, all possible candidate road locations near each trajectory point on the single travel trajectory are obtained. Second, the transition probabilities between candidate locations are calculated based on road connectivity and trajectory point motion. Then, the emission probability of each trajectory point at a candidate location is calculated based on the distance from the trajectory point to the candidate location. Finally, the path probability matrix is ​​continuously updated by combining the transition and emission probabilities. Dynamic programming is used to find the optimal path and output it. This process is repeated for each subsequent travel trajectory.

[0060] S14: Based on the matched floating vehicle trajectory points and the distribution of signalized intersections along the path, the distance from each trajectory point to the downstream signalized intersection and the turn information after passing the intersection are extracted. Distance calculation uses the Manhattan distance method, and turn determination uses the vector method.

[0061] For example, the specific process of step S2 is as follows:

[0062] S21: Combining traffic wave theory with probability statistics theory, the probability model of the maximum queue length and the spatial distribution of floating vehicles is derived. The traffic states of signalized intersections corresponding to undersaturated and saturated states are as follows: Figure 3 As shown in the figure, v a is the traffic wave velocity generated when the upstream arriving traffic flow changes from the moving state to the stopped state, w is the traffic wave velocity generated when the queuing traffic flow changes from the stopped state to the saturated traffic state, R, C, and τ represent the red light duration, cycle duration, and the time required to clear the queue during the green light period, respectively. The area within the triangle and polygon represents the stopped queue state, and the other areas are the free-flow driving state.

[0063] (1) The average traffic density at any location on a road section within a unit cycle can be expressed as:

[0064] ρ i =ρ max ,i=0

[0065] ρ i =ρ c ,i=1

[0066] ρ i =ρ a ,i=2

[0067]

[0068] Where d(x) is the average traffic density at the stop line x from the downstream intersection, ρ i is the density of the traffic flow at position x when it is in state i, i = 0 means the vehicles begin to queue, i = 1 means the vehicle queue begins to dissipate, i = 2 means the vehicles have not yet entered the queue area, ρ max is the blocking density, ρ c is the critical density, ρ a is the arrival density, t i Indicates the time the vehicle is in state i at position x.

[0069] (2) The probability density of floating vehicle sampling at any position on the road section within a unit period:

[0070]

[0071] Among them, f X (x) is the probability density of floating vehicle sampling at position x, and L is the total length of the study section.

[0072] (3) The probability density of the floating car sampling at position x in the undersaturated state is:

[0073]

[0074]

[0075] Among them, Z0 is the normalization coefficient, l max It is the maximum queue length of the period under the undersaturated state.

[0076] (4) The probability density of the floating car sampling at position x in the saturated state is:

[0077]

[0078] Among them, Z1 is the normalization coefficient, l r is the remaining queue length under saturation.

[0079] (5) General model construction

[0080] Since the undersaturated state is a special saturated state, the remaining queue length l r The saturated state is when the vehicle density is 0, so the spatial distribution of vehicles in the saturated state is considered as the general case. In the derivation of the formulas for the above two states, it is assumed that the average density of traffic in the queue area increases linearly, so a general model can be constructed that is applicable to the above two states.

[0081]

[0082] in, is the normalized arrival density, is the increment of normalized arrival density in the queue.

[0083] By calibrating the parameters of the distribution, we can get the maximum queue length l, that is, l = l max +l r .

[0084] S22: Calibrate the model parameters and obtain the mean of the maximum queue length in the historical period. This embodiment aggregates the floating vehicle data in the same historical period, uses the maximum likelihood as the criterion, and adopts the simulated annealing algorithm to obtain the mean of the maximum queue length in the historical period. The objective function of the maximum likelihood estimation is as follows:

[0085]

[0086] Among them, x0 is a sample in the historical floating vehicle spatial position distribution dataset on the research section, x O This is a dataset of the spatial location distribution of historical floating vehicles on the research road section.

[0087]

[0088] For example, the specific process of step S3 is as follows:

[0089] S31: Based on the matched floating vehicle trajectory points, period-divided floating vehicle data is extracted according to the intersection signal cycle. First, the matched trajectory points are divided by intersection and entrance lane. Second, the floating vehicle data for each intersection and entrance lane is time-sliced, with a time window of 15 minutes. Finally, the floating vehicle data within each slice time is period-divided according to the signalized intersection timing information.

[0090] S32: Vehicle status is determined for the floating vehicle data in each cycle. In this embodiment, a speed threshold is set to 5 m / s. Vehicles with a speed less than or equal to the threshold are in a parked state, and vice versa.

[0091] S33: Extract the trajectory point in each cycle that is farthest from the downstream intersection and in a parked state as the observation sample of the maximum queue length of the cycle.

[0092] S34: Gather observation samples of the maximum queue length during the same period in history, fit them according to the normal distribution, and use the obtained variance as the variance of the maximum queue length during the historical period.

[0093] Exemplarily, the specific process of step S4 is as follows:

[0094] S41: Setting the prior distribution of probability distribution model parameters, including the maximum queue length l of the period, the remaining queue length l of the period r , period normalized arrival density The prior distributions of the first two parameters are set to normal distribution, and the prior distribution of the last parameter is set to truncated normal distribution, as follows:

[0095] l~N(l prior ,σ 2 )

[0096] Among them, l prior is the mean of the maximum queue length obtained by calibrating the simulated annealing algorithm in S2, σ 2 is the variance of the maximum queue length distribution of floating vehicles.

[0097] l r ~N(l r,prior ,10 2 )

[0098] Among them, l r,prior It is the cycle remaining queue length obtained by calibrating the simulated annealing algorithm in S2.

[0099]

[0100] in, is the period-normalized arrival density obtained by calibrating the simulated annealing algorithm in S2, L road To study the road length, m is the numerical input of the period-normalized arrival density, represents the probability density function of the standard normal distribution, Represents the cumulative distribution function of the standard normal distribution. The mean and standard deviation of the normal distribution before standardization are and 1.

[0101] S42: According to Bayesian theory, the posterior distribution estimation model of the maximum queue period is as follows:

[0102]

[0103] Among them, l is the maximum queue length of the cycle, l r is the remaining queue length of the cycle, is the period-normalized arrival density, is the likelihood function, f L (l) l, l respectively r 、 The prior distribution of , x is the sample data of the floating vehicle spatial position observation in the current analysis period.

[0104] S43: Based on the floating car data observed in the current period, the Markov Chain Monte Carlo (MCMC) method is used to estimate the posterior distribution of the maximum queue length of the period. The formula for solving the complex posterior integral using the MCMC method is as follows:

[0105]

[0106] Among them, g(D|x) is the likelihood function in Bayesian theory, p(x) is the probability distribution of x in the interval [a, b] (the prior distribution of the parameter), and x t is the sample generated on the Markov chain according to p(x), n is the total number of samples generated, and D is the observed sample data.

[0107] In calculation When Consider it as g(D|x), and f L (l) Treat it as p(x) and solve it according to the above formula.

[0108] This method uses the observations in the current period and the prior distribution derived from historical data to generate multiple posterior distribution samples, and calculates the acceptance probability of each posterior distribution sample. When the number generated meets the iteration condition, it can stop and use the final posterior distribution sample to estimate the posterior distribution, thereby realizing the estimation of the maximum queuing distribution of the period.

[0109] Case Study

[0110] (1) This paper selects floating vehicle data from a central urban area of ​​a certain city as a method example scenario, uses a 15-minute time window length, and studies the queue length of through traffic in the central urban area under typical scenarios for nine days from May 28, 2018 to June 8, 2018. The research objects are as follows:

[0111] Table 1 Study subjects

[0112]

[0113] (2) The method proposed by the present invention is used to visualize the estimated historical maximum queue length and the mean of the remaining queue length, as shown in the following example: Figures 4 to 7 shown.

[0114] (3) The method proposed in this invention is used to visualize the estimated maximum queue length posterior distribution, as shown in Figure 8 As shown in the figure, the MCMC method parameters are set as follows: number of MCMC chains: 4, number of formal samples: 8,000, number of burn-in samples: 2,000, and sample acceptance probability: 0.95. Specifically, the maximum queue lengths for through vehicles at the north entrance of Intersection 1 from 8:30 AM to 8:45 AM, the maximum queue lengths for through vehicles at the south entrance of Intersection 2 from 11:45 AM to 12:00 PM, and the maximum queue lengths for through vehicles at the west entrance of Intersection 3 from 8:45 AM to 9:00 AM.

[0115] By the attached Figure 4 、5 , 6, and 7 show that the estimated change trend of the historical mean of the intersection queue length under different traffic scenarios shows that the mean queue length during peak hours first increases and then decreases, while the mean queue length during off-peak hours fluctuates within a certain range and shows a stable trend. Both are in line with the objective law of change. Figure 8 It can be seen that the estimated distribution and the true distribution of the cycle average maximum queue length obtained by the present invention can be considered to be derived from the same distribution within a certain confidence range, and the estimation result can reflect the actual situation of vehicle queues at the intersection to a certain extent.

[0116] Based on the same inventive concept, an embodiment of the present invention discloses a signalized intersection queue length estimation system based on floating vehicle data, comprising: a preprocessing module for performing map matching on the floating vehicle data to obtain the distance and direction from the floating vehicle trajectory matching point to the downstream signalized intersection; a priori information calculation module for constructing a probability model of the periodic maximum queue length and the spatial position distribution of the floating vehicle based on traffic wave theory and probability statistics theory, calibrating the model parameters, and obtaining the mean of the periodic maximum queue length within a historical period; and, based on the historical floating vehicle data and according to the intersection signal timing data, extracting the periodic maximum queue length of the floating vehicle samples and obtaining the variance of the periodic maximum queue length within the historical period; and a posterior estimation module for defining the prior distribution of the periodic maximum queue length using the mean and variance obtained from the historical data, constructing a posterior estimation model for the periodic maximum queue length based on Bayesian theory, and solving the posterior estimation model parameters using the MCMC method.

[0117] Based on the same inventive concept, an embodiment of the present invention discloses a computer program product, including a computer program / instruction. When the computer program / instruction is executed by a processor, the steps of the method for estimating the queue length of a signalized intersection based on floating vehicle data are implemented.

Claims

1. A method for estimating queue length at a signalized intersection based on floating vehicle data, characterized in that: The steps include: Perform map matching on the floating vehicle data to obtain the distance and direction from the floating vehicle trajectory matching point to the downstream signal intersection; Based on traffic wave theory and probability statistics theory, a probability model of the maximum queue length and the spatial distribution of floating vehicles is constructed. The model parameters are calibrated to obtain the mean of the maximum queue length in the historical period. The probability model of the maximum queue length and the spatial distribution of floating vehicles is expressed as: Among them, f X (x) is the probability density of floating vehicle sampling at position x, L is the total length of the research section, l r is the remaining queue length under saturation, l max is the distance from the remaining queue length to the maximum queue length of the period, is the normalized arrival density, To normalize the increment of arrival density in the queue, the maximum queue length l = l max +l r ; Based on historical floating vehicle data and intersection signal timing data, the maximum queue length of the floating vehicle sample is extracted, and the variance of the maximum queue length in the historical period is obtained; The mean and variance obtained from historical data are used to define the prior distribution of the maximum queue length of a period. A posterior estimation model of the maximum queue length of a period is constructed based on Bayesian theory, and the Markov Chain Monte Carlo (MCMC) method is used to solve the parameters of the posterior estimation model.

2. The method for estimating queue length at a signalized intersection based on floating vehicle data according to claim 1, characterized in that: The map matching of the floating vehicle data includes: The floating vehicle data is screened to remove data outside the study area, duplicate data, and abnormal drift trajectory points. The screened floating vehicle data is then divided into multiple travel trajectories using the time threshold method. The road network is represented as a topological graph with intersections as nodes and road sections as edges; Based on a single floating vehicle travel trajectory, a map matching algorithm based on the hidden Markov model is used to obtain the trajectory points that match the road section.

3. The method for estimating queue length at a signalized intersection based on floating vehicle data according to claim 1, characterized in that: Based on the historical floating vehicle spatial position distribution dataset on the research section, the maximum likelihood criterion was used and the simulated annealing algorithm was adopted to obtain the mean of the maximum queue length in the historical period. The objective function of maximum likelihood estimation is expressed as: Among them, x0 is the historical floating vehicle spatial position distribution dataset x on the research section O The samples in .

4. The method for estimating queue length at a signalized intersection based on floating vehicle data according to claim 1, characterized in that: The method of extracting the maximum queue length of a floating vehicle sample based on historical floating vehicle data and intersection signal timing data, and obtaining the variance of the maximum queue length of the period within the historical period, includes: The matched trajectory points are divided into intersections and entrances, and the floating vehicle data of each intersection and entrance is time-sliced. According to the timing information of the signalized intersection, the floating vehicle data within each slice time is divided into periods. The floating vehicle data in each cycle is used to determine the vehicle state. Vehicles with a speed less than or equal to the preset threshold are considered to be in a parked state, and vice versa. The trajectory point farthest from the downstream intersection and in a parked state in each cycle is extracted as the observation sample of the maximum queue length of the cycle; Aggregate the observation samples of the maximum queue length in the same period of history, fit them according to the normal distribution, and use the obtained variance as the variance of the maximum queue length in the historical period.

5. The method for estimating queue length at a signalized intersection based on floating vehicle data according to claim 1, characterized in that: The a posteriori estimation model of the maximum queue length of the period is expressed as: Among them, l is the maximum queue length of the cycle, l r is the remaining queue length of the cycle, is the period-normalized arrival density, is the likelihood function, f L (l) l, l respectively r 、 The prior distribution of , x represents the sample data of the floating vehicle spatial position observation in the current analysis period.

6. The method for estimating queue length at a signalized intersection based on floating vehicle data according to claim 5, characterized in that: The prior distributions of the maximum queue length and the remaining queue length in a period are set to normal distribution, and the prior distribution of the normalized arrival density is set to truncated normal distribution.

7. The method for estimating queue length at a signalized intersection based on floating vehicle data according to claim 5, characterized in that: The Markov Chain Monte Carlo (MCMC) method is used to estimate the posterior distribution of the maximum queue length in a period, including: generating multiple posterior distribution samples using the observations in the current period and the prior distribution derived from historical data, and calculating the acceptance probability of each posterior distribution sample. The method stops when the number of generated samples meets the iteration condition, and uses the final posterior distribution sample to estimate the posterior distribution, thereby achieving the estimation of the maximum queue length in the period.

8. A signal intersection queue length estimation system based on floating vehicle data, characterized in that: include: The pre-processing module is used to perform map matching on the floating vehicle data and obtain the distance and direction from the floating vehicle trajectory matching point to the downstream signal intersection; The prior information calculation module is used to construct a probability model of the maximum queue length and the spatial distribution of floating vehicles based on traffic wave theory and probability statistics theory, calibrate the model parameters, and obtain the average maximum queue length over the historical period; Based on historical floating vehicle data and intersection signal timing data, the maximum queue length of the floating vehicle sample is extracted, and the variance of the maximum queue length in the historical period is obtained. The probability model of the maximum queue length and the spatial distribution of floating vehicles is expressed as: Among them, f X (x) is the probability density of floating vehicle sampling at position x, L is the total length of the research section, l r is the remaining queue length under saturation, l max is the distance from the remaining queue length to the maximum queue length of the period, is the normalized arrival density, To normalize the increment of arrival density in the queue, the maximum queue length l = l max +l r ; And the posterior estimation module is used to define the prior distribution of the maximum queue length of the period based on the mean and variance obtained from historical data, construct a posterior estimation model of the maximum queue length of the period based on Bayesian theory, and use the MCMC method to solve the parameters of the posterior estimation model.

9. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the steps of the method for estimating the queue length of a signalized intersection based on floating vehicle data according to any one of claims 1 to 7 are implemented.

Images